CN117541291A - An electricity price prediction method and system based on EMD decomposition and SSA-SVM model - Google Patents

Abstract

The invention discloses an electricity price prediction method and system based on EMD decomposition and SSA-SVM model. It relates to the technical field of electricity price prediction and includes obtaining original electricity price data, decomposing the electricity price data through the EMD algorithm; constructing an SSA-SVM model, and applying SSA Optimize the key parameters in the SVM model, perform predictions, and obtain the predicted value under each characteristic signal; accumulate and reconstruct the predicted values to obtain the final prediction result. The invention can effectively handle nonlinearity and non-stationarity, improve the accuracy of electricity price prediction, and better adapt to the real characteristics of electricity price data. It can better deal with outliers and fluctuations in electricity price data, improve the performance of the model through parameter optimization, and better fit the characteristics of electricity price data. It can better predict future electricity price trends and fluctuations, improve the utilization efficiency of energy resources, and achieve better economic benefits.

Description

A method and system for predicting electricity prices based on EMD decomposition and SSA-SVM model

Technical Field

The invention relates to the technical field of electricity price prediction, in particular to an electricity price prediction method and system based on EMD decomposition and SSA-SVM model.

Background Art

Based on artificial intelligence and data analysis technology, many electricity price prediction models have been established. Given that a single model often has certain limitations and precision errors, combining the characteristics of different algorithms to build a hybrid electricity price prediction model has become a research hotspot.

Common hybrid prediction methods include BP neural network optimized by genetic algorithm (GA-BP), ARIMA decomposed by wavelet transform (WT-ARIMA), LS-SVM optimized by particle swarm algorithm (PSO-LSSVM), LSTM network decomposed by wavelet packet (WPD-LSTM), hybrid Bayesian support vector machine method (BE-SVM), etc.

The above hybrid models still have shortcomings when dealing with non-stationary, nonlinear time series and complex electricity price data: GA-BP lacks generalization ability and may fall into local optimality, resulting in poor prediction performance; WT-ARIMA does not adequately handle nonlinear and non-stationary processing characteristics, and its high computational complexity may limit its prediction ability; although PSO-LSSVM can handle nonlinear problems to a certain extent, for highly nonlinear electricity price data, LS-SVM may still not be able to fully capture its inherent laws, thus affecting the prediction performance; the performance of the BE-SVM method depends to a large extent on the selection of parameters, which is difficult to tune and sensitive to noise. Although SVM has strong robustness, in electricity price forecasting, if the noise is large or there are outliers, it may affect the prediction performance of BE-SVM.

Summary of the invention

In view of the problems existing in the existing electricity price prediction methods based on EMD decomposition and SSA-SVM models, the present invention is proposed. By utilizing the advantages of EMD in processing nonlinear and non-stationary data, the implicit information of the electricity price series is fully mined, and the accuracy of electricity price prediction is improved by combining the strong adaptability of SSA and the good generalization performance of SVM. Therefore, the problem to be solved by the present invention is how to provide an electricity price prediction method and system based on EMD decomposition and SSA-SVM models.

In order to solve the above technical problems, the present invention provides the following technical solutions:

In the first aspect, an embodiment of the present invention provides an electricity price prediction method based on EMD decomposition and SSA-SVM model, which includes obtaining original electricity price data, dividing the original electricity price data into a training set and a test set, and decomposing the electricity price data through an empirical mode decomposition algorithm EMD; constructing an SSA-SVM model, applying a sparrow search algorithm SSA to optimize key parameters in the SVM model, inputting the decomposed electricity price data into the model for prediction, and obtaining a predicted value under each characteristic signal; accumulating and reconstructing the predicted values output by the model to obtain a final prediction result, and performing electricity price prediction based on EMD decomposition and SSA-SVM model.

As a preferred solution of the electricity price prediction method based on EMD decomposition and SSA-SVM model described in the present invention, the decomposition by the empirical mode decomposition algorithm EMD specifically includes determining all the maximum and minimum points on the original electricity price data signal X(t), using cubic spline interpolation to construct the upper envelope x _max (t) and lower envelope x _min (t) of the signal; calculating the envelope mean The relevant calculation formula is as follows:

The intermediate condition function h ¹ (t) of the original signal is calculated based on the original signal and the envelope mean. The relevant calculation formula is as follows:

If h ¹ (t) satisfies the IMF condition, the first characteristic component of IMF is obtained; if h ¹ (t) does not meet the IMF condition, the above calculation is repeated until the IMF condition is met and the IMF component is obtained. The relevant formula is as follows:

IMF ₁ (t) = h ¹ (t)

The IMF ₁ (t) is stripped from the signal to obtain the residual component r ₁ (t). The relevant calculation formula is as follows:

r ₁ (t) = x (t) - IMF ₁ (t)

Taking r ₁ (t) as the signal to be processed, the relevant calculation formula is as follows:

x(t)＝ _r1 (t)

Until it can no longer be decomposed, the final decomposition result is obtained. The relevant calculation formula is as follows:

In the formula, x(t) is the original electricity price data signal, and r _n (t) is the residual component, that is, the signal to be processed after the final decomposition is completed.

