CN117639107B

CN117639107B - Power quality assessment method and system for power distribution network

Info

Publication number: CN117639107B
Application number: CN202311742527.7A
Authority: CN
Inventors: 张强; 王晶; 马楠楠
Original assignee: Qingdao Texaron Electric Co ltd
Current assignee: Qingdao Texaron Electric Co ltd
Priority date: 2023-12-18
Filing date: 2023-12-18
Publication date: 2024-05-17
Anticipated expiration: 2043-12-18
Also published as: CN117639107A

Abstract

The application provides a power quality assessment method and a system for a power distribution network, wherein the method comprises the following steps: acquiring initial electric energy data to be evaluated in a target power distribution network; performing incremental interpolation processing on the initial electric energy data to obtain first electric energy data; performing feature extraction on the first electric energy data by adopting a power distribution network feature analysis model to obtain electric energy time domain features and/or electric energy frequency domain features; the power time domain features include at least: waveform distortion values of the voltage or the current; the electric energy frequency domain features include at least: harmonic content characteristics, frequency offset characteristics, and power spectrum characteristics; and predicting the electric energy time domain characteristics and/or the electric energy frequency domain characteristics through an electric energy quality evaluation model to obtain an electric energy quality evaluation result of the target power distribution network. The method can obtain the power quality evaluation result of the power distribution network, provides an important reference for scheduling power transmission resources in the power distribution network, is beneficial to improving the stability of the power distribution network and improves the energy utilization efficiency.

Description

Power quality assessment method and system for power distribution network

Technical Field

The application belongs to the field of electric energy evaluation, and particularly relates to an electric energy quality evaluation method and an electric energy quality evaluation system for a power distribution network.

Background

In recent years, rapid economic development and population growth have led to a significant increase in domestic, commercial and industrial electricity. In order to balance the power supply requirements and reduce carbon emissions, the development of electric power technology is receiving close attention.

The distribution network (Distribution network) refers to a power system network that receives a portion of power from the power transmission network (Transmission network) and delivers the power to the end user. With the development of renewable energy sources and distributed energy sources, the reasonable allocation of power distribution network resources becomes a fundamental problem for the power distribution network. And one premise of reasonable allocation is that the power quality evaluation of the power distribution network is accurately evaluated, and the power transmission resources in the power distribution network can be reasonably scheduled based on actual demands through accurate evaluation results, so that the stability of the power distribution network is improved, and the power loss problem caused by fluctuation of the power distribution network is reduced, even the power safety problem is solved.

However, in the related art, the coverage of data collection and monitoring is limited, and the whole power distribution network cannot be covered completely, so that a part of nodes may not obtain accurate power quality data, and accuracy of the evaluation result of the power distribution network is affected.

In summary, a technical solution is needed to overcome the above technical problems in the related art.

Disclosure of Invention

The application provides a power quality evaluation method and a power quality evaluation system for a power distribution network, which are used for realizing power quality evaluation of the power distribution network, and are beneficial to improving the stability of the power distribution network and improving the energy utilization efficiency.

In a first aspect, the present application provides a method for evaluating power quality of a power distribution network, the method comprising:

acquiring initial electric energy data to be evaluated in a target power distribution network;

performing incremental interpolation processing on the initial electric energy data to obtain first electric energy data;

Performing feature extraction on the first electric energy data by adopting a power distribution network feature analysis model to obtain electric energy time domain features and/or electric energy frequency domain features; the electrical energy time domain features include at least: waveform distortion values of the voltage or the current; the electric energy frequency domain features at least comprise: harmonic content characteristics, frequency offset characteristics, and power spectrum characteristics;

And predicting the electric energy time domain characteristics and/or the electric energy frequency domain characteristics through an electric energy quality evaluation model to obtain an electric energy quality evaluation result of the target power distribution network.

In a second aspect, an embodiment of the present application provides a power quality evaluation system for a power distribution network, including:

The acquisition module is used for acquiring initial electric energy data to be evaluated in the target power distribution network;

The supplementing module is used for performing incremental interpolation processing on the initial electric energy data so as to obtain first electric energy data;

The extraction module is used for carrying out feature extraction on the first electric energy data by adopting a power distribution network feature analysis model so as to obtain electric energy time domain features and/or electric energy frequency domain features; the electrical energy time domain features include at least: waveform distortion values of the voltage or the current; the electric energy frequency domain features at least comprise: harmonic content characteristics, frequency offset characteristics, and power spectrum characteristics;

and the prediction module is used for predicting the electric energy time domain characteristics and/or the electric energy frequency domain characteristics through an electric energy quality evaluation model so as to obtain an electric energy quality evaluation result of the target power distribution network.

According to the technical scheme provided by the embodiment of the application, initial electric energy data to be evaluated in the target power distribution network are obtained. By acquiring initial electric energy data to be evaluated, subsequent processing and evaluation can be performed based on actual electric energy conditions, and accurate basic data can be provided for electric energy quality evaluation. Then, incremental interpolation processing is performed on the initial power data to obtain first power data. Compared with the related art, the method can fill partial data loss and correct partial data errors and deviations by performing incremental interpolation processing on the initial electric energy data. And then, performing feature extraction on the first electric energy data by adopting a power distribution network feature analysis model to obtain electric energy time domain features and/or electric energy frequency domain features. Here, the electrical energy time domain features include at least: waveform distortion values of the voltage or the current; the electric energy frequency domain features include at least: harmonic content characteristics, frequency offset characteristics, power spectrum characteristics. And the characteristic analysis model of the power distribution network is adopted to perform characteristic extraction on the first electric energy data, so that characteristic information of the electric energy data can be obtained from two aspects of time domain and frequency domain. The power time domain features may reflect the degree of waveform distortion of the voltage or current, while the power frequency domain features include harmonic content features, frequency offset features, and power spectrum features, which can describe the state of the power quality. And finally, predicting the electric energy time domain characteristics and/or the electric energy frequency domain characteristics through an electric energy quality evaluation model to obtain an electric energy quality evaluation result of the target power distribution network. And predicting the time domain features and the frequency domain features of the electric energy through the electric energy quality evaluation model, so that an electric energy quality evaluation result of the target power distribution network can be obtained. The result can assist in dispatching and operation of a power system, provide important reference information, and is beneficial to improving stability of a power distribution network and improving energy utilization efficiency. According to the embodiment of the application, the power quality evaluation of the power distribution network can be realized, and a data basis is provided for the scheduling of power transmission resources in the power distribution network.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute a limitation on the application. In the drawings:

fig. 1 is a schematic flow chart of a method for evaluating power quality of a power distribution network according to an embodiment of the present application;

