CN109359665B - Household appliance load identification method and device based on support vector machine - Google Patents
Household appliance load identification method and device based on support vector machine Download PDFInfo
- Publication number
- CN109359665B CN109359665B CN201810990146.3A CN201810990146A CN109359665B CN 109359665 B CN109359665 B CN 109359665B CN 201810990146 A CN201810990146 A CN 201810990146A CN 109359665 B CN109359665 B CN 109359665B
- Authority
- CN
- China
- Prior art keywords
- load
- data
- household appliance
- training
- identification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention provides a household appliance load identification method and a household appliance load identification device based on a support vector machine, wherein the method comprises the following steps: acquiring total load data of started household appliances; and inputting the total load data into a trained load identification model to obtain identification data. The method and the device for identifying the load of the household appliance based on the support vector machine adopt non-invasive load identification, do not need to monitor the running state and the power consumption of each electric appliance, do not need to enter the electric appliance during installation, have small investment, do not influence normal production work, are suitable for comprehensive popularization, apply the support vector machine algorithm to the identification of the household appliance, and improve the accuracy of the load identification of the household appliance.
Description
Technical Field
The invention relates to the field of non-invasive load detection, in particular to a household appliance load identification method and device based on a support vector machine.
Background
For an electric power system, the electric power load monitoring has great significance, and the method is not only beneficial to improving load composition, guiding reasonable consumption of users and reducing power consumption cost, but also beneficial to optimal configuration of national electric power resources. The power load decomposition data can enable power consumers to know the power consumption of various electric equipment at different time intervals in more detail, help the electric consumers to make reasonable energy-saving plans, adjust the use of the electric equipment, reduce the power consumption and the electricity expense expenditure, and simultaneously help power companies to know the load composition of a power system more truly, standardize the load power consumption, reasonably arrange the service time of various loads, improve the utilization efficiency of a power grid, reduce the investment of the power system and reduce the running grid loss of the system.
In recent years, research on transient waveform identification home appliances is gradually increased, and with the increase of sampling rate, the requirement of people on home appliance identification precision is also continuously increased, and the accuracy of a load identification model is influenced by large-batch load data and complex conditions.
The existing resident power load monitoring technology is to equip a sensor for each electrical appliance to obtain the electricity utilization information of the electrical appliance, and belongs to intrusive load monitoring. A sensor with a digital communication function is installed on an interface of each electric appliance and a power grid in an intrusive mode, so that the operation state and the power consumption of each electric appliance are monitored. The method has accurate measurement, but the investment is large, the installation work needs to enter the interior of the electric appliance, the normal production work is influenced, and the method is not suitable for comprehensive popularization.
Disclosure of Invention
In order to overcome the technical defects, embodiments of the present invention provide a method and an apparatus for identifying a load of a home appliance based on a support vector machine.
In a first aspect, an embodiment of the present invention provides a method for identifying a load of a home appliance based on a support vector machine, including: acquiring total load data of started household appliances; and inputting the total load data into a trained load identification model to obtain identification data.
In a second aspect, an embodiment of the present invention provides a device for identifying a load of a home appliance based on a support vector machine, including: the acquisition module is used for acquiring total load data of the started household appliance; and the identification module is used for inputting the total load data into a trained load identification model to obtain identification data.
In a third aspect, an embodiment of the present invention provides an electronic device, including a memory and a processor, where the processor and the memory complete communication with each other through a bus; the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the method for identifying the load of the household appliance based on the support vector machine according to the first aspect.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to implement a method for identifying a load of an appliance based on a support vector machine according to the first aspect.
The method and the device for identifying the load of the household appliance based on the support vector machine adopt non-invasive load identification, do not need to monitor the running state and the power consumption of each electric appliance, do not need to enter the electric appliance during installation, have small investment, do not influence normal production work, are suitable for comprehensive popularization, apply the support vector machine algorithm to the identification of the household appliance, and improve the accuracy of the load identification of the household appliance.
Drawings
Fig. 1 is a schematic flow chart of a method for identifying a load of a household appliance based on a support vector machine according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an algorithm flow for optimizing SVM parameters by using PSO according to an embodiment of the present invention;
fig. 3 is a schematic flowchart of a method for identifying a load of a home appliance based on a support vector machine according to an embodiment of the present invention;
FIG. 4 is a waveform diagram of the turn-on period of the household appliance to be identified according to the embodiment of the present invention;
fig. 5 is a 3D view of SVM parameter selection and accuracy distribution of a training model of the home appliance 1 according to the embodiment of the present invention;
FIG. 6 is a graph of a particle swarm algorithm parameter optimization fitness curve according to an embodiment of the present invention;
fig. 7 is a 3D view of SVM parameter selection and accuracy when the event detection error of the home appliance 1 is within 500 data points according to the embodiment of the present invention;
fig. 8 is a 3D view of SVM parameter selection and accuracy distribution when the event detection error of the home appliance 1 is within 1000 data points according to the embodiment of the present invention;
fig. 9 is a schematic structural diagram of a household appliance load identification apparatus based on a support vector machine according to an embodiment of the present invention;
fig. 10 is a schematic physical structure diagram of an electronic device according to an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
Fig. 1 is a schematic flowchart of a method for identifying a load of a home appliance based on a support vector machine according to an embodiment of the present invention, as shown in fig. 1, including:
and step 12, inputting the total load data into a trained load identification model to obtain identification data.