As a preferred solution of the electricity price prediction method based on EMD decomposition and SSA-SVM model described in the present invention, wherein: the construction of the SSA-SVM model specifically includes: constructing an SVM model, selecting the kernel function of the model, and determining the key parameters that need to be selected in the kernel function; optimizing the sparrow search algorithm SSA, and using the optimized sparrow search algorithm to determine the key parameters in the kernel function; inputting the signal to be processed after EMD decomposition into the sparrow search algorithm to obtain the prediction parameter result, substituting the obtained prediction parameter result into the SVM model for training, and after the model training is completed, inputting each signal to be processed into the model for prediction.

As a preferred solution of the electricity price prediction method based on EMD decomposition and SSA-SVM model described in the present invention, the SVM model specifically includes adding a Lagrangian function and obtaining the dual form of the original optimization problem according to the optimal conditions. The relevant expressions are as follows:

The constraints are:

Where _xi is the input vector, _yi is the output vector, l is the training sample, ε is the insensitive loss function, C is the penalty parameter, _αi is the Lagrange multiplier and _αi ≥ 0; the nonlinear mapping value φ is set to map the training sample to the high-dimensional space and perform linear regression, and the kernel function K(x,y) is introduced to realize nonlinear regression, and its optimization equation is transformed into:

K(x,x _i )=exp(-||xx _i || ² /2g ² )

Where _xi is the input vector and g is the prediction parameter.

As a preferred solution of the electricity price prediction method based on EMD decomposition and SSA-SVM model described in the present invention, the key parameters include the prediction parameter g and the penalty parameter C in the kernel function, and the key parameters in the kernel function are determined using the optimized sparrow search algorithm, specifically including: decomposing the original electricity price data sequence to obtain subsequences, intrinsic mode IMF and residual components as prediction inputs; normalizing each subsequence, and the relevant calculation formula is as follows:

Wherein, x′ _i is the normalized result of the i-th point of the subsequence, _xi is the electricity price of the i-th point of the subsequence, and x _max and x _min are the maximum and minimum electricity prices in the subsequence; the number of optimization parameters is determined to be 2, the number of iterations, the position vector of the sparrow population is initialized, the number of sparrow populations, the ratio of discoverers to warners, and the warning threshold are set; the fitness of each individual sparrow is calculated to obtain the current optimal position of the sparrow with a high fitness value; whether the position update condition is met, if so, the position is updated; if not, the current position is retained; whether the stop search condition is met, if so, the optimization is stopped to obtain the optimal solution.

As a preferred solution of the electricity price prediction method based on EMD decomposition and SSA-SVM model described in the present invention, the optimized sparrow search algorithm specifically includes initializing the population, the number of iterations, and the ratio of discoverers to followers. The corresponding expressions are as follows:

In the formula, n is the number of sparrow populations, and d represents the dimension of individual sparrows. The fitness of individual sparrows is calculated as follows:

In the formula, F _X is the fitness value of each row of sparrows; the discoverer with stronger search ability, that is, with better fitness value, will get food first during the search process. As an explorer, it can obtain a larger foraging search range than the follower; the position update of the discoverer during the iteration process is as follows:

In the formula, represents the position of the i-th sparrow in the j-dimension after iteration t, α represents a random number in the range of (0,1], T _max represents the maximum number of iterations, R ₂ ∈[0,1] represents the warning value, ST∈[0,1] represents the safety value, Q is a random number that obeys the normal distribution, L ₁ represents a 1×d column matrix, and all elements are 1; when R ₂ <ST, it means that there are no natural enemies around the sparrow, and the explorer conducts a global search; if R2 ≥ ST, it means that some sparrows have found the predator, and all sparrows take relevant actions to update their individual positions; update the follower position:

In the formula, represents the best position occupied by the producer in the j-th dimension at the (t+1)th iteration, represents the worst position in the entire population at the tth iteration; A represents a matrix in which each element is randomly assigned 1 or -1, and A ⁺ = ^AT (AA ^T ) ^-1 ; when i＞n/2, it indicates that the follower is in a very hungry state, and the product of a standard normal distribution random number and an exponential function with natural logarithm as the base is used to control its value to conform to the normal distribution, that is, to obtain more energy; when i≤n/2, a position is randomly found near the current optimal position; update the position of the vigilant:

In the formula, β represents the step size control coefficient of the normal distribution with a mean of 0 and a variance of 1, K represents a random number in the interval [-1,1], _fi represents the fitness value of the current sparrow individual, _fg represents the current global best fitness value, _fw represents the current global worst fitness value, and c is a constant; when the warning sparrow is in the current optimal position, it will escape to a position near itself. If the sparrow's current position is not the optimal position, it will escape to the vicinity of the current optimal position.