Fig. 2 is a schematic diagram of a power quality evaluation method of a power distribution network according to an embodiment of the present application;

Fig. 3 is a schematic diagram of a power quality evaluation method of a power distribution network according to an embodiment of the present application;

FIG. 4 is a schematic diagram of a power quality assessment system for a power distribution network according to an embodiment of the present application;

fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the description of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

In order to solve at least one of the above technical problems, the embodiment of the application provides a power quality evaluation scheme for a power distribution network.

The power quality evaluation scheme of the power distribution network provided by the embodiment of the application can be executed by an electronic device, and the electronic device can be a server, a server cluster and a cloud server. The electronic device may also be a terminal device such as a mobile phone, a computer, a tablet, a wearable device, or a dedicated device (e.g. a dedicated terminal device with a distribution network management system, etc.). In an alternative embodiment, the electronic device may have installed thereon a service program for performing the power quality assessment scheme of the power distribution network.

Fig. 1 is a flowchart of a power quality evaluation method of a power distribution network according to an embodiment of the present application, where, as shown in fig. 1, the method includes:

101. and acquiring initial electric energy data to be evaluated in the target power distribution network.

Where a distribution grid refers to a power system network that receives a portion of power from a transmission grid and delivers the power to an end user. The system is the last section of power supply in the power system and is responsible for reducing, distributing and transmitting the electric energy transmitted by the high-voltage power transmission network so as to meet the power requirements of different users. In the distribution network, the electric energy is reduced to a suitable voltage level (such as low voltage, medium voltage, etc.) through a transformer substation, and is transmitted and distributed through equipment such as cables, wires, transformers, etc., and finally supplied to various end users such as families, commercial buildings, industrial enterprises, etc. The power distribution network mainly comprises the following two main parts:

High voltage distribution network (High Voltage Distribution Network): the high-voltage distribution network is used for reducing the high-voltage electric energy transmitted by the power transmission network to medium voltage or low voltage, and then transmitting the electric energy to areas such as a central business area, an industrial park and the like through the power transformation and distribution station. High voltage distribution networks typically use higher voltage levels (e.g., 10 kV, 35, kV, etc.).

Low voltage distribution network (Low Voltage Distribution Network): the low voltage distribution network further steps down the power from the substation to the low voltage level required by the user (e.g., 220V, 380V, etc.), and delivers it to the end users of the home, commercial building, industrial enterprise, etc. via cables, wires, etc. The low voltage distribution network is mainly responsible for delivering electrical energy to consumers and providing a stable supply of electrical energy.

The stability of the power distribution network in the operation process is considered in the design and the operation of the power distribution network, so that the power distribution network can ensure stable supply of electric energy and ensure that the electricity demand of a user is met. Various devices and technologies are used in distribution networks, such as substations, transformers, distribution cabinets, protection devices, etc., to ensure power quality, voltage stability and power reliability. Meanwhile, with the development of renewable energy sources and distributed energy sources, the power distribution network also faces new challenges and opportunities for improvement to adapt to the change and demand of future energy systems.

Aiming at the electric energy quality assessment requirement, initial electric energy data to be assessed needs to be acquired first, and a data base is provided for subsequent processing steps. In an embodiment of the present application, the initial power data includes, but is not limited to: voltage data, current data, frequency data, harmonic data, transient data, asymmetry data.

The voltage data includes voltage data of each node in the power supply system, such as node voltage amplitude, phase angle and the like. The voltage is one of important parameters for power quality evaluation, and can be used for evaluating the conditions of stability, voltage deviation, voltage fluctuation and the like of a power supply system. The current data includes current data for each node in the power supply system, such as node current magnitude, phase angle, etc. The current data may be used to evaluate power quality problems in terms of load conditions, current imbalance, harmonic content, etc. of the power supply system. The frequency data includes a change in the grid frequency in the power supply system. The deviation of the power grid frequency can be used for indicating the stability degree of the power supply system, and the intensity degree of frequency change can be used for evaluating the power quality of the power supply system. Transient data includes transient conditions occurring in the power supply system, such as voltage transients, current transients, etc. Transients are a power quality problem and information such as the magnitude, duration, etc. of the transients can help assess the reliability and stability of the power supply system. The asymmetry data includes an asymmetry of the voltage and current in the power supply system. The asymmetry may cause problems such as unbalance of the power grid, overheating of equipment, increase of power loss, etc., so that the asymmetry needs to be evaluated.

In practical applications, these initial electrical energy data may be acquired by several means:

Firstly, the data of electric energy parameters such as voltage, current, power and the like can be monitored and recorded in real time by installing an electric energy instrument at a key node of the power distribution network. These meters can provide accurate power data as an initial source of data for evaluation.

Second, the data acquisition system can be adopted to acquire the electric energy data of each node of the power distribution network in real time. The data acquisition systems transmit the measurement results of the electric energy meters to a data processing center for storage and analysis to obtain initial electric energy data to be evaluated.

Thirdly, the electric energy parameters can be monitored in real time through intelligent power grid sensors arranged at key positions of transformers, switch equipment and the like, and data are transmitted to a central control system. And through the intelligent power grid sensor, the electric energy data of each node in the power distribution network can be obtained and used as an initial data source for evaluation.

Fourth, the remote monitoring platform may interact with devices in the distribution network via telecommunication technology. Related electrical energy data may be obtained through communication with the distribution network equipment and used to evaluate the electrical energy quality of the distribution network.