In the embodiment of the invention, an adopted household appliance identification method belongs to the field of non-invasive load identification, and a non-invasive load monitoring (NILM) device can measure and obtain signals of load-bearing electric power information such as voltage, current and the like of a total load, wherein the information comprises information of different characteristic load components. By extracting the characteristic information of these electrical quantities, the NILM system can achieve load shedding.
That is, by detecting the electric power signal such as the voltage, current or power of the total load, it is possible to detect what electric appliance the user is using.
By obtaining non-intrusive load detection data from a main circuit outside a building or customer premises, the type of load detection data can be selected as a power signal. The method comprises the steps of detecting a total power signal after the active household appliances are started and aggregated at an instrument entrance outside a building.
The embodiment of the invention carries out recognition based on the load recognition model, and inputs the total power signal into the trained load recognition model to obtain recognition data, namely the power signal of each special device, thereby realizing the decomposition and recognition of each household appliance.
The household appliance load identification method based on the support vector machine provided by the embodiment of the invention adopts non-invasive load identification, does not need to monitor the running state and power consumption of each electric appliance, does not need to enter the electric appliance during installation work, has small investment, does not influence normal production work, is suitable for comprehensive popularization, applies the support vector machine algorithm to household appliance identification, and improves the accuracy of household appliance load identification.
On the basis of the above embodiment, the trained load recognition model is obtained by the following steps:
constructing a training sample set by utilizing the load identification type;
inputting the data in the training sample set into a load recognition model for training, and obtaining a pre-trained load recognition model based on a K-CV method in cross validation;
and calculating the classification accuracy by using the pre-trained load identification model, and adjusting the pre-trained load identification model based on the classification accuracy and a particle swarm optimization algorithm to obtain the trained load identification model.
Firstly, a training sample set is constructed according to a load identification type, wherein the load identification type can reflect unique information reflecting the power utilization state, such as voltage, a working waveform, starting current and the like, of power utilization equipment in the running process. The characteristics are determined by the working conditions of the electric equipment, and the load identification types can be classified into 3 types of steady state, transient state and operation mode, wherein the steady state and the transient state depend on the characteristics of components inside the equipment; the operating mode is determined by the operating control strategy of the device. During the operation of the equipment, the load identification types can repeatedly appear, and therefore, each electric appliance can be identified. In the embodiment of the present invention, the transient type is used.
And after the training sample set is obtained, inputting the data in the training sample set into a load recognition model for training according to a support vector machine algorithm. The penalty parameter c and the kernel function parameter g are main parameters needing to be adjusted in the SVM, c represents the tolerance of the load recognition model when the load waveform and the learning process are different, and g is the parameter of the mapped kernel function. And taking the training sample set as an original data set, and obtaining the classification accuracy of the training sample set under the punishment parameter c and the kernel function parameter g and a corresponding pre-trained load identification model by utilizing K-fold Cross Validation (K-CV) in Cross Validation (CV).
And setting parameter optimizing ranges of c and g, and performing parameter optimizing by using a particle swarm optimization algorithm. The fitness of the PSO is the classification accuracy obtained by a K-CV method after the parameters c and g are determined, so that the optimal parameters c and g are found to enable the optimal parameters c and g to obtain the highest classification accuracy under CV, and the pre-trained load identification model is adjusted according to the optimal parameters c and g to obtain the trained load identification model.
The household appliance load identification method based on the support vector machine provided by the embodiment of the invention adopts non-invasive load identification, does not need to monitor the running state and power consumption of each electric appliance, does not need to enter the electric appliance during installation work, has small investment, does not influence normal production work, is suitable for comprehensive popularization, applies the support vector machine algorithm to household appliance identification, and improves the accuracy of household appliance load identification. Meanwhile, a load recognition model is trained by using a K-CV method, parameter optimization is assisted by using a particle swarm optimization algorithm, main parameters in the SVM are adjusted, the load recognition model with higher classification accuracy is obtained, and the precision of the model is improved.
On the basis of the above embodiment, the constructing a training sample set by using the load recognition type specifically includes:
generating a corresponding simulation data set from the preprocessed load data according to the load identification type;
dividing the simulation data set into a training set and a test set according to a preset proportion;
and carrying out normalization processing on the training set and the test set to obtain the training sample set.