As a preferred solution of the electricity price prediction method based on EMD decomposition and SSA-SVM model described in the present invention, the prediction result includes modeling each subsequence separately, repeatedly obtaining their respective prediction results, accumulating and reconstructing all the prediction results to obtain the final electricity price prediction result; when evaluating the prediction performance of the model, the evaluation indicators include root mean square error, mean square error, deviation error, determination coefficient, mean absolute percentage error and Nash coefficient.

In the second aspect, an embodiment of the present invention provides an electricity price prediction system based on EMD decomposition and SSA-SVM model, which includes: an acquisition module, used to obtain original electricity price data, divide the original electricity price data into a training set and a test set, and decompose the electricity price data through the empirical mode decomposition algorithm EMD; construct a prediction module, construct an SSA-SVM model, apply the sparrow search algorithm SSA to optimize the key parameters in the SVM model, input the decomposed electricity price data into the model for prediction, and obtain the predicted value under each characteristic signal; an output module, used to accumulate and reconstruct the predicted values output by the model to obtain the final prediction result, and perform electricity price prediction based on EMD decomposition and SSA-SVM model.

In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: the processor implements any step of the above method when executing the computer program.

In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: the computer program implements any step of the above method when executed by a processor.

The beneficial effect of the present invention is that by using empirical mode decomposition EMD to decompose the original electricity price data, nonlinearity and non-stationarity can be effectively handled, thereby improving the accuracy of electricity price prediction. Each decomposed subsequence can be modeled according to its specific frequency component, so that the model can better adapt to the real characteristics of electricity price data. High-frequency noise components can be separated from electricity price data, thereby reducing the interference of noise on prediction. This helps to improve the robustness of the model, making it more capable of dealing with outliers and fluctuations in electricity price data. EMD decomposition allows different frequency components to be processed separately, which helps to extract important information from electricity price data. Analysis of different components can reveal trends, seasonality and periodicity in electricity price data, so that the model can better understand the reasons for electricity price fluctuations. The sparrow search algorithm SSA is used to optimize the key parameters in the support vector machine SVM model. Through parameter optimization, the performance of the model can be improved to ensure that it better fits the characteristics of electricity price data. The SSA-SVM model has high flexibility and can adapt to different types of electricity price data. Whether it is market price, consumer electricity price or other electricity price data type, it can be adapted to different scenarios through parameter adjustment. By decomposing and accumulating the forecast results, we can better understand the contribution of different frequency components in the electricity price data. It is helpful for time series analysis and can better predict future electricity price trends and fluctuations. High-precision electricity price forecasts provide strong support for energy market participants to optimize electricity procurement strategies, rationally plan electricity demand and reduce costs. By improving the accuracy of electricity price forecasts, electricity market participants can manage costs more effectively, reduce unnecessary expenses, improve the utilization efficiency of energy resources, and achieve better economic benefits.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

FIG1 is a flow chart of an electricity price prediction method based on EMD decomposition and SSA-SVM model.

Figure 2 is a flowchart of EMD decomposition.

Figure 3 is a diagram of the support vector regression structure.

Figure 4 is a flow chart of the sparrow optimization algorithm.

Figure 5 is a graph of historical electricity price data.

Figure 6 is a comparison of wholesale electricity price predictions of various algorithms.

Figure 7 is a comparison chart of retail electricity price predictions of various algorithms.

Fig. 8 Comparison of wholesale electricity price prediction indicators of various algorithms.

Figure 9 Comparison of retail electricity price prediction indicators of various algorithms.

Fig. 10 Box plot of wholesale electricity price prediction error of each algorithm.

Fig. 11 Box plot of retail electricity price prediction error of each algorithm.

DETAILED DESCRIPTION

In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

Example 1

Referring to FIG. 1 , which is a first embodiment of the present invention, the embodiment provides an electricity price prediction method based on EMD decomposition and SSA-SVM model, including:

S1: Obtain the original electricity price data, divide the original electricity price data into a training set and a test set, and decompose the electricity price data using the empirical mode decomposition algorithm EMD;

Specifically, the electricity price data is subjected to EMD decomposition, and the signal is decomposed into a number of intrinsic mode functions (IMFs) and residual components (Res);

Empirical Mode Decomposition (EMD) is a powerful tool for analyzing and processing nonlinear and non-stationary signals. It can decompose complex data signals into a series of basic, physically meaningful oscillating characteristic components, which are also called Intrinsic Mode Functions (IMFs). Each IMF is a component of the original signal, and together they constitute the overall structure of the original signal. Each IMF has a unique frequency characteristic, and these frequency characteristics change over time. This makes EMD very suitable for processing signals that change significantly on a time scale, such as power load signals and the power price signals involved in this invention, and many other practical signals in scientific research and engineering.