In summary, the method for acquiring the initial electric energy data to be evaluated in the target power distribution network can utilize various modes such as electric energy meter data, a data acquisition system, a smart grid sensor, a remote monitoring platform and the like to monitor and record the data of the electric energy parameters in real time, so that accurate initial data is provided for subsequent electric energy quality evaluation.

By acquiring the initial electrical energy data to be evaluated in step 101, subsequent processing and evaluation can be performed based on actual electrical energy conditions, and accurate basic data can be provided for electrical energy quality evaluation.

102. And performing incremental interpolation processing on the initial power data to obtain first power data.

Because the coverage of data collection and monitoring may be limited in the related art, and the whole power distribution network cannot be covered entirely, certain important nodes may not be able to obtain accurate power quality data, so that accuracy and comprehensiveness of subsequently obtained evaluation results are limited.

In order to solve the problem, in the embodiment of the application, the initial electric energy data is subjected to incremental interpolation processing to fill partial data loss, correct partial data errors and deviations, and improve the problem of data loss in the related technology.

In practical application, the electric energy data of the missing partial nodes can be supplemented by adopting modes such as linear interpolation, polynomial interpolation, K-nearest neighbor interpolation (K-Nearest Neighbors Interpolation, KNN) interpolation and the like. The electric energy data of the missing important nodes can be filled up based on a model method (such as a regression model, a time sequence model and the like) so as to improve the integrity and the reliability of the electric energy data.

As an alternative embodiment, step 102 may be implemented using KNN interpolation. Specifically, determining missing power data points to be filled from the initial power data; calculating a distance between the missing power data point and a power data point existing in the initial power data; and selecting K nearest adjacent electric energy data points to interpolate the missing electric energy data points so as to obtain first electric energy data.

Specifically, first, the K value in KNN interpolation needs to be determined. Here, selecting an appropriate K value may determine the number of neighboring data points used to fill in the missing value during interpolation. The selection of the value of K may be based on experience, domain knowledge, or by cross-validation, etc. Further, it is necessary to calculate the distance between the missing power data point and other initial power data. In the embodiment of the application, euclidean distance, manhattan distance and the like can be used as distance measurement. Further, K power data points nearest to the distance between the missing power data points are determined based on the calculated distances. Finally, interpolation is performed on the missing power data points using numerical information of the K nearest power data points, including but not limited to: mean interpolation processing, or weighted average interpolation processing.

Further alternatively, the interpolation process in KNN interpolation may be expressed as the following formula:

；

where h { y } is the interpolation of the missing power data points, y_i is the numerical information of the K nearest neighboring power data points, frac {1} { K } represents the averaging operation on the K nearest neighboring power data points, sum { i=1 } { K } represents the summing operation on the K nearest neighboring power data points, and i is a numerical value greater than 1 and less than K. Frac { } { } is a fractional form representation in mathematics, representing the form of a numerator and a denominator, and this symbol is typically used to represent the ratio or fraction of two numbers.

The missing electric energy data points can be obtained through the numerical information supplement of the K nearest adjacent electric energy data points through the formula. The K value here represents the number of selected nearest neighbor power data points, and can be adjusted according to actual needs. The KNN interpolation method is an interpolation method based on nearest adjacent electric energy data points, and certain continuity of the data characteristic space needs to be ensured when the method is applied.

It should be noted that outliers may also have an effect on the interpolation result, so that other methods are needed to handle the effect of outliers when interpolation is performed using KNN interpolation.

For example, assuming that there is power data of a power distribution network, the power consumption of each node is recorded. These nodes include sensors, measuring instruments, etc. that are distributed throughout the distribution network at various locations. Assuming that there is a missing electrical energy data point for a node in the distribution network, the missing data can be filled in by KNN interpolation.

First, the K value in KNN interpolation is determined. Let k=5 be chosen, i.e. 5 data points in the neighborhood are chosen for interpolation. The choice of this K value may be determined empirically or by cross-validation. Next, the distance between the missing power data point and the other initial power data points needs to be calculated. Euclidean distance or manhattan distance may be used as the distance measure. The assumed missing data point is node a, which is at the following distance from the other data points:

the distance from the node B is 5;

the distance from the node C is 7;

The distance from the node D is 3;

The distance from the node E is 9;

the distance from the node F is 8;

according to the distance, 5 data points closest to the node A are selected for interpolation processing, namely, the node D, the node B, the node C, the node F and the node E. Finally, interpolation processing is carried out by adopting numerical information of the nearest 5 data points. The mean interpolation process may be employed herein.

Assume that the power data of node D is 10, the power data of node B is 8, the power data of node C is 12, the power data of node F is 9, and the power data of node E is 11. Then according to the formula:

The interpolation result of node a can be calculated:

；

i.e. the missing power data points of node a are interpolated to 10 by KNN interpolation. By way of example, it can be shown that the KNN interpolation is used to fill in a specific embodiment of a missing electrical energy data point in the distribution network.

As an alternative embodiment, the initial power data may be further preprocessed before performing the incremental interpolation processing on the initial power data to obtain the first power data in 102; the pretreatment at least comprises: and (5) data cleaning and data normalization. Further, outlier detection processing is performed on the initial electric energy data obtained through preprocessing, and detected abnormal electric energy data are removed from the initial electric energy data. The outlier detection process adopts the following formula:

；

Where X is the initial power data point, \mu is the mean of the power data, \sigma is the standard deviation of the power data, and z is the normalized score. Similar to above, \frac { (X- \mu) } represents an averaging operation on (X- \mu). If the absolute value of z exceeds the set threshold, the electrical energy data point corresponding to z is abnormal electrical energy data.

103. And carrying out feature extraction on the first electric energy data by adopting a power distribution network feature analysis model so as to obtain electric energy time domain features and/or electric energy frequency domain features. Wherein the electrical energy time domain features include at least: waveform distortion value of voltage or current. The electric energy frequency domain features include at least: harmonic content characteristics, frequency offset characteristics, power spectrum characteristics.

In the above steps, the feature extraction is performed on the first electric energy data through the power distribution network feature analysis model, so that feature information of the electric energy data can be obtained from two aspects of time domain and frequency domain. The power time domain features may reflect the degree of waveform distortion of the voltage or current, while the power frequency domain features include harmonic content features, frequency offset features, and power spectrum features, which can describe the state of the power quality.