The generating of the corresponding simulation data set from the preprocessed load data according to the load identification type specifically includes:
acquiring load data of a certain number of household appliances which are independently started in unit time, and screening the load data according to a preset rule to obtain generated simulation data;
synthesizing the simulation data set using the generated simulation data.
The load identification type selected in the embodiment of the invention is a transient power signal, and the power is equal to the product of voltage and current. The total power signal is expressed in terms of a mathematical expression that can be approximated as:
P(t)＝p1(t)+p2(t)+...+pn(t)
where n is the total number of active home devices that have been turned on during the measurement period, pi(t) represents the power consumption of a single device that the ith home device contributes to the overall aggregate measurement at time t. The purpose of the NILM system is to decompose the total power signal p (t) measured by the meter into power signals of each specific device, thereby realizing the decomposition and identification of each household appliance in the building.
Firstly, voltage and current data of a certain number of household appliances which are independently started in unit time are obtained, then the voltage and current data are screened, and data with obvious fluctuation are selected. The predetermined rule of the screening is based on event detection.
The event detection is simply to determine whether a new event is generated according to a certain rule based on the change of the signal. The simplest method is to calculate the change of load data of adjacent time or time periods, compare the change with a set threshold value, and judge that an event occurs when the change exceeds the threshold value. The method is simple and easy to operate, and has the problem that the setting of the threshold value is skillful, and the too large or too small threshold value can cause wrong event detection results. For this reason, the parameters need to be trained on a large number of samples. In the embodiment of the invention, a threshold value is set according to experience, when the fluctuation exceeds the threshold value, the event occurs, otherwise, no event occurs. After screening, the generated simulation data is obtained.
Then synthesizing a simulation data set according to the generated simulation data, wherein the synthesis method comprises the following steps:
the load data of each home appliance may be expressed as:
q1(n)＝{q1(1),q1(2),...,q1(N1)}
q2(n)＝{q2(1),q2(2),...,q2(N2)}
......
qm(n)＝{qm(1),qm(2),...,qm(Nm)}
where m is the total number of data sets, NiAnd q (i) is the sampling value of the apparent power of the household appliance. Synthesizing the household appliance load data, and assuming that the household appliances corresponding to the synthesized household appliance data are m and n, the synthesizing mode is as follows:
Sr(n)＝{s(1),s(2),...,s(br),s(br+1),...,s(br+ar),s(br+ar+1),
...s(Nm-cr+br+1)}
＝{0,0,...,qm(cr),qm(cr+1),...,qm(cr+ar)+qn(1),
qm(cr+ar+1)+qn(2),...,qm(Nm-cr+1)+qn(Nn-ar)}
in the formula, Sr(n) the synthesized simulated data set, s (i) the apparent power sample values of the simulated data set, brThe position of the moment when the household appliance m is turned on, arIs the distance from the location of the m starting time of the household appliance, crThe activated reading position is extracted for the home appliance n. Most of the time cr1, when the household appliance m is turned on before the household appliance n and the turn-on characteristic is incomplete, c occursr>1.
After obtaining the simulation data set, dividing the simulation data set into a simulation set and a test set, and uniformly carrying out normalization processing on the simulation set and the test set to obtain a training sample set, wherein the normalization mapping is as follows:
xmin＝min(x)
xmax＝max(x)
wherein x represents the training set and the test set data, and y is the data set after normalization processing. The purpose of the normalization process is to normalize the data. The data standardization process is to scale the data according to the proportion to enable the data to fall into a smaller characteristic interval, and to convert the data into a dimensionless pure value by removing the unit limitation of the data, and the data processing process is beneficial to comparing and weighting the data of different units or measurement levels. In general, if the input data has a mean value, the efficiency and accuracy of the machine learning algorithm are higher, and therefore a common data normalization means is normalization processing. The result of the normalization is to convert the raw data into numbers in the interval [0,1], i.e., y ∈ [0,1], i ═ 1, 2.
The household appliance load identification method based on the support vector machine provided by the embodiment of the invention adopts non-invasive load identification, does not need to monitor the running state and power consumption of each electric appliance, does not need to enter the electric appliance during installation work, has small investment, does not influence normal production work, is suitable for comprehensive popularization, applies the support vector machine algorithm to household appliance identification, and improves the accuracy of household appliance load identification.
On the basis of the above embodiment, the obtaining of the pre-trained load recognition model based on the K-CV method in the cross validation specifically includes:
equally dividing the training sample set into K training subsets;
and (3) making a primary verification set for each training subset, taking the rest K-1 training subsets as training sets, and respectively performing training verification to obtain K pre-trained load identification models.