The decomposition process of EMD is an iterative process, which extracts the highest frequency IMF from the original signal in turn, and then subtracts it from the original signal to obtain a new residual signal. This process is repeated until the residual signal can no longer be decomposed or the preset stop condition is reached. Finally, each IMF and residual signal can be further transformed using Hilbert transform to obtain its instantaneous frequency and instantaneous amplitude over time. In this way, we can extract the main features of the signal and have a deeper understanding and analysis of the dynamic behavior and internal structure of the original signal.

The present invention uses EMD to decompose the electricity price signal in the following way:

All the maximum and minimum points on the original electricity price signal X(t) are determined, and the upper envelope x _max (t) and lower envelope x _min (t) of the signal are constructed using the cubic spline interpolation method.

The envelope mean can be calculated based on the upper and lower envelope lines

The intermediate condition function h ¹ (t) of the original signal is obtained based on the original signal and the envelope mean:

If h ¹ (t) satisfies the IMF condition, the first component of IMF is obtained; if h ¹ (t) does not satisfy the IMF condition, repeat steps 1)-2) until the IMF condition is satisfied, and the IMF component is obtained:

IMF ₁ (t) = h ¹ (t)

The IMF ₁ (t) is stripped from the signal to obtain the residual component r ₁ (t):

r ₁ (t) = x (t) - IMF ₁ (t)

Let r ₁ (t) be the signal to be processed:

x(t)＝ _r1 (t)

Repeat steps 1)-7) until no further decomposition is possible, and the final decomposition result is obtained:

S2: Construct the SSA-SVM model, apply the sparrow search algorithm SSA to optimize the key parameters in the SVM model, input the decomposed electricity price data into the model for prediction, and obtain the predicted value under each characteristic signal.

Support vector regression applied to regression problems is a powerful regression model based on the principle of interval maximization and kernel method. Support vector regression attempts to find a function so that the error between the predicted value and the true value does not exceed the preset threshold, and at the same time minimizes the complexity of the model. The advantage of support vector regression is that it can effectively handle linear and nonlinear regression problems. It can map the original feature space to a higher-dimensional space by introducing a kernel function and find the optimal regression function in this new feature space. Compared with logistic regression and neural networks, support vector machines provide a clearer and more powerful way to learn complex nonlinear equations.

In SVM, the minimum value of the precision ω in the linear regression function is obtained, and the norm of the Euclidean space is minimized, which is transformed into an optimization problem. The constrained optimization problem is not easy to solve because it involves a nonlinear optimization problem. By introducing the Lagrangian function and according to the optimal conditions, the dual form of the original optimization problem is obtained:

The constraints are:

Where _xi is the input vector, _yi is the output vector, l is the training sample, ε is the insensitive loss function, C is the penalty parameter, _αi is the Lagrange multiplier and _αi ≥0.

When solving nonlinear regression problems, it is necessary to set the nonlinear mapping value φ first, so that the training samples can be effectively mapped to the high-dimensional space and linear regression can be performed on it. Since the above solution can only calculate the inner product of the high-dimensional space, the kernel function K(x,y) can be introduced here to replace <φ(x),φ(x)> so that nonlinear regression can be achieved, and its optimization equation can be transformed into:

Choosing different kernel functions for SVM will affect the quality of the algorithm. The present invention chooses radial basis function (RBF) as the kernel function:

K(x,x _i )=exp(-||xx _i || ² /2g ² )

Electricity prices are related to multiple factors (such as weather conditions, supply and demand, coal prices, etc.), and these relationships are often nonlinear. The RBF kernel can effectively handle nonlinear problems and has a certain robustness to some noise data. When the parameters are selected appropriately, the RBF kernel SVM has a good generalization ability and can also show good prediction performance on unseen data. After determining the kernel function, the parameters g and penalty parameter C of the radial basis function (RBF) are selected. The support vector regression structure is shown in Figure 2.

Sparrow Search Algorithm (SSA) is an intelligent optimization algorithm proposed in 2020. The design of this algorithm is inspired by the foraging behavior and anti-predation behavior of sparrows. In the foraging process of sparrows, the sparrow population can be divided into two roles: "finders" and "followers". The "finders" are responsible for finding food in the environment and providing foraging areas and directions for the entire sparrow population, while the "followers" rely on the "finders" to obtain food.

In the SSA algorithm, the behaviors of these two types of sparrows are simulated as two search strategies to find the optimal solution in the solution space of the optimization problem. At the same time, an early warning mechanism is also introduced in the algorithm. In each iteration, a part of the sparrow individuals will be selected for early warning. If they find that the current solution may cause the optimization process to fall into a local optimum or other undesirable situations, they can choose to abandon the current solution and explore other areas of the solution space. This makes it have good global optimization capabilities and adaptability. The specific implementation steps are as follows:

Initialize the population, number of iterations, and the ratio of discoverers to followers

Where: n is the number of sparrow populations, and d represents the dimension attached to the individual sparrows.

Calculating fitness

Where: F _X is the fitness value of each row of sparrows

A discoverer with a stronger search ability, i.e. a better fitness value, will have priority in obtaining food during the search process. As an explorer, it can obtain a larger foraging search range than a follower.

The finder's position is updated during the iteration as follows:

Where: represents the position of the i-th sparrow in the j-dimension at the t-th iteration, α represents a random number in the range of (0,1], T _max represents the maximum number of iterations, R ₂ ∈[0,1] represents the warning value, ST∈[0,1] represents the safety value, Q is a random number that obeys the normal distribution, and L ₁ represents a 1×d-column matrix with all elements being 1.

When _R2 <ST, it means there are no natural enemies around and the explorer can conduct a global search. If R2≥ST, it means some sparrows have discovered the predator and all sparrows must take relevant actions.

Update follower location:

In the formula, represents the best position occupied by the producer in the j-th dimension at the (t+1)th iteration, represents the worst position in the entire population at the t-th iteration; A represents a matrix (1 row and d columns) in which each element is randomly assigned 1 or -1, and A ⁺ = ^AT (AA ^T ) ^-1 .

When i＞n/2, it indicates that the pursuer is in a very hungry state. The product of a standard normal distribution random number and an exponential function with natural logarithm as the base is used to control its value to conform to the normal distribution, that is, to obtain more energy. When i≤n/2, the process can be explained as randomly finding a position near the current optimal position, and the variance of each dimension according to the optimal position is small and the value is relatively stable.

Updated Sentinel location:

Where: β represents the step size control coefficient that obeys the normal distribution with a mean of 0 and a variance of 1, K represents a random number in the interval [-1,1], _fi represents the fitness value of the current sparrow individual, _fg represents the current global best fitness value, _fw represents the current global worst fitness value, and c represents a constant, which is set to avoid the denominator being zero.

When the warning sparrow is in the current optimal position, it will escape to a position near itself. If it is not the optimal position, it will escape to the vicinity of the current optimal position.

Use SSA to search for the best values of g and C, specifically,

The original electricity price series is decomposed by EMD to obtain the intrinsic mode IMF and a residual component IMF-Res as its subsequence, which is used as the input for subsequent predictions.

Each subsequence is normalized, and the standardized value is between [0,1]. The calculation method is to subtract the data from the minimum value of the column and then divide it by the range:

Where: _x′i is the normalized result of the i-th point in the subsequence, _xi is the electricity price of the i-th point in the subsequence, and _xmax and _xmin are the maximum and minimum electricity prices in the subsequence.

Determine the number of optimization parameters as 2, the number of iterations, initialize the sparrow population position vector within the C and g empirical value range, set the sparrow population size, the ratio of discoverers to warners, and the warning threshold.

Calculate the fitness of each individual sparrow and get the best position of the current sparrow with high fitness value.

Determine whether the location update conditions are met. If so, update the location; if not, keep the current location.

Determine whether the stop search condition is met. If so, stop searching to obtain the optimal solution.

The obtained optimal parameter g and penalty parameter C are substituted into the support vector machine model for training.

Each subsequence is modeled separately and the respective prediction results are obtained repeatedly. Then all the prediction results are accumulated and reconstructed to obtain the final electricity price prediction result. The prediction results are judged by calculating indicators such as RMSE, R2, MAE, and MBE.

Construct an SVM model, select the kernel function of the model, and determine the key parameters that need to be selected in the kernel function; optimize the sparrow search algorithm SSA, and use the optimized sparrow search algorithm to determine the key parameters in the kernel function; input the signal to be processed after EMD decomposition into the sparrow search algorithm to obtain the prediction parameter results, and substitute the obtained prediction parameter results into the SVM model for training. After the model training is completed, the obtained signals to be processed are input into the model for prediction.

S3: The predicted values output by the model are accumulated and reconstructed to obtain the final prediction result, and the electricity price prediction is performed based on EMD decomposition and SSA-SVM model.