As an optional embodiment, 103, the feature extraction of the first electrical energy data by using the power distribution network feature analysis model to obtain the electrical energy time domain feature and/or the electrical energy frequency domain feature, as shown in fig. 2, may be implemented as the following steps 201 to 202:

201. And inputting the first electric energy data into a first power distribution network characteristic analysis model to obtain electric energy time domain characteristics.

In the embodiment of the application, the first distribution network characteristic analysis model is constructed by a variation self-encoder (Variational Autoencoder, VAE). VAE is an unsupervised learning method from which low-dimensional representations can be extracted for feature extraction and generation of new data samples by learning the underlying spatial distribution of the input data.

Further alternatively, a variant of VAE, DISENTANGLED VAE, is used in the present application to construct the first distribution network characterization model. This variant aims at learning a potential representation of the interpretable factor.

In the constructed model, first electric energy data is acquired and input to an encoder (Encoder) of the first power distribution network characteristic analysis model, and potential space vectors are obtained so that electric energy data points in the first electric energy data are mapped into potential space. And mapping the potential space vector into the electric energy data space through a decoder (Encoder) of the first power distribution network characteristic analysis model, and reconstructing to obtain reconstructed data corresponding to the first electric energy data. Finally, determining a waveform distortion value between the reconstruction data and the first electric energy data; the waveform distortion value is used to evaluate a degree of distortion of the first power data waveform.

For example, assuming that there are N power data points in the first power data, then x= { X ₁,x₂,...,x_N }, where X _i represents one power data point in the set. It is assumed that a vector representation corresponding to each electrical energy data point in the potential space, namely z= { Z ₁,z₂,...,z_N }, can be learned through the first distribution network feature analysis model, wherein each vector has decoupling property in one potential space vector in the Z _i set.

Based on the above assumption, in the encoder of the first distribution network signature analysis model, the encoder maps the input power data points x _i to the potential space, thereby outputting a potential space vector z _i. The encoder may employ a multi-layer perceptron (Multilayer Perceptron, MLP) structure. Further alternatively, the encoder maps x _i into a mean vector μ _i and a variance vector σ _i in potential space, namely:

；

furthermore, in the decoder of the first power distribution network feature analysis model, the decoder remaps the potential space vector z _i into the original power data space to reconstruct and obtain reconstructed data x _i corresponding to the first power data, namely:

；

Finally, a waveform distortion value between the reconstructed data and the first power data is determined. In practice, waveform distortion values include root mean square error (Root Mean Square Error, RMSE) and total harmonic distortion (Total Harmonic Distortion, THD).

Root Mean Square Error (RMSE): calculate the difference between the reconstructed data x _i and the original power data point x _i and square, sum, and open the difference, i.e

；

Total Harmonic Distortion (THD): the ratio of harmonic components in the reconstruction data x _i is calculated and used to measure the degree of distortion of the waveform. It can be calculated by calculating the ratio of the effective value to the total power of the non-fundamental component, namely:

；

Where N _h is the harmonic order, and H _k represents the kth harmonic component.

The potential representation of the electrical energy data is learned by a first power distribution network signature analysis model to calculate waveform distortion values, and the degree of waveform distortion of the voltage or current can be evaluated. These evaluation metrics are intended to help analyze and improve the power performance and quality of the distribution network.

202. And inputting the first electric energy data into a second power distribution network characteristic analysis model to obtain electric energy frequency domain characteristics.

In the embodiment of the application, the second distribution network characteristic analysis model is constructed by a two-dimensional convolutional neural network (2D-Convolutional Neural Network, 2D-CNN) and a Fourier transform network (Fourier Transform Network).

The 2D-CNN model is suitable for extracting frequency domain characteristics of a plurality of sequences of electric energy data, such as electric energy data of a plurality of nodes and a plurality of devices in a power distribution network. The electrical energy data can be directly converted into the frequency domain feature space through a fourier transform network. This method can explicitly extract harmonic content features, frequency offset features, and power spectrum features. The structure of the network can be similar to a convolutional neural network or a full-connection layer, and the frequency domain feature extraction of the data is realized by learning the parameters of the network.

Wherein, as an alternative embodiment, 202, the first power data is converted into a two-dimensional image; inputting the two-dimensional image into a 2D-CNN to extract the corresponding time domain features of the two-dimensional image; inputting the extracted time domain features of the two-dimensional image into a Fourier transform network; and obtaining a frequency domain feature map through a Fourier transform network, and inputting the frequency domain feature map into different feature extraction models to extract and obtain harmonic content features, frequency offset features and power spectrum features.

Specifically, as an alternative embodiment, 202, the first sequence of electrical energy data is converted into a two-dimensional image X for the harmonic content characteristic. For example, the sequence of electrical energy data is represented in the form of a thermodynamic diagram. Further, the two-dimensional image X is processed using a 2D-CNN model to obtain a time domain feature map F (X), and the time domain feature map F (X) is fourier transformed to obtain a frequency domain feature map H (F (X)). In the frequency domain feature map H (F (X)), a harmonic amplitude spectrum is obtained by analyzing the amplitude of the frequency domain image. For each frequency bin k, its corresponding amplitude Ak may be calculated, representing the signal strength at that frequency. Wherein, the calculation formula of harmonic amplitude spectrum:

；

where Re denotes a real part, im denotes an imaginary part, and H (F (X)) (k) denotes a complex value at a kth frequency point in the frequency domain feature map H (F (X)).

By means of this model, harmonic amplitude spectrum features of the electrical energy data can be extracted, which will help to evaluate the individual harmonic intensity information in the electrical energy data.

Specifically, as an alternative embodiment, 202, the first sequence of electrical energy data is converted into a two-dimensional image X for the frequency offset feature. For example, the sequence of electrical energy data is represented in the form of a thermodynamic diagram. Further, the two-dimensional image X is processed using a 2D-CNN model to obtain a time domain feature map F (X), and the time domain feature map F (X) is fourier transformed to obtain a frequency domain feature map H (F (X)). In the frequency domain feature map H (F (X)), a phase spectrum is obtained by analyzing the phase of the frequency domain feature map H (F (X)). For each frequency bin k, its corresponding phase phik may be calculated, representing the phase shift information at that frequency. Wherein, the calculation formula of the phase spectrum:

；

Extracting frequency offset characteristics of the power data from this model will help to evaluate the relative position and phase information at different frequencies in the power data.