The penalty parameter c and the kernel function parameter g are main parameters needing to be adjusted in the SVM, c represents the tolerance of the model when the load waveform and the learning process are different, and g is the parameter of the mapped kernel function. The training sample set is used as an original data set, and the classification accuracy of the training set under the c and g groups is obtained by using K-fold Cross Validation (K-CV) in Cross Validation (CV). The K-CV method is realized by equally dividing original data into K groups, making a primary verification set for each subset data, taking the rest K-1 groups of subset data as training sets, and obtaining K training models after respectively performing training and verification. And taking the average of the classification accuracy of the K model final verification sets as the performance index of the classifier under K-CV. Wherein, K generally begins to take a value from small to big, and the minimum is 3, and can take 2 when the data is less.
On the basis of the foregoing embodiment, the adjusting the pre-trained load recognition model based on the classification accuracy and the particle swarm optimization algorithm to obtain a trained load recognition model specifically includes:
setting a parameter optimizing range, wherein the parameter optimizing range comprises a punishment parameter range and a kernel function parameter range;
optimizing and iterating the punishment parameters and the kernel function parameters, and calculating the classification accuracy after each iteration based on a K-CV method in cross validation;
and after the iteration times are reached, outputting a group of corresponding first punishment parameters and first kernel function parameters with the highest classification accuracy to obtain the trained load identification model corresponding to the first punishment parameters and the first kernel function parameters.
After outputting a set of corresponding first penalty parameters and first kernel function parameters with the highest classification accuracy, the method further comprises:
if the highest classification accuracy rate corresponds to multiple groups, outputting a first punishment parameter and a first kernel function parameter corresponding to one group with the smallest punishment parameter in the multiple groups;
and if the minimum punishment parameters correspond to multiple groups, outputting a group of first searched corresponding first punishment parameters and first kernel function parameters.
Parameter Optimization is carried out by adopting a Particle Swarm Optimization (PSO) for assistance, the fitness of the PSO is the classification accuracy obtained by a K-CV method after the parameters c and g are determined, and therefore the optimal parameters c and g are found out so that the highest classification accuracy can be obtained under CV.
Fig. 2 is a schematic flow chart of an algorithm for optimizing SVM parameters by using PSO according to an embodiment of the present invention, as shown in fig. 2, including:
in step 212, the optimal solution (optimal parameters c and g) is output.
The fitness of PSO is defined as the classification accuracy obtained by a K-CV method after parameters c and g are determined, after parameter optimizing ranges and related parameters are set according to experience, population positions (values of c and g) are generated randomly according to the set population number and the optimizing ranges, and initial speed is defined.
And next, calculating the corresponding fitness (classification accuracy) according to the c and g values in each population, recording the optimal values between the population and the individuals, updating the population and the speed based on the fitness, and changing the position of the population to the c and g values corresponding to the optimal and global optimal values of the individual. In order to avoid the influence of local optimization on the final result in the process of population migration, random variation with small probability is given to partial individuals on the basis that the population position is controlled by the speed.
When the final iteration times are reached, the optimizing process is terminated, and the population positions (optimal parameters c and g) reaching the optimal fitness (highest classification accuracy) are output. If the result has the highest classification accuracy rate corresponding to a plurality of groups of c and g, considering that c is used as a penalty parameter, if the value of c is too high, an overfitting state is easily caused, namely the problem that the accuracy rate of a training sample set is very high, but the generalization capability of a model obtained by training is too low is easily caused, so that the group of c and g with the minimum parameter c is selected as an optimal parameter, and if the parameter corresponding to the minimum c also has a plurality of groups, the group searched first is selected as the optimal parameter.
The household appliance load identification method based on the support vector machine provided by the embodiment of the invention adopts non-invasive load identification, does not need to monitor the running state and power consumption of each electric appliance, does not need to enter the electric appliance during installation work, has small investment, does not influence normal production work, is suitable for comprehensive popularization, applies the support vector machine algorithm to household appliance identification, and improves the accuracy of household appliance load identification. Meanwhile, the pre-trained load identification model is adjusted by assisting parameter optimization through a particle swarm optimization algorithm, so that the load identification model has higher classification accuracy and the performance of the model is improved.
Fig. 3 is a schematic flow chart of a method for identifying a load of a home appliance based on a support vector machine according to an embodiment of the present invention, as shown in fig. 3, the method includes three stages, wherein the stage-one data preparation includes:
the phase two data training comprises:
310, optimizing and iterating SVM parameters;
311, training an SVM load recognition model according to cross validation;
313, obtaining optimal parameters c and g and a trained load identification model according to the highest accuracy;
the stage three identification verification comprises the following steps:
and step 317, outputting the accuracy rate and the recognition result of the test set.
The method provided by the embodiment of the invention is described in detail by an example.
In the calculation example, the load data sampling rate is 30000Hz, it is assumed that the opening instantaneous characteristic action time lengths of all the household appliances in the original data are 1.5s, the sequence time length of the household appliance time window is 2s, the opening time interval of two household appliance devices is within 0.5s, and the data length of the time sequence corresponding to each household appliance opening event is 60000 units. The specific rule of the simulation data set synthesized by the extraction activated generated simulation data is shown in table 1:
TABLE 1 detailed rules for synthesizing simulation datasets
Wherein, ai、biIs a random number of 1-15000, andi、bisatisfies ai+biThe data type with the activated equipment is not more than 15001 selected household appliance data to be detected, the data type without the activated equipment is used as the data of other household appliances except the identification household appliance which are interfered, and the selected household appliance data are randomly selected from the screened data set.