Furthermore, this embodiment also provides an electricity price prediction system based on EMD decomposition and SSA-SVM model, including: an acquisition module, used to obtain original electricity price data, divide the original electricity price data into a training set and a test set, and decompose the electricity price data through the empirical mode decomposition algorithm EMD; construct a prediction module, construct an SSA-SVM model, apply the sparrow search algorithm SSA to optimize the key parameters in the SVM model, input the decomposed electricity price data into the model for prediction, and obtain the predicted value under each characteristic signal; an output module, used to accumulate and reconstruct the predicted values output by the model to obtain the final prediction result, and perform electricity price prediction based on EMD decomposition and SSA-SVM model.

This embodiment also provides a computer device, which is suitable for the case of an electricity price prediction method based on EMD decomposition and SSA-SVM model, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement all or part of the steps of the method described in the embodiment of the present invention as proposed in the above embodiment.

This embodiment also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the method in any optional implementation of the above embodiment is executed. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

As can be seen from the above, the present invention can effectively handle nonlinearity and non-stationarity, improve the accuracy of electricity price prediction, and can model according to specific frequency components to better adapt to the real characteristics of electricity price data by decomposing the original electricity price data using empirical mode decomposition EMD. Separate the high-frequency noise components from the electricity price data to reduce the interference of noise on the prediction. It helps to improve the robustness of the model and better cope with outliers and fluctuations in electricity price data. Allowing different frequency components to be processed separately helps to extract important information from electricity price data. The analysis of different components can reveal the trends, seasonality and periodicity in electricity price data, so that the model can better understand the reasons for electricity price fluctuations. Through parameter optimization, the performance of the model can be improved to ensure better fit to the characteristics of electricity price data. It can adapt to different scenarios through parameter adjustment, and the contribution of different frequency components in electricity price data can be better understood by decomposing and accumulating the prediction results. It is helpful for time series analysis and can better predict future electricity price trends and fluctuations. High-precision electricity price forecasts provide strong support for energy market participants, optimize electricity procurement strategies, rationally plan electricity demand and reduce costs. By improving the accuracy of electricity price forecasts, electricity market participants can manage costs more effectively, reduce unnecessary expenses, improve the utilization efficiency of energy resources, and achieve better economic benefits.

Example 2

2 to 11 , a second embodiment of the present invention is shown. This embodiment provides an electricity price prediction method based on EMD decomposition and SSA-SVM model. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

Taking the wholesale electricity price and retail electricity price of the power market in Guizhou Province as an example, the historical electricity price data from January 2021 to June 2023 were collected, and at the same time, the external factors affecting the electricity price - the grid-connected electricity volume and the coal price were input as feature quantities into the SVM, POS-SVM, EMD-SVM and EMD-SSA-SVM prediction models of the present invention for prediction comparison. The root mean square error (RSME), mean square error (MAE), bias error (MBE), mean absolute percentage error (MAPE), determination coefficient (R2), and Nash coefficient (NSE) were used as evaluation indicators to compare the prediction results of each model. The specific indicators are shown in the following table:

Table 1 Comparison of wholesale electricity price prediction indicators of various models

Judging from the forecast results of wholesale electricity prices, the EMD-SSA-SVM forecasting model is smaller than the other three models in RMSE, MAE, MBE, and MAPE, indicating that it has the lowest sensitivity to large errors, the smallest average of the absolute value of the forecast error, and the smallest percentage of the forecast error, and the highest forecast accuracy.

Table 2 Comparison of retail electricity price prediction indicators of various models

Judging from the retail electricity price forecast results, EMD-SVR and EMD-SSA-SVR have the lowest RMSE. The average of the square of the prediction errors of these two models is the smallest, and they are less sensitive to large errors.

EMD-SSA-SVR has the lowest MAE, the smallest average of the absolute value of its prediction error, and its MBE value is also closer to zero, indicating that the deviation between its predicted value and the true value is the smallest. The MAPE of all models is very close, but the MAPE of EMD-SSA-SVR is the smallest, indicating that its relative prediction error percentage is the smallest and the prediction accuracy is the highest. In general, the prediction results of EMD-SSA-SVR are more accurate and have better prediction performance. From the box plot of the prediction error, it can also be seen that the distance between the quartiles of the EMD-SSA-SVM prediction error is the smallest, the distribution of the error is concentrated, and the stability is good. The median is also closest to zero, indicating that the error deviation is small and the prediction is more accurate. This method can effectively improve the prediction accuracy of electricity prices.

As can be seen from the above, the present invention can effectively handle nonlinearity and non-stationarity, improve the accuracy of electricity price prediction, and can model according to specific frequency components to better adapt to the real characteristics of electricity price data by decomposing the original electricity price data using empirical mode decomposition EMD. Separate the high-frequency noise components from the electricity price data to reduce the interference of noise on the prediction. It helps to improve the robustness of the model and better cope with outliers and fluctuations in electricity price data. Allowing different frequency components to be processed separately helps to extract important information from electricity price data. The analysis of different components can reveal the trends, seasonality and periodicity in electricity price data, so that the model can better understand the reasons for electricity price fluctuations. Through parameter optimization, the performance of the model can be improved to ensure better fit to the characteristics of electricity price data. It can adapt to different scenarios through parameter adjustment, and the contribution of different frequency components in electricity price data can be better understood by decomposing and accumulating the prediction results. It is helpful for time series analysis and can better predict future electricity price trends and fluctuations. High-precision electricity price forecasts provide strong support for energy market participants, optimize electricity procurement strategies, rationally plan electricity demand and reduce costs. By improving the accuracy of electricity price forecasts, electricity market participants can manage costs more effectively, reduce unnecessary expenses, improve the utilization efficiency of energy resources, and achieve better economic benefits.

It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims (10)