Specifically, as an alternative embodiment, 202, for the power spectral feature, the first sequence of electrical energy data is converted into a two-dimensional image X. For example, the sequence of electrical energy data is represented in the form of a thermodynamic diagram. Further, the two-dimensional image X is processed using a 2D-CNN model to obtain a time domain feature map F (X), and the time domain feature map F (X) is fourier transformed to obtain a frequency domain feature map H (F (X)). In the frequency domain feature map H (F (X)), the power spectral density can be obtained by squaring the amplitude of the frequency domain feature map H (F (X)). For each frequency bin k, its corresponding power spectrum Pk may be calculated, representing the power distribution over that frequency. Wherein, the calculation formula of the power spectrum density:

；

Where |h (F (X)) (k) | represents the amplitude value at the kth frequency point in the frequency domain feature map H (F (X)).

The power spectrum characteristics of the power data are extracted through the model, which can help evaluate the power distribution of the power data at different frequencies.

104. And predicting the electric energy time domain characteristics and/or the electric energy frequency domain characteristics through an electric energy quality evaluation model to obtain an electric energy quality evaluation result of the target power distribution network.

In the step, the electric energy time domain characteristics and the electric energy frequency domain characteristics are predicted through the electric energy quality evaluation model, so that an electric energy quality evaluation result of the target power distribution network can be obtained. The result can guide the dispatching and operation of the power system, provide important reference information, and is beneficial to improving the stability of the power distribution network and improving the energy utilization efficiency.

Specifically, in an alternative embodiment, 104, the waveform distortion value of the voltage, the waveform distortion value of the current, the harmonic content characteristic, the frequency offset characteristic, and the power spectrum characteristic are input into a Random Forest model (Random Forest) to obtain the power quality evaluation result of the target power distribution network.

The random forest is an integrated learning model and consists of a plurality of decision trees. By training multiple decision trees and integrating, predictive performance and robustness can be improved.

Further alternatively, referring to fig. 3, the above steps may be further implemented as:

301. Inputting the waveform distortion value of the voltage, the waveform distortion value of the current, the harmonic content characteristic, the frequency offset characteristic and the power spectrum characteristic into a random forest model;

302. Traversing all decision trees in the random forest model, and predicting according to the decision rules of each decision tree to obtain the prediction result of each decision tree;

303. the prediction results of all the decision trees are adopted to obtain a power quality assessment result by adopting a first decision model; the first decision model is constructed based on a multi-classification voting mechanism; or adopting a second decision model to the prediction result of each decision tree to obtain a power quality assessment result; the second decision model is constructed based on a regression mechanism.

For a multi-class voting mechanism, the mathematical expression of the first decision model can be expressed as:

；

Where at { y } is the final power quality assessment result, h (x) is the predicted result for each decision tree, and Nt is the number of decision trees in the random forest. mode represents the mode from among a plurality of predictions, i.e., the classification prediction category that occurs the most often is selected.

For the regression mechanism, the mathematical expression of the second decision model can be expressed as:

；

Where at { y } is the final power quality assessment result, h (x) is the predicted result for each decision tree, and Nt is the number of decision trees in the random forest. sum represents the sum calculation, which results in an average of the predicted results for each decision tree.

Besides the abnormal value processing mode, the asymmetric phenomenon in the power distribution network can be evaluated to further exclude abnormal situations because the asymmetric phenomenon can cause abnormal situations such as unbalanced power grid, overheat equipment, increased power loss and the like.

In an embodiment of the present application, it is optionally assumed that the initial power data further includes asymmetric data. The asymmetry data comprises voltage values or current values of the unbalance of the respective local areas in the distribution network. Based thereon, an evaluation index is determined based on the asymmetry data, including but not limited to: negative sequence coefficient of distribution network, zero sequence current unbalance coefficient, asymmetric S content and voltage unbalance loss. The unbalance degree of the asymmetric data in the power distribution network can be comprehensively estimated from multiple angles through the estimation indexes.

The power distribution network negative sequence coefficient can reflect the unbalance degree between different phases through the ratio of the negative sequence component and the zero sequence component to the positive sequence component which are shown in the negative sequence coefficient. Furthermore, the degree of asymmetry of the power quality of the power distribution network can be estimated by comparing the degree of asymmetry with a standard value. In general, the larger the negative sequence coefficient, the more serious the asymmetric data, and the worse the power quality of the distribution network. Further alternatively, the calculation formula of the negative sequence coefficient of the power distribution network is as follows:

；

Wherein NSF is a negative sequence coefficient of the power distribution network; i_1 is the amplitude of the positive sequence component and is used for representing the current component of normal operation in the power distribution network; i_2 is the amplitude of the negative sequence component and is used for representing the asymmetric current component caused by the negative sequence voltage in the power distribution network; i_0 is the magnitude of the zero sequence component and is used to represent the asymmetric current component in the distribution network caused by the zero sequence voltage.

In the above formula, the square root\sqrt { i_2^2 +i_ 0^2} of the square sum of the magnitudes of the negative sequence component i_2 and the zero sequence component i_0 is calculated, and then the ratio is performed on the square root\sqrt { i_2^2 +i_ 0^2}, and the obtained ratio result is multiplied by 100% to obtain the final negative sequence coefficient NSF.