The addition of the simulation data containing only a single activation device to the combined simulation data set is intended to guide the learning process of the algorithm, the homogeneous ones generated in the tableThe waveform data of the type has the same proportion of data of the activated equipment and data of the activated equipment, and the purpose is to reduce the situations of over-fitting and under-fitting in the training process. In any case, it is desirable to keep the turn-on characteristics of the activated device and the deactivated device that is "incorrect" intact in the generated simulation data set to facilitate the training learning process of the model, where a is satisfiedi+biThe purpose of ≦ 15001 and the on position being a random number between 1 and 15000 is to ensure that the time between an active device and a disturbing inactive device does not exceed 0.5 s.
The simulation randomly selects 4 household appliances corresponding to data from an original data set as activation equipment, and the experiment respectively learns and deduces the simulation data of each household appliance, wherein fig. 4 is a waveform diagram of the starting period of the household appliance to be recognized in the embodiment of the invention, in fig. 4, (a) a corresponding household appliance 1, (b) a corresponding household appliance 2, (c) a corresponding household appliance 3, and (d) a corresponding household appliance 4, the waveform of the apparent power of the starting period of each household appliance is shown in fig. 4, the simulation set containing the waveform of the household appliance to be recognized is marked to generate a simulation data set, wherein 80% of the data are uniformly and randomly extracted as a training set, and the rest 20% of the data are used as a test set for verifying the accuracy of the recognition model. Fig. 5 is a 3D view of SVM parameter selection and accuracy distribution of a training model of the home appliance 1 according to the embodiment of the present invention, and fig. 6 is a parameter optimization fitness curve diagram of a particle swarm algorithm according to the embodiment of the present invention, as shown in fig. 5 and fig. 6. The recognition results are shown in table 2.
TABLE 2 load identification results
The above experimental process is established under the conditions of accurate event detection, namely model training and load identification under an ideal condition, and as can be seen from table 2, the performance of the load identification method under the ideal condition is better, but it is considered that in practical application, due to feature superposition and possible interference of a large amount of loads during working, activation and detection of a load opening event become more difficult. In order to deal with the situation that the load opening event detection has errors, a certain error needs to be added during the process of carrying out model training by using data, meanwhile, in order to keep effective household appliance opening characteristics, the household appliance opening position obtained by event detection is set to be a random position 500 or 1000 data points away from the actual opening position, a load identification model with event detection error tolerance can be obtained after the model training process is carried out, fig. 7 is a 3D view of SVM parameter selection and accuracy when the event detection error of the electric appliance 1 is within 500 data points according to the embodiment of the present invention, fig. 8 is a 3D view of SVM parameter selection and accuracy distribution when the event detection error of the electric appliance 1 is within 1000 data points according to the embodiment of the present invention, the SVM parameter selection and accuracy distribution of the home appliance 1 model obtained by training as shown in fig. 7 and 8 are shown in table 3.
TABLE 3 load identification results with event detection errors
As can be seen from the recognition results in table 3, the load recognition accuracy of the training model is reduced in consideration of the event detection error when comparing the case where the event detection is accurate, and the recognition results are different for the household appliance load waveforms with different characteristics. The identification accuracy rate of the event detection error is compared, and the larger the error is, the lower the accuracy rate is. As can be seen by comparing fig. 7 and 8 with fig. 5 and 6, the training environment changes, the span of the variation range of the SVM parameter distribution is wider, the variation rule tends to be gentle, the overall accuracy of the training set is reduced, and the learning efficiency is poorer in the training process compared with the ideal case. But the accuracy of the training set can be effectively improved by increasing the training data, and the accuracy of the test set is also improved to a certain extent.
Fig. 9 is a schematic structural diagram of a household electrical appliance load identification apparatus based on a support vector machine according to an embodiment of the present invention, as shown in fig. 9, including an obtaining module 91 and an identifying module 92, where the obtaining module 91 is configured to obtain total load data of a started household electrical appliance; the recognition module 92 is configured to input the total load data into the trained load recognition model to obtain recognition data.
In the embodiment of the invention, an adopted household appliance identification method belongs to the field of non-invasive load identification, and a non-invasive load monitoring (NILM) device can measure and obtain signals of load-bearing electric power information such as voltage, current and the like of a total load, wherein the information comprises information of different characteristic load components. By extracting the characteristic information of these electrical quantities, the NILM system can achieve load shedding.
That is, by detecting the electric power signal such as the voltage, current or power of the total load, it is possible to detect what electric appliance the user is using.