1. A method for predicting electricity prices based on EMD decomposition and SSA-SVM model, characterized by: comprising: Obtain the original electricity price data, divide the original electricity price data into a training set and a test set, and decompose the electricity price data using the empirical mode decomposition algorithm EMD; Construct the SSA-SVM model, apply the sparrow search algorithm SSA to optimize the key parameters in the SVM model, input the decomposed electricity price data into the model for prediction, and obtain the predicted value under each characteristic signal; The predicted values output by the model are accumulated and reconstructed to obtain the final prediction result, and electricity price prediction is performed based on EMD decomposition and SSA-SVM model. 2. The electricity price prediction method based on EMD decomposition and SSA-SVM model according to claim 1, characterized in that: the decomposition by the empirical mode decomposition algorithm EMD specifically includes: Determine all the maximum and minimum points on the original electricity price data signal X(t), and use cubic spline interpolation to construct the upper envelope x _max (t) and lower envelope x _min (t) of the signal; Calculate the envelope mean The relevant calculation formula is as follows: The intermediate condition function h ¹ (t) of the original signal is calculated based on the original signal and the envelope mean. The relevant calculation formula is as follows: If h ¹ (t) satisfies the IMF condition, the first characteristic component of the IMF is obtained; If h ¹ (t) does not meet the IMF condition, repeat the above calculation until it meets the IMF condition and obtain the IMF component. The relevant formula is as follows: IMF ₁ (t) = h ¹ (t) The IMF ₁ (t) is stripped from the signal to obtain the residual component r ₁ (t). The relevant calculation formula is as follows: r ₁ (t) = x (t) - IMF ₁ (t) Taking r ₁ (t) as the signal to be processed, the relevant calculation formula is as follows: x(t)＝ _r1 (t) Until it can no longer be decomposed, the final decomposition result is obtained. The relevant calculation formula is as follows: In the formula, x(t) is the original electricity price data signal, and r _n (t) is the residual component, that is, the signal to be processed after the final decomposition is completed. 3. The electricity price prediction method based on EMD decomposition and SSA-SVM model according to claim 2, characterized in that: the construction of the SSA-SVM model specifically includes: Build the SVM model, select the kernel function of the model, and determine the key parameters that need to be selected in the kernel function; Optimize the sparrow search algorithm SSA, and use the optimized sparrow search algorithm to determine the key parameters in the kernel function; The signal to be processed after EMD decomposition is input into the sparrow search algorithm to obtain the prediction parameter results, and the obtained prediction parameter results are substituted into the SVM model for training. After the model training is completed, the obtained signals to be processed are input into the model for prediction. 4. The electricity price prediction method based on EMD decomposition and SSA-SVM model according to claim 3, characterized in that: the SVM model specifically includes: By adding the Lagrangian function and according to the optimal conditions, we can get the dual form of the original optimization problem. The relevant expressions are as follows: The constraints are: Where _xi is the input vector, _yi is the output vector, l is the training sample, ε is the insensitive loss function, C is the penalty parameter, _αi is the Lagrange multiplier and _αi ≥ 0; The nonlinear mapping value φ is set to map the training samples to the high-dimensional space and perform linear regression. The kernel function K(x, y) is introduced to realize nonlinear regression, and its optimization equation is transformed into: K(x,x _i )=exp(-||xx _i || ² /2g ² ) Where _xi is the input vector and g is the prediction parameter. 5. The electricity price prediction method based on EMD decomposition and SSA-SVM model according to claim 4, characterized in that: the key parameters include the prediction parameter g and the penalty parameter C in the kernel function, and the key parameters in the kernel function are determined using the optimized sparrow search algorithm, specifically including: The subsequences, intrinsic mode IMF and residual components obtained by decomposing the original electricity price data sequence are used as the input for prediction; Each subsequence is normalized, and the relevant calculation formula is as follows: In the formula, x′ _i is the normalized result of the i-th point in the subsequence, _xi is the electricity price of the i-th point in the subsequence, x _max and x _min are the maximum and minimum electricity prices in the subsequence; Determine the number of optimization parameters as 2, the number of iterations, initialize the sparrow population position vector, set the number of sparrow populations, the ratio of discoverers to alarms, and the alarm threshold; Calculate the fitness of each individual sparrow and obtain the best position of the current sparrow with a high fitness value; Determine whether the location update conditions are met. If so, update the location; if not, keep the current location; Determine whether the stop search condition is met. If so, stop searching and get the optimal solution. 6. The electricity price prediction method based on EMD decomposition and SSA-SVM model according to claim 5, characterized in that: the optimized sparrow search algorithm specifically includes: Initialize the population, number of iterations, and the ratio of discoverers to followers. The corresponding expressions are as follows: In the formula, n is the number of sparrow populations, and d represents the dimension attached to individual sparrows; Calculate the individual fitness of sparrows. The relevant expression is as follows: In the formula, F _X is the fitness value of each row of sparrows; A discoverer with a stronger search ability, i.e. a better fitness value, will have priority in obtaining food during the search process. As an explorer, they can obtain a larger foraging search range than followers; The finder's position is updated during the iteration as follows: In the formula, represents the position of the i-th sparrow in the j-dimension at the t-th iteration, α represents a random number in the range of (0,1], T _max represents the maximum number of iterations, R ₂ ∈[0,1] represents the warning value, ST∈[0,1] represents the safety value, Q is a random number that obeys the normal distribution, L ₁ represents a 1×d-column matrix, and all elements are 1; When R ₂ <ST, it means there are no natural enemies around the sparrow, and the explorer conducts a global search; If R2 ≥ ST, it means that some sparrows have discovered the predator, and all sparrows take relevant actions and update their individual positions; Update follower location: In the formula, represents the best position occupied by the producer in the j-th dimension at the (t+1)th iteration, represents the worst position in the entire population at the t-th iteration; A represents a matrix in which each element is randomly assigned 1 or -1, and A ⁺ = ^AT (AA ^T ) ^-1 ; When i＞n/2, it indicates that the follower is in a very hungry state. The product of a standard normal distribution random number and an exponential function with natural logarithm as the base is used to control its value to conform to the normal distribution, that is, to obtain more energy; when i≤n/2, a position is randomly found near the current optimal position; Updated Sentinel location: In the formula, β represents the step size control coefficient that obeys the normal distribution with a mean of 0 and a variance of 1, K represents a random number in the interval [-1,1], _fi represents the fitness value of the current sparrow individual, _fg represents the current global best fitness value, _fw represents the current global worst fitness value, and c is a constant; When the warning sparrow is in the current optimal position, it will escape to a position near itself. If the sparrow's current position is not the optimal position, it will escape to the vicinity of the current optimal position. 7. The electricity price prediction method based on EMD decomposition and SSA-SVM model as described in claim 6 is characterized in that: the prediction result includes modeling each subsequence separately, repeatedly obtaining their own prediction results, accumulating and reconstructing all prediction results to obtain the final electricity price prediction result; when evaluating the model prediction performance, the evaluation indicators include root mean square error, mean square error, bias error, determination coefficient, mean absolute percentage error and Nash coefficient. 8. An electricity price prediction system based on EMD decomposition and SSA-SVM model, based on the electricity price prediction method based on EMD decomposition and SSA-SVM model according to any one of claims 1 to 7, characterized in that: it comprises: An acquisition module is used to acquire original electricity price data, divide the original electricity price data into a training set and a test set, and decompose the electricity price data using an empirical mode decomposition algorithm (EMD); Construct a prediction module, build an SSA-SVM model, apply the sparrow search algorithm SSA to optimize the key parameters in the SVM model, input the decomposed electricity price data into the model for prediction, and obtain the predicted value under each characteristic signal; The output module is used to accumulate and reconstruct the predicted values output by the model to obtain the final prediction result, and to perform electricity price prediction based on EMD decomposition and SSA-SVM model. 9. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the electricity price prediction method based on EMD decomposition and SSA-SVM model as claimed in any one of claims 1 to 7 are implemented. 10. A computer-readable storage medium having a computer program stored thereon, characterized in that: when the computer program is executed by a processor, the steps of the electricity price prediction method based on EMD decomposition and SSA-SVM model according to any one of claims 1 to 7 are implemented.