And a zero sequence current imbalance coefficient (Zero Sequence Current Unbalance Factor) for evaluating the imbalance degree of the zero sequence current. Alternatively, waveforms of the zero sequence currents may be acquired and the magnitude differences of the zero sequence currents calculated for evaluating the unbalance degree of the zero sequence currents. Specifically, the zero sequence current imbalance coefficient may be calculated by:

first, a zero sequence current value of each phase in a three-phase system needs to be acquired. This can be obtained by means of a corresponding measuring device, for example a zero sequence current transformer. Then, the amplitude difference of the zero sequence current is calculated. The zero sequence current imbalance coefficient i.e. I0UF can be calculated according to the following formula:

；

Wherein Imax is the maximum value of the three-phase zero-sequence current, imin is the minimum value of the three-phase zero-sequence current, and Imean is the average value of the three-phase zero-sequence current. Here, imean can be obtained by summing the zero sequence current values of all phases and dividing by the number of phases. Finally, the unbalance degree of the zero sequence current can be estimated through the calculated zero sequence current unbalance coefficient. If the zero sequence current imbalance coefficient is close to 0, the zero sequence current is almost unbalanced; whereas if the coefficient is close to 1, it indicates that the zero sequence current is very unbalanced.

The asymmetric S content (Unsymmetrical S Content, USC) refers to a quantitative indicator of the degree of asymmetry of the three-phase power. As an alternative embodiment, first, it is necessary to obtain a power value for each phase in a three-phase system. For example, by a power measuring instrument or a power monitoring system, to obtain the values of the three-phase active power (P) and the reactive power (Q). Next, the total active power (Ptotal) and reactive power (Qtotal) are calculated:

；

Wherein, P1, P2 and P3 respectively represent active power of A, B, C phases, and Q1, Q2 and Q3 respectively represent reactive power of A, B, C phases. Then, a single-phase power imbalance coefficient (UPC) is calculated:

；

Wherein max (P1, P2, P3) represents the maximum active power value in A, B, C phases, and min (P1, P2, P3) represents the minimum active power value in A, B, C phases.

Finally, the asymmetric S content (USC) is calculated by the following formula:

；

Where USC represents the overall asymmetric S content of the distribution network, where UPC ² represents the degree of active power imbalance, (Qtotal/Ptotal) ² represents the ratio of reactive power to active power.

Through the steps, the asymmetric S content can be calculated and used for evaluating the degree of asymmetry of the three-phase power. It will be appreciated that a higher value of USC indicates a greater degree of asymmetry of the three-phase power.

The voltage unbalance loss (Unbalanced Voltage Loss) is an extra loss caused by three-phase voltage unbalance and is also an index for reflecting the degree of asymmetry of the power grid. Specifically, the voltage unbalance loss can be calculated by:

First, it is necessary to acquire a voltage value of each phase in the three-phase system. This may be obtained by a voltage measuring instrument or a power monitoring system to obtain a voltage value for A, B, C phases. Then, an average value (Uavg) of the three-phase voltage amplitudes is calculated:

；

Wherein U1, U2 and U3 respectively represent the voltage values of A, B, C phases.

Next, the sum of squares of the voltage amplitude differences (Ud) is calculated:

；

finally, the voltage imbalance loss (Ploss) is calculated:

；

where k is the voltage imbalance coefficient, suitably chosen according to the particular case, typically between 0.5 and 1; ptotal is the total active power.

Through the above steps, the voltage unbalance loss can be calculated to evaluate the extra power loss due to the voltage unbalance. In general, the higher the value of the voltage imbalance loss, the greater the degree of imbalance in the voltage, and the higher the additional power loss.

Furthermore, by means of the above-described evaluation indexes, the degree of unbalance of the asymmetric data in the power distribution network can be comprehensively evaluated from multiple angles. Further alternatively, the method of comprehensive evaluation may be that n indexes are assumed, the normalized values thereof are x1, x2, … …, xn, the corresponding weights are w1, w2, … …, wn, and the comprehensive evaluation index is Y. The calculation formula of the comprehensive evaluation index of the unbalance degree of the asymmetric data in the power distribution network is expressed as follows:

；

by sensitivity analysis, the change response of the comprehensive evaluation index to each weight can be evaluated. If the weight of the ith index is changed from wi to (wi+Δwi), the amount by which the overall evaluation index changes can be expressed as:

；

According to the formula, the influence degree of the weight change of different indexes on the comprehensive evaluation index can be calculated. Where xi represents the value of the i-th index after normalization, and wi represents the weight of the i-th index. If delta Y is larger, the importance of the index to the comprehensive evaluation index is higher; if ΔY is smaller, this indicates that the index is of lower importance to the overall evaluation index.

It should be noted that since each index may be affected by other indexes, the calculation of the sensitivity analysis needs to take such interactions into consideration. For example, using the monte carlo method, a set of random weights are generated using a random number generator and simulator, and then the corresponding composite evaluation index is calculated. Through multiple simulations, the varying response of the comprehensive evaluation index to each weight can be evaluated to achieve the effect of weight setting on the comprehensive evaluation index.

The calculation formula of Y and the calculation formula of delta Y are combined to form a multi-index calculation comprehensive evaluation index network, so that the whole asymmetry condition of the power distribution network can be evaluated through the network, and weight adjustment can be carried out according to requirements and priorities.

In the embodiment of the application, initial electric energy data to be evaluated in a target power distribution network is acquired; performing incremental interpolation processing on the initial electric energy data to obtain first electric energy data; performing feature extraction on the first electric energy data by adopting a power distribution network feature analysis model to obtain electric energy time domain features and/or electric energy frequency domain features; and predicting the electric energy time domain characteristics and/or the electric energy frequency domain characteristics through an electric energy quality evaluation model to obtain an electric energy quality evaluation result of the target power distribution network. The application can realize the power quality evaluation of the power distribution network, provides a data basis for the dispatching of power transmission resources in the power distribution network, is beneficial to improving the stability of the power distribution network and improves the energy utilization efficiency.

It should be noted that, the execution subjects of each step of the method provided in the above embodiment may be the same device, or the method may also be executed by different devices. In addition, in some of the flows described in the above embodiments and the drawings, a plurality of operations appearing in a specific order are included, but it should be clearly understood that the operations may be performed out of the order in which they appear herein or performed in parallel, the sequence numbers of the operations such as 101, 102, 103, 104, etc. are merely used to distinguish between the various operations, and the sequence numbers themselves do not represent any order of execution. In addition, the flows may include more or fewer operations, and the operations may be performed sequentially or in parallel.

It should be noted that, the descriptions of "first" and "second" herein are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, and are not limited to the "first" and the "second" being different types.

A power distribution network power quality assessment system according to one or more embodiments of the present application will be described in detail below. Those skilled in the art will appreciate that these power distribution network power quality assessment systems may be configured using commercially available hardware components through the steps taught by the present solution.

In yet another embodiment of the present application, there is also provided a power quality evaluation system of a power distribution network, as shown in fig. 4, the system including:

the acquisition module 401 is configured to acquire initial electrical energy data to be evaluated in the target power distribution network;

a supplement module 402, configured to perform incremental interpolation processing on the initial power data to obtain first power data;

The extracting module 403 is configured to perform feature extraction on the first electrical energy data by using a power distribution network feature analysis model, so as to obtain an electrical energy time domain feature and/or an electrical energy frequency domain feature; the electrical energy time domain features include at least: waveform distortion values of the voltage or the current; the electric energy frequency domain features at least comprise: harmonic content characteristics, frequency offset characteristics, and power spectrum characteristics;

and the prediction module 404 is configured to predict the electric energy time domain feature and/or the electric energy frequency domain feature through an electric energy quality evaluation model, so as to obtain an electric energy quality evaluation result of the target power distribution network.

Optionally, the augmentation module 402 is specifically configured to: determining missing electric energy data points to be filled from the initial electric energy data; calculating the distance between the missing power data point and the existing power data point in the initial power data; and selecting K nearest adjacent electric energy data points to interpolate the missing electric energy data points so as to obtain the first electric energy data.

Alternatively, the interpolation process may be expressed as the following formula:

；

Where h { y } is the interpolation of the missing power data points, y_i is the numerical information of the K nearest neighboring power data points, frac {1} { K } represents the averaging operation on the K nearest neighboring power data points, sum { i=1 } { K } represents the summing operation on the K nearest neighboring power data points, and i is a numerical value greater than 1 and less than K.

Optionally, the extraction module 403 is specifically configured to: inputting the first electric energy data into a first power distribution network characteristic analysis model to obtain the electric energy time domain characteristics; the first power distribution network characteristic analysis model is constructed by a variation self-encoder; and/or inputting the first electric energy data into a second power distribution network characteristic analysis model to obtain the electric energy frequency domain characteristic; the second power distribution network characteristic analysis model is constructed by a two-dimensional convolutional neural network 2D-CNN and a Fourier transform network;

The prediction module 404 is specifically configured to: and inputting the waveform distortion value of the voltage, the waveform distortion value of the current, the harmonic content characteristic, the frequency offset characteristic and the power spectrum characteristic into a random forest model to obtain a power quality evaluation result.

Optionally, when the extraction module 403 inputs the first electrical energy data into the first power distribution network feature analysis model to obtain the electrical energy time domain feature, the method is specifically used for:

acquiring the first electric energy data, inputting the first electric energy data into an encoder of a first power distribution network feature analysis model, and obtaining potential space vectors so that electric energy data points in the first electric energy data are mapped into potential spaces;

Mapping the potential space vector into an electric energy data space through a decoder of a first power distribution network characteristic analysis model, and reconstructing to obtain reconstruction data corresponding to the first electric energy data;

Determining a waveform distortion value between the reconstruction data and the first power data; the waveform distortion value is used for evaluating the distortion degree of the first electric energy data waveform; the waveform distortion values include root mean square error and total harmonic distortion.

Optionally, the extracting module 403 inputs the first electrical energy data into a second power distribution network feature analysis model, so as to obtain the electrical energy frequency domain feature, which is specifically configured to:

converting the first electric energy data into a two-dimensional image;

inputting the two-dimensional image into the 2D-CNN to extract corresponding two-dimensional image time domain features;

Inputting the extracted time domain features of the two-dimensional image into a Fourier transform network;

and obtaining a frequency domain feature map through the Fourier transform network, and inputting the frequency domain feature map into different feature extraction models to extract and obtain harmonic content features, frequency offset features and power spectrum features.

Optionally, when the prediction module 404 inputs the waveform distortion value of the voltage, the waveform distortion value of the current, the harmonic content feature, the frequency offset feature, and the power spectrum feature into the random forest model to obtain the power quality evaluation result, the prediction module is specifically configured to:

Inputting the waveform distortion value of the voltage, the waveform distortion value of the current, the harmonic content characteristic, the frequency offset characteristic and the power spectrum characteristic into a random forest model;

traversing all decision trees in the random forest model, and predicting according to the decision rules of each decision tree to obtain the prediction result of each decision tree;

the prediction results of all the decision trees are adopted to obtain the electric energy quality assessment results by adopting a first decision model; the first decision model is constructed based on a multi-classification voting mechanism; or alternatively

The prediction results of all the decision trees are adopted to obtain the electric energy quality assessment results by adopting a second decision model; the second decision model is constructed based on a regression mechanism.

Optionally, the system further includes an anomaly detection module for preprocessing the initial power data before the augmentation module 402 performs incremental interpolation processing on the initial power data to obtain first power data; the pretreatment at least comprises: data cleaning and data normalization; performing outlier detection processing on the initial electric energy data obtained through pretreatment, and removing detected abnormal electric energy data from the initial electric energy data; the outlier detection process adopts the following formula:

；

Where X is the initial power data point, (\mu) is the mean of the power data, (\sigma) is the standard deviation of the power data, and z is the normalized score; frac { (X- \mu) } represents an averaging operation on (X- \mu); if the absolute value of z exceeds the set threshold, the electrical energy data point corresponding to z is abnormal electrical energy data.

Optionally, the initial power data further comprises asymmetric data; the asymmetry data comprises voltage values or current values of the unbalance of the respective local areas in the distribution network.

The system also comprises an asymmetry evaluation module, a compensation module and a compensation module, wherein the asymmetry evaluation module is used for determining a negative sequence coefficient, a zero sequence current imbalance coefficient, an asymmetric S content and a voltage imbalance loss of the power distribution network based on the asymmetry data; the calculation formula of the negative sequence coefficient of the power distribution network is as follows:

；

Wherein NSF is the negative sequence coefficient of the power distribution network; i_1 is the amplitude of the positive sequence component and is used for representing the current component of normal operation in the power distribution network; i_2 is the amplitude of the negative sequence component and is used for representing the asymmetric current component caused by the negative sequence voltage in the power distribution network; i_0 is the amplitude of the zero sequence component and is used for representing the asymmetric current component caused by the zero sequence voltage in the power distribution network;

Based on the negative sequence coefficient of the power distribution network, the zero sequence current unbalance coefficient, the asymmetric S content and the voltage unbalance loss, calculating a comprehensive evaluation index of unbalance degree of asymmetric data in the power distribution network by adopting a sensitivity analysis method; the calculation formula of the comprehensive evaluation index is as follows:

；

Where xi represents the value of the i-th index after normalization, and wi represents the weight of the i-th index.

According to the power quality evaluation system for the power distribution network, the power quality evaluation of the power distribution network can be realized, a data basis is provided for the dispatching of power transmission resources in the power distribution network, the stability of the power distribution network is improved, and the energy utilization efficiency is improved.

In still another embodiment of the present application, there is also provided an electronic apparatus including: the device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;

A memory for storing a computer program;

and the processor is used for realizing the power quality evaluation method of the power distribution network according to the embodiment of the method when executing the program stored in the memory.

The communication bus 1140 referred to above for electronic devices may be a peripheral component interconnect standard (PERIPHERAL COMPONENT INTERCONNECT, PCI) bus, or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, or the like. The communication bus 1140 may be divided into an address bus, a data bus, a control bus, and the like.

For ease of illustration, only one thick line is shown in fig. 5, but not only one bus or one type of bus.

The communication interface 1120 is used for communication between the electronic device and other devices described above.

Memory 1130 may include random access memory (Random Access Memory, RAM) or non-volatile memory (non-volatil ememory), such as at least one disk memory. Optionally, the memory may also be at least one memory device located remotely from the aforementioned processor.

The processor 1110 may be a general-purpose processor including a Central Processing Unit (CPU), a network processor (Network Processor, NP), and the like; but may also be a digital signal processor (DIGITAL SIGNAL Processing, DSP), application specific integrated circuit (Application SpecificIntegrated Circuit, ASIC), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components.

Accordingly, the present application also provides a computer readable storage medium storing a computer program, where the computer program is executed to implement the steps executable by the electronic device in the above method embodiments.

Claims

1. A method for evaluating the power quality of a power distribution network, the method comprising:

Predicting the electric energy time domain features and/or the electric energy frequency domain features through an electric energy quality evaluation model to obtain an electric energy quality evaluation result of a target power distribution network;

Performing feature extraction on the first electric energy data by adopting a power distribution network feature analysis model to obtain electric energy time domain features and/or electric energy frequency domain features, wherein the method comprises the following steps:

inputting the first electric energy data into a first power distribution network characteristic analysis model to obtain the electric energy time domain characteristics; the first power distribution network characteristic analysis model is constructed by a variation self-encoder;

inputting the first electric energy data into a second power distribution network characteristic analysis model to obtain the electric energy frequency domain characteristics; the second power distribution network characteristic analysis model is constructed by a two-dimensional convolutional neural network 2D-CNN and a Fourier transform network;

Predicting the electric energy time domain feature and/or the electric energy frequency domain feature through an electric energy quality evaluation model to obtain an electric energy quality evaluation result of a target power distribution network, wherein the electric energy quality evaluation result comprises:

And inputting the waveform distortion value of the voltage, the waveform distortion value of the current, the harmonic content characteristic, the frequency offset characteristic and the power spectrum characteristic into a random forest model to obtain a power quality evaluation result.

2. The power quality assessment method of a power distribution network according to claim 1, wherein performing incremental interpolation processing on the initial power data to obtain first power data includes:

Determining missing electric energy data points to be filled from the initial electric energy data;

calculating the distance between the missing power data point and the existing power data point in the initial power data;

and selecting K nearest adjacent electric energy data points to interpolate the missing electric energy data points so as to obtain the first electric energy data.

3. The power quality assessment method of a power distribution network according to claim 2, wherein the interpolation process can be expressed as the following formula:

；

4. The power distribution network power quality assessment method according to claim 1, wherein inputting the first power data into a first power distribution network feature analysis model to obtain the power time domain feature comprises:

5. The power distribution network power quality assessment method according to claim 1, wherein inputting the first power data into a second power distribution network feature analysis model to obtain the power frequency domain feature comprises:

converting the first electric energy data into a two-dimensional image;

6. The power quality assessment method of a power distribution network according to claim 1, wherein inputting waveform distortion values of voltages, waveform distortion values of currents, harmonic content characteristics, frequency offset characteristics, and power spectrum characteristics into a random forest model to obtain a power quality assessment result comprises:

the prediction results of all the decision trees are adopted to obtain the electric energy quality assessment results by adopting a first decision model; the first decision model is constructed based on a multi-class voting mechanism.

7. The power quality assessment method of a power distribution network according to claim 1, further comprising, before performing incremental interpolation processing on the initial power data to obtain first power data:

preprocessing the initial electric energy data; the pretreatment at least comprises: data cleaning and data normalization;

Performing outlier detection processing on the initial electric energy data obtained through pretreatment, and removing detected abnormal electric energy data from the initial electric energy data;

the outlier detection process adopts the following formula:

；

8. The power distribution network power quality assessment method according to claim 1, wherein the initial power data further comprises asymmetric data; the asymmetric data comprises unbalanced voltage values or current values of various local areas in the power distribution network; the method further comprises the steps of:

Determining a negative sequence coefficient, a zero sequence current imbalance coefficient, an asymmetric S content and a voltage imbalance loss of the power distribution network based on the asymmetric data; the calculation formula of the negative sequence coefficient of the power distribution network is as follows:

；

9. A power distribution network power quality assessment system, the system comprising:

The prediction module is used for predicting the electric energy time domain characteristics and/or the electric energy frequency domain characteristics through an electric energy quality evaluation model so as to obtain an electric energy quality evaluation result of the target power distribution network;