By obtaining non-intrusive load detection data from a main circuit outside a building or customer premises, the type of load detection data can be selected as a power signal. The obtaining module 91 detects the total power signal after the aggregation of the activated household appliances is started at the instrument entrance outside the building.
In the embodiment of the invention, the identification is carried out based on the load identification model, the identification module 92 inputs the total power signal into the trained load identification model to obtain identification data, namely the power signal of each special device, thereby realizing the decomposition and identification of each household appliance. The apparatus provided in the embodiment of the present invention is used for executing the above method embodiments, and for detailed descriptions and specific processes, reference is made to the above method embodiments, which are not described herein again.
The household appliance load identification device based on the support vector machine provided by the embodiment of the invention adopts non-invasive load identification, does not need to monitor the running state and power consumption of each electric appliance, does not need to enter the electric appliance during installation work, has small investment, does not influence normal production work, is suitable for comprehensive popularization, applies the support vector machine algorithm to household appliance identification, and improves the accuracy of household appliance load identification.
The embodiment of the invention provides a household appliance load identification system based on a support vector machine, which comprises the following modules:
data preparation module
In the data preparation module, a corresponding simulation data set is generated according to the load identification type by utilizing the preprocessed load data, then the activation position is extracted according to the event occurrence position, the simulation data set is marked according to the load identification content, and the simulation data set is divided into a training set and a testing set according to the proportion.
Data training module
Determining algorithm parameters and a training method, setting a parameter optimization range, optimizing the generated load identification model through a training set based on a cross validation method, and searching parameters c and g which can ensure the accuracy of the training set to be optimal and a corresponding load identification model.
Test verification module
And determining the event occurrence position of the test set data by using event detection, and identifying by using the generated load identification model.
The embodiment of the invention provides a household appliance load identification system based on a support vector machine, which utilizes an implementation method of analyzing non-invasive load detection in house load identification, provides a household appliance identification method based on the support vector machine aiming at defects in the load identification method, designs a data training process for generating a simulation data set according to a limited data set by making a house meet the particularity of an identification problem, optimizes parameters of the support vector machine by utilizing a particle swarm algorithm on the basis of a grid search method, and determines optimal parameters according to the accuracy of a training set obtained by the parameters. The invention has the advantages of low cost, convenient installation and good identification effect.
Fig. 10 illustrates a physical structure diagram of an electronic device, and as shown in fig. 10, the electronic device may include: a processor (processor)101, a communication Interface (communication Interface)102, a memory (memory)103 and a bus 104, wherein the processor 101, the communication Interface 102 and the memory 103 complete communication with each other through the bus 104. Bus 104 may be used for information transfer between the electronic device and the sensor. The processor 101 may call logic instructions in the memory 103 to perform the following method: acquiring total load data of started household appliances; and inputting the total load data into a trained load identification model to obtain identification data.
In addition, the logic instructions in the memory 103 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
An embodiment of the present invention provides a non-transitory computer-readable storage medium, which stores computer instructions, where the computer instructions cause a computer to execute the pseudo base station positioning method provided in the foregoing embodiment, for example, including: acquiring total load data of started household appliances; and inputting the total load data into a trained load identification model to obtain identification data.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Various modifications and additions may be made to the described embodiments by those skilled in the art without departing from the spirit of the invention or exceeding the scope as defined in the appended claims.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (8)
1. A household appliance load identification method based on a support vector machine is characterized by comprising the following steps:
acquiring total load data of started household appliances;
inputting the total load data into a trained load identification model to obtain identification data;
the trained load recognition model is obtained through the following steps:
constructing a training sample set by utilizing the load identification type;
inputting the data in the training sample set into a load recognition model for training, and obtaining a pre-trained load recognition model based on a K-CV method in cross validation;
calculating a classification accuracy by using the pre-trained load identification model, and adjusting the pre-trained load identification model based on the classification accuracy and a particle swarm optimization algorithm to obtain a trained load identification model;
the method for constructing the training sample set by utilizing the load identification type specifically comprises the following steps:
generating a corresponding simulation data set from the preprocessed load data according to the load identification type;
dividing the simulation data set into a training set and a test set according to a preset proportion;
carrying out normalization processing on the training set and the test set to obtain a training sample set;
the synthesis method of the simulation data set comprises the following steps:
the load data of each home appliance may be expressed as:
q1(n)＝{q1(1),q1(2),...,q1(N1)}
q2(n)＝{q2(1),q2(2),...,q2(N2)}
......
qm(n)＝{qm(1),qm(2),...,qm(Nm)}
where m is the data setTotal number, NiThe number of sampling points of the equipment is q (i), and the sampling value of the apparent power of the household appliance is q (i);
synthesizing the household appliance load data, setting the household appliances corresponding to the synthesized household appliance data as m and n, wherein the synthesizing mode is as follows:
Sr(n)＝{s(1),s(2),...,s(br),s(br+1),...,s(br+ar),s(br+ar+1),...s(Nm-cr+br+1)}
＝{0,0,...,qm(cr),qm(cr+1),...,qm(cr+ar)+qn(1),qm(cr+ar+1)+qn(2),...,qm(Nm-cr+1)+qn(Nn-ar)}
wherein S isr(n) the synthesized simulated data set, s (i) the apparent power sample values of the simulated data set, brThe position of the moment when the household appliance m is turned on, arIs the distance from the location of the m starting time of the household appliance, crThe activated reading position is extracted for the home appliance n.
2. The method according to claim 1, wherein the generating the preprocessed load data into the corresponding simulation data set according to the load identification type specifically includes:
acquiring load data of a certain number of household appliances which are independently started in unit time, and screening the load data according to a preset rule to obtain generated simulation data;
synthesizing the simulation data set using the generated simulation data.
3. The method according to claim 1, wherein the obtaining of the pre-trained load recognition model based on the K-CV method in the cross validation specifically comprises:
equally dividing the training sample set into K training subsets;
and (3) making a primary verification set for each training subset, taking the rest K-1 training subsets as training sets, and respectively performing training verification to obtain K pre-trained load identification models.
4. The method according to claim 1, wherein the adjusting the pre-trained load recognition model based on the classification accuracy and the particle swarm optimization algorithm to obtain the trained load recognition model specifically comprises:
setting a parameter optimizing range, wherein the parameter optimizing range comprises a punishment parameter range and a kernel function parameter range;
optimizing and iterating the punishment parameters and the kernel function parameters, and calculating the classification accuracy after each iteration based on a K-CV method in cross validation;
and after the iteration times are reached, outputting a group of corresponding first punishment parameters and first kernel function parameters with the highest classification accuracy to obtain the trained load identification model corresponding to the first punishment parameters and the first kernel function parameters.
5. The method of claim 4, further comprising, after outputting a set of corresponding first penalty parameters and first kernel function parameters for which the classification accuracy is highest:
if the highest classification accuracy rate corresponds to multiple groups, outputting a first punishment parameter and a first kernel function parameter corresponding to one group with the smallest punishment parameter in the multiple groups;
and if the minimum punishment parameters correspond to multiple groups, outputting a group of first searched corresponding first punishment parameters and first kernel function parameters.
6. A household appliance load identification device based on a support vector machine is characterized by comprising:
the acquisition module is used for acquiring total load data of the started household appliance;
the identification module is used for inputting the total load data into a trained load identification model to obtain identification data;
the trained load recognition model is obtained through the following steps:
constructing a training sample set by utilizing the load identification type;
inputting the data in the training sample set into a load recognition model for training, and obtaining a pre-trained load recognition model based on a K-CV method in cross validation;
calculating a classification accuracy by using the pre-trained load identification model, and adjusting the pre-trained load identification model based on the classification accuracy and a particle swarm optimization algorithm to obtain a trained load identification model;
the method for constructing the training sample set by utilizing the load identification type specifically comprises the following steps:
generating a corresponding simulation data set from the preprocessed load data according to the load identification type;
dividing the simulation data set into a training set and a test set according to a preset proportion;
carrying out normalization processing on the training set and the test set to obtain a training sample set;
the synthesis method of the simulation data set comprises the following steps:
the load data of each home appliance may be expressed as:
q1(n)＝{q1(1),q1(2),...,q1(N1)}
q2(n)＝{q2(1),q2(2),...,q2(N2)}
......
qm(n)＝{qm(1),qm(2),...,qm(Nm)}
where m is the total number of data sets, NiThe number of sampling points of the equipment is q (i), and the sampling value of the apparent power of the household appliance is q (i);
synthesizing the household appliance load data, setting the household appliances corresponding to the synthesized household appliance data as m and n, wherein the synthesizing mode is as follows:
Sr(n)＝{s(1),s(2),...,s(br),s(br+1),...,s(br+ar),s(br+ar+1),...s(Nm-cr+br+1)}
＝{0,0,...,qm(cr),qm(cr+1),...,qm(cr+ar)+qn(1),qm(cr+ar+1)+qn(2),...,qm(Nm-cr+1)+qn(Nn-ar)}
wherein S isr(n) the synthesized simulated data set, s (i) the apparent power sample values of the simulated data set, brThe position of the moment when the household appliance m is turned on, arIs the distance from the location of the m starting time of the household appliance, crThe activated reading position is extracted for the home appliance n.
7. An electronic device, comprising a memory and a processor, wherein the processor and the memory communicate with each other via a bus; the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute a method for identifying a load of an electric appliance based on a support vector machine according to any one of claims 1 to 5.
8. A non-transitory computer readable storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements a method for identifying a load of an appliance based on a support vector machine according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810990146.3A CN109359665B (en) | 2018-08-28 | 2018-08-28 | Household appliance load identification method and device based on support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810990146.3A CN109359665B (en) | 2018-08-28 | 2018-08-28 | Household appliance load identification method and device based on support vector machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109359665A CN109359665A (en) | 2019-02-19 |
| CN109359665B true CN109359665B (en) | 2020-11-06 |
Family
ID=65350049
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810990146.3A Active CN109359665B (en) | 2018-08-28 | 2018-08-28 | Household appliance load identification method and device based on support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109359665B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110135452B (en) * | 2019-03-27 | 2023-08-08 | 广东浩迪创新科技有限公司 | Illegal electric appliance identification method and system based on intelligent electric meter |
| CN110504679B (en) * | 2019-07-25 | 2021-05-18 | 深圳供电局有限公司 | KM matching algorithm-based non-invasive load identification method |
| CN110414839A (en) * | 2019-07-29 | 2019-11-05 | 四川长虹电器股份有限公司 | Load recognition methods and system based on quantum genetic algorithm and SVM model |
| CN111242391B (en) * | 2020-03-06 | 2023-10-31 | 云南电网有限责任公司电力科学研究院 | Machine learning model training method and system for power load identification |
| CN112379159B (en) * | 2020-11-09 | 2021-10-26 | 北华航天工业学院 | Non-invasive household load decomposition method based on electric appliance operation mode |
| CN113076985B (en) * | 2021-03-29 | 2022-12-16 | 上海电力大学 | Power consumer load identification method and device and computer readable medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105956525A (en) * | 2016-04-22 | 2016-09-21 | 中国人民解放军总装备部军械技术研究所 | Static gesture recognizing method and system for AR auxiliary maintenance |
| CN106952183A (en) * | 2017-04-24 | 2017-07-14 | 武汉大学 | A kind of short-term load forecasting method based on particle group optimizing least square method supporting vector machine |
| CN107832705A (en) * | 2017-11-08 | 2018-03-23 | 天津大学 | A kind of non-intrusion type load recognition methods based on canonical correlation analysis |
-
2018
- 2018-08-28 CN CN201810990146.3A patent/CN109359665B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105956525A (en) * | 2016-04-22 | 2016-09-21 | 中国人民解放军总装备部军械技术研究所 | Static gesture recognizing method and system for AR auxiliary maintenance |
| CN106952183A (en) * | 2017-04-24 | 2017-07-14 | 武汉大学 | A kind of short-term load forecasting method based on particle group optimizing least square method supporting vector machine |
| CN107832705A (en) * | 2017-11-08 | 2018-03-23 | 天津大学 | A kind of non-intrusion type load recognition methods based on canonical correlation analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109359665A (en) | 2019-02-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109359665B (en) | Household appliance load identification method and device based on support vector machine | |
| CN111612651A (en) | Abnormal electric quantity data detection method based on long-term and short-term memory network | |
| Cheng et al. | Enhanced state estimation and bad data identification in active power distribution networks using photovoltaic power forecasting | |
| CN110416995B (en) | Non-invasive load decomposition method and device | |
| CN109257383B (en) | BGP anomaly detection method and system | |
| CN114970633B (en) | LSTM-based non-invasive electrical appliance identification method, system and equipment | |
| CN111932051A (en) | Malicious behavior detection method based on non-invasive power terminal time sequence monitoring | |
| CN112085111A (en) | Load identification method and device | |
| Liu et al. | A novel dual-attention optimization model for points classification of power quality disturbances | |
| CN115186012A (en) | Power consumption data detection method, device, equipment and storage medium | |
| CN116681186A (en) | Power quality analysis method and device based on intelligent terminal | |
| Azizi et al. | Quantification of disaggregation difficulty with respect to the number of smart meters | |
| CN113327008A (en) | Electricity stealing detection method, system and medium based on time sequence automatic encoder | |
| US20220243347A1 (en) | Determination method and determination apparatus for conversion efficiency of hydrogen production by wind-solar hybrid electrolysis of water | |
| CN109858667A (en) | It is a kind of based on thunder and lightning weather to the short term clustering method of loading effects | |
| CN111239484A (en) | Non-invasive load electricity consumption information acquisition method for non-resident users | |
| CN111090679A (en) | Time sequence data representation learning method based on time sequence influence and graph embedding | |
| CN116561569A (en) | Industrial power load identification method based on EO feature selection and AdaBoost algorithm | |
| CN116467631A (en) | Power fingerprint identification model training method, power fingerprint identification method and device | |
| CN212646862U (en) | Resident electricity consumption collection system | |
| Qin | Software reliability prediction model based on PSO and SVM | |
| CN110322063B (en) | Power consumption simulation prediction method and storage medium | |
| Xiong et al. | Research on Industrial Load Decomposition Based on Correlation Analysis and DNN | |
| Ucar et al. | Online power quality events detection using weighted Extreme Learning Machine | |
| CN118549823B (en) | Lithium battery electrical performance testing method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |