Power load demand prediction method and device and electronic equipment

Abstract

The invention discloses a power load demand prediction method and device and electronic equipment. The method comprises the following steps: acquiring power load demand data of a target area in a current sampling time period; performing variational mode decomposition on the power load demand data to obtain a plurality of target components; based on the plurality of target components, adopting a power load demand prediction model to obtain a plurality of predicted load demands, the plurality of predicted load demands being in one-to-one correspondence with the plurality of target components; and obtaining a power load demand prediction result of the target area based on the plurality of predicted load demands. According to the invention, the technical problem of low prediction accuracy of a power load prediction method in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of smart grids, and in particular to a method, device and electronic equipment for predicting power load demand. Background Art

[0002] The stable operation and efficient dispatching of power systems are crucial to ensuring power supply, promoting economic development and social life. As a key link in power system management, power load demand forecasting is of great significance to power dispatching, load management, power market transactions and power grid planning. With the widespread application of renewable energy and the development of smart grid technology, the characteristics and forecasting difficulty of power load have changed significantly. The forecasting methods in related technologies are often difficult to adapt to this complex and changeable power load pattern, resulting in low accuracy of power load forecasting results.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present invention provide a method, device and electronic device for predicting electric load demand, so as to at least solve the technical problem of low prediction accuracy existing in the electric load prediction method in the related art.

[0005] According to one aspect of an embodiment of the present invention, there is provided a method for predicting electric load demand, comprising: obtaining electric load demand data of a target area in a current sampling period; performing variational mode decomposition on the electric load demand data to obtain a plurality of target components; based on the plurality of target components, using an electric load demand prediction model to obtain a plurality of predicted load demands, wherein the plurality of predicted load demands correspond one-to-one to the plurality of target components; based on the plurality of predicted load demands, obtaining an electric load demand prediction result for the target area.

[0006] According to another aspect of an embodiment of the present invention, there is also provided an apparatus for predicting electric load demand, comprising: a data acquisition module for acquiring electric load demand data of a target area in a current sampling period; a variational mode decomposition module for performing variational mode decomposition on the electric load demand data to obtain a plurality of target components; a first prediction module for obtaining a plurality of predicted load demands based on the plurality of target components and adopting a power load demand prediction model, wherein the plurality of predicted load demands correspond one-to-one to the plurality of target components; and a second prediction module for obtaining a prediction result of the electric load demand of the target area based on the plurality of predicted load demands.

[0007] According to another aspect of an embodiment of the present invention, there is also provided an electronic device, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the methods for predicting electric power load demand.

[0008] In an embodiment of the present invention, the power load demand data of the target area in the current sampling period is obtained; the power load demand data is subjected to variational mode decomposition to obtain multiple target components; based on the multiple target components, a power load demand prediction model is used to obtain multiple predicted load demands, wherein the multiple predicted load demands correspond one-to-one to the multiple target components; based on the multiple predicted load demands, the power load demand prediction result of the target area is obtained, thereby achieving the purpose of accurately predicting the power load demand based on the multiple target components obtained by performing variational mode decomposition on the power load demand data and using the power load demand prediction model, thereby achieving the technical effect of improving the accuracy of power load demand prediction, and further solving the technical problem of low prediction accuracy existing in the power load prediction method in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0010] Figure 1 is a flow chart of a method for predicting power load demand according to an embodiment of the present invention;

[0011] Figure 2 is a schematic diagram of a power load demand prediction device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0012] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0013] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0014] First, to facilitate understanding of the embodiments of the present invention, some terms or nouns involved in the present invention are explained below:

[0015] Variational Mode Decomposition (VMD) is a signal processing method that is mainly used to decompose complex signals into a set of physically meaningful and simpler intrinsic mode functions (IMFs).

[0016] Harris' hawk optimization algorithm (HEO), an optimization algorithm based on the foraging behavior of hawks, simulates the collaborative behavior and strategy of Harris' hawks when hunting, and finds the optimal solution through cooperation and competition among individuals.

[0017] Flower pollination optimization algorithm (PSO), an optimization algorithm based on the plant pollination process, simulates the pollination process between flowers and searches for the optimal solution through information communication and exchange between flowers.

[0018] Differential Evolution algorithm (DE), an optimization algorithm based on evolutionary algorithm, optimizes the solution of problems by simulating the evolutionary process in nature, including individual mutation, crossover and selection operations to find the optimal solution.

[0019] According to an embodiment of the present invention, a method embodiment of power load demand forecasting is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0020] Figure 1 is a flow chart of a method for predicting power load demand according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0021] Step S102, obtaining the power load demand data of the target area in the current sampling period;

[0022] In step S102 of the present application, the power load demand data is used to indicate the record of the power demand of the target area (or users in the target area) for the power system during the current sampling period. The power load demand data can be collected in real time or periodically by smart meters, data acquisition and monitoring systems, distribution automation systems or other power monitoring equipment.

[0023] Step S104, performing variational mode decomposition on the power load demand data to obtain multiple target components;

[0024] In step S104 of the present application, a complex signal can be decomposed into multiple simpler components with specific frequency characteristics through variational mode decomposition (VMD). For example, the power load demand data can be decomposed into multiple intrinsic mode functions (IMF). For the power load demand data, VMD can decompose it into multiple target components, each of which represents a mode or frequency characteristic in the power load demand data. This decomposition helps the subsequent prediction model to more accurately capture and analyze the fluctuation characteristics of the power load.

[0025] Step S106, based on the multiple target components, using the power load demand forecasting model to obtain multiple predicted load demands, wherein the multiple predicted load demands correspond to the multiple target components one by one;

[0026] In step S106 of the present application, based on the target components decomposed by VMD, the power load demand forecasting model can more accurately predict the future load demand of each component. Since VMD can separate different frequency modes in the power load data, the power load demand forecasting model can make predictions based on the characteristics of each mode, thereby improving the accuracy of the prediction. The prediction results of each target component are independent and match the characteristics of the component.

[0027] Step S108, based on the multiple predicted load demands, obtain the power load demand prediction result of the target area.

[0028] In step S108 of the present application, the predicted load demand results of each target component are integrated to obtain the final power load demand prediction result. This integration takes into account the various characteristics of the power load and can provide a more comprehensive and accurate prediction value.

[0029] Optionally, the power load demand forecast result of the target area can be obtained by, but not limited to, summing up multiple predicted load demands, performing weighted calculations, etc.

[0030] Through the above steps S102 to S108, the purpose of accurately predicting the power load demand by using the power load demand prediction model based on the variational mode decomposition of the power load demand data to obtain multiple target components can be achieved, thereby achieving the technical effect of improving the accuracy of power load demand prediction, and further solving the technical problem of low prediction accuracy existing in the power load prediction method in the related technology.

[0031] In an optional embodiment, variational modal decomposition is performed on power load demand data to obtain multiple target components, including: using a target variational modal decomposition model to perform variational modal decomposition on power load demand data to obtain multiple target components; wherein the target variational modal decomposition model is obtained by optimizing parameters of an initial variational modal decomposition model based on multiple algorithms.

[0032] Optionally, by adopting a target variational modal decomposition model optimized based on multiple algorithms, the power load data can be more accurately decomposed into different frequency components, i.e., multiple target components. This is because the optimized model can better set key parameters (such as Lagrange penalty parameter α, decomposition mode number K), thereby avoiding modal aliasing and improving the clarity and accuracy of the decomposition results. The optimization process enables the variational modal decomposition model to better adapt to the characteristics of power load demand data in a specific area. The combined use of different algorithms (such as Harris Hawk optimization algorithm, flower pollination optimization algorithm, differential evolution algorithm, etc.) can make up for the shortcomings of a single algorithm, improve the robustness and adaptability of the model, and make it perform better when processing complex and changeable power load data. The multiple target components obtained by decomposition are closer to the real mode of power load demand data, which provides higher quality input for the subsequent use of prediction models for power load demand prediction. So that the power load demand prediction model can more accurately capture the prediction law of each component, thereby improving the accuracy and reliability of the overall prediction results.

[0033] In an optional embodiment, before using the target variational modal decomposition model to perform variational modal decomposition on the power load demand data to obtain multiple target components, the method also includes: obtaining the power load demand data of the target area in the historical sampling period to obtain a power load demand data set; based on the power load demand data set, using multiple algorithms to optimize the parameters of the initial variational modal decomposition model to obtain the target variational modal decomposition model, wherein the multiple algorithms include at least: Harris Hawk optimization algorithm, flower pollination optimization algorithm, differential evolution algorithm.

[0034] Optionally, by collecting the power load demand data of the target area during the historical sampling period, a power load demand data set is formed. This data set contains information about past load patterns and is an important basis for training and optimizing models. Based on these historical data, multiple algorithms are used to optimize the parameters of the initial variational mode decomposition model, and the most suitable parameter settings for the load characteristics of the region can be found, thereby improving the accuracy of decomposition and the reliability of prediction. By using multiple algorithms to optimize the parameters of the initial variational mode decomposition model, these algorithms have different search strategies and optimization capabilities, and can optimize the model parameters from multiple angles. For example: the Harris Hawk optimization algorithm can simulate the hunting behavior of the eagle, and can perform global and local searches to find the optimal solution in the parameter space. The flower pollination optimization algorithm can update and optimize parameters based on the pollination process of flowers through global pollination and local pollination behaviors, and is suitable for solving complex optimization problems. The differential evolution algorithm can mutate through individual differences in the population, and can quickly explore the parameter space to avoid falling into the local optimal solution. The combined use of these algorithms can make up for the limitations of a single algorithm, improve the efficiency and effect of parameter optimization, and thus obtain a better target variational modal decomposition model. The target variational modal decomposition model obtained through optimization can more effectively separate signal components of different frequencies and modes when decomposing power load demand data. This is crucial for the subsequent training of power demand forecasting models, which can reduce the prediction errors caused by modal aliasing and improper parameter settings and improve the overall prediction accuracy.

[0035] In an optional embodiment, based on the power load demand data set, multiple algorithms are used to optimize the parameters of the initial variational modal decomposition model to obtain a target variational modal decomposition model, including: initializing a group of individuals representing the variational modal decomposition model parameters in a first search space as the current population, wherein the variational modal decomposition model parameters include at least: a Lagrange penalty parameter and a number of decomposition modes; based on the power load demand data set and the current population, using the Harris Hawk optimization algorithm to find the optimal solution in the first search space; in the process of finding the optimal solution in the first search space, updating the positions of the individuals included in the current population based on the flower pollination optimization algorithm, and updating the individuals included in the current population based on the differential evolution algorithm; repeating the above operations until a preset first termination condition is reached; determining the target variational modal decomposition model based on the individuals included in the current population when the preset first termination condition is reached.

[0036] Optionally, by initializing a group of individuals representing VMD model parameters (such as Lagrange penalty parameters, decomposition modal number) in the first search space and using them as the current population, it is possible to systematically explore and optimize in the parameter space, avoid blind selection or rely on experience to set parameters, thereby improving the decomposition effect and prediction accuracy of the model. The Harris Hawk optimization algorithm is used to find the optimal solution in the first search space. The algorithm can simulate the hunting behavior of the eagle, combine global search and local search strategies, effectively explore the parameter space, and quickly locate the optimal solution. This helps to improve the efficiency of parameter optimization and shorten the time of model training and optimization. In the process of finding the optimal solution, the flower pollination optimization algorithm and the differential evolution algorithm are combined to update the position and individual update of the individuals included in the current population. The flower pollination optimization algorithm can promote the global optimization of parameters, while the differential evolution algorithm helps to explore the differences between individuals and avoid the trap of local optimal solutions. This collaborative optimization strategy can make full use of the advantages of different algorithms to improve the comprehensiveness and accuracy of optimization. The termination condition can be, but is not limited to, reaching the maximum number of iterations, meeting a specific error threshold or fitness standard, etc. By setting the termination condition, it is possible to avoid the waste of computing resources caused by over-optimization, while ensuring that the optimization process achieves the desired effect within a limited time. When the termination condition is reached, the target variational mode decomposition model is determined based on the individuals included in the current population. These individuals represent the most suitable parameter settings for the power load demand data set after optimization, so that the target variational mode decomposition model has higher decomposition accuracy and prediction ability. This model construction method based on the determination of the optimal solution can ensure that the model can more accurately capture and decompose the different frequency components in the signal when processing actual power load data.

[0037] In the above methods, through the systematic parameter optimization process, combined with the Harris Eagle optimization algorithm, the flower pollination optimization algorithm and the differential evolution algorithm, the parameter optimization effect of the variational mode decomposition model can be effectively improved, thereby improving the accuracy and reliability of power load demand forecasting. This method can not only accelerate the training and optimization of the model, but also ensure that the model maintains a high prediction performance when facing complex and changeable power load data, providing strong technical support for the dispatch and planning of the power system.

[0038] Optionally, in the process of searching for the optimal solution in the first search space using the Harris Hawk optimization algorithm based on the power load demand data set and the current population, the exploration phase of the variational mode decomposition algorithm can be adjusted and optimized based on the Harris Hawk optimization algorithm. The eagle in the Harris Hawk optimization algorithm represents the particles included in the current population. The specific formula is as follows:

[0039]

[0040] Among them, X(t+1) is the next iteration position, X(t) is the current eagle position, and X m (t) is the average position of all eagles in the current population, X targeted (t) is the target eagle, X rabbit (t) is the current optimal solution, r1, r2, r3, r4, q are 5 random numbers evenly distributed from 0 to 1, UB is the upper limit of the search space, LB is the lower limit of the search space, N is the population of Harris hawks, X Ni is the set of i hawks selected from N Harris hawks.

[0041] Optionally, the global pollination and local pollination behaviors of the flower pollination optimization algorithm can be used to replace the progressive fast dive and soft surround behaviors of the variational mode decomposition algorithm to update the positions of the individuals included in the current population. The specific formula is as follows:

[0042] Y=X(t)+LF(X rabbit (txt))

[0043] Z=X(t)+∈(X j (t)-X k (t))

[0044]

[0045] Among them, X(t) is the current iteration position, LF() is the Levy flight function, ∈ is a random number evenly distributed from 0 to 1, and X j (t), X k (t) are two random Harris hawks, and F() is the fitness function.

[0046] Optionally, in the process of updating individuals included in the current population based on the differential evolution algorithm, the mutation behavior of the differential evolution algorithm can be used to replace the progressive rapid dive and hard encirclement to update the individuals of the current population. The specific formula is as follows:

[0047]

[0048] F=1.2×β t -1

[0049] Among them, X α (t), X β (t) and X γ (t) is randomly selected from the current population. t is an arbitrary number between 0 and 1.

[0050] Optionally, before reaching the preset first termination condition, that is, before the final output result, a chaotic local search can be performed on the global optimal solution. The search method is as follows:

[0051] X′ rabbit =(1-λ)X rabbit +λ(X rand1 -X rand2 )

[0052]

[0053] Among them, X′ rabbit is the global optimal solution, X rand1 , X rand2 are two randomly selected Harris Hawks. λ represents the value generated by the piecewise chaos map, P = 0.4.

[0054] In an optional embodiment, based on the power load demand data set, multiple algorithms are used to optimize the parameters of the initial variational modal decomposition model to obtain a target variational modal decomposition model, including: determining the first fitness function as: the envelope entropy of the components obtained by variational modal decomposition is minimized, wherein the envelope entropy is used to quantify the complexity of the components obtained by variational modal decomposition; based on the power load demand data set and the first fitness function, multiple algorithms are used to optimize the parameters of the initial variational modal decomposition model to obtain a target variational modal decomposition model.

[0055] Optionally, envelope entropy is a metric that quantifies signal complexity, which can measure the uncertainty of a signal in both frequency and time dimensions. In the VMD decomposition process, envelope entropy, as a fitness function, can help the algorithm evaluate the clarity and complexity of each component (IMF) obtained by decomposition. Lower envelope entropy values ​​indicate purer IMFs and clearer oscillation modes, which helps improve the quality of the decomposition results. By determining envelope entropy minimization as the goal, during the parameter optimization process, the algorithm will give priority to those parameter combinations that can produce IMF components with lower envelope entropy, thereby obtaining a variational mode decomposition model with better decomposition effect. In addition, the parameter optimization process with envelope entropy minimization as the goal can enhance the decomposition capability of the VMD model, so that it can more effectively separate signal components of different frequencies and modes when processing power load demand data. This helps to build subsequent prediction models, avoid modal aliasing problems, and improve the accuracy of power load demand prediction. Moreover, the VMD model optimized by the envelope entropy minimization strategy can provide clearer and purer IMF components. These components, as the input of the prediction model, can help the prediction model to more accurately capture the changing rules in the signal, thereby effectively improving the accuracy of power load demand prediction. In the above method, by introducing envelope entropy as a fitness function to guide the optimization process of VMD model parameters, it is possible to quantify the complexity of the decomposed components and improve the decomposition ability and prediction accuracy of the model.

[0056] Optionally, the optimization target is determined to be the VMD model parameters: Lagrange penalty parameter α, decomposition mode number K; the fitness function is determined to be the minimum envelope entropy of the component IMF obtained by variational mode decomposition, and the fitness function is constructed as follows:

[0057] fitness=minE e

[0058]

[0059] Wherein, a(j) is obtained by Hilbert demodulation of the original signal and represents the value of the jth element; j=1, 2, ..., N represents the total number of elements; e j represents the proportion of the jth element, E e represents the minimum envelope entropy, and fitness represents the value of the fitness function.

[0060] In an optional embodiment, before using a power load demand prediction model based on multiple target components to obtain multiple predicted load demands, the method also includes: using a target variational mode decomposition model to perform variational mode decomposition on the power load demand data set to obtain multiple historical components; based on the multiple historical components and the actual load demands corresponding to the multiple historical components, using multiple algorithms to optimize the parameters of the support vector regression model to obtain the power load demand prediction model, wherein the multiple algorithms include at least: Harris Hawk optimization algorithm, flower pollination optimization algorithm, differential evolution algorithm.

[0061] Optionally, before decomposing the power load demand data set using the target variational mode decomposition model, multiple historical components are generated. These historical components contain detailed information about the past load patterns of the target area, providing a specific and comprehensive reference for subsequent parameter optimization and the construction of the prediction model. Based on the historical components and their corresponding actual load demands, the key parameters of the support vector regression (SVR) model (such as the penalty slack variable C and the kernel function parameter γ) are optimized using multivariate optimization strategies such as the Harris Hawk optimization algorithm, the flower pollination optimization algorithm, and the differential evolution algorithm. These algorithms can explore the parameter space and find parameter combinations that can minimize the prediction error, thereby obtaining a power load demand prediction model with better performance. By optimizing the support vector regression model parameters, the power load demand prediction model can more accurately capture the changing laws of each decomposed historical component, reduce the prediction error, and improve the accuracy of the prediction results. For smart grid management, power dispatching, and market decision-making, accurate prediction results are crucial, which can help better cope with future load demands and optimize resource allocation. The optimization process can not only improve the accuracy of the power load demand forecasting model, but also enhance its robustness in dealing with noise, outliers and different load patterns. This means that the optimized power load demand forecasting model can still maintain a high forecasting performance when facing complex and changeable power load data, which is crucial for the stable operation of the power system.

[0062] It should be noted that since each historical component represents a specific frequency and pattern in the power load demand data set, the optimized power load demand forecasting model can also make personalized forecasts based on the characteristics of these components. This helps to analyze power load demand more carefully and provide more accurate support for load forecasting at different time points, different seasons or different weather conditions.

[0063] In an optional embodiment, based on multiple historical components and actual load demands corresponding to the multiple historical components, multiple algorithms are used to optimize the parameters of the support vector regression model to obtain a power load demand prediction model, including: initializing a group of individuals representing the support vector regression model parameters in a second search space as a target population, wherein the support vector regression model parameters include at least: penalty slack variables, parameters in the kernel function; based on multiple historical components, actual load demands corresponding to the multiple historical components and the target population, using the Harris Hawk optimization algorithm to find the optimal solution in the second search space; in the process of finding the optimal solution in the second search space, updating the positions of individuals included in the target population based on the flower pollination optimization algorithm, and updating the individuals included in the target population based on the differential evolution algorithm; repeating the above operations until a preset second termination condition is reached; determining the power load demand prediction model based on the individuals included in the target population when the preset second termination condition is reached.

[0064] Optionally, a group of individuals representing SVR model parameters (such as penalty slack variable C, kernel function parameter γ) are initialized in the second search space to form a target population. Through the algorithm's exploration in the search space, these parameters can be systematically adjusted and optimized to improve the prediction performance of the model. The Harris Hawk optimization algorithm can simulate the hunting behavior of the eagle, combine global search with local search strategies, and efficiently find the optimal solution in the parameter space. This helps to ensure that the optimized parameter combination can maximize the accuracy of the power load demand forecasting model. The flower pollination optimization algorithm and the differential evolution algorithm are based on the pollination process of flowers and the population evolution mechanism, respectively, to update the position and individual of the individuals in the target population, which can promote the diversity of parameters, avoid overfitting, and enhance the adaptability and robustness of the model to different power load demand data. The preset second termination condition provides a clear end criterion for the iterative process, such as reaching the maximum number of iterations or the prediction error is lower than a certain threshold. This helps to control the optimization process, avoid endless calculations, and ensure that the optimization results meet the expected performance standards. When the termination condition is reached, the parameter settings of the power load demand forecasting model are determined according to the optimal individuals contained in the target population. These optimized parameters enable the model to learn the relationship between historical components and actual load demand more effectively, improving the prediction accuracy and stability. The optimized power load demand prediction model can perform a more detailed analysis of different frequencies and patterns in the historical components and realize personalized power load demand prediction. By optimizing the parameters, the constructed power load demand prediction model can provide more accurate and reliable prediction results, thereby improving the operating efficiency of the power system.

[0065] It should be noted that the specific implementation process of using multiple algorithms to optimize the parameters of the support vector regression model to obtain the power load demand forecasting model is the same as the aforementioned process of using multiple algorithms to optimize the parameters of the initial variational mode decomposition model to obtain the target variational mode decomposition model. The only difference lies in the data used for model training, as well as the training parameters, optimization objectives, fitness function and other parameter settings. Therefore, the specific acquisition process of the power load demand forecasting model will not be repeated here.

[0066] In an optional embodiment, based on multiple historical components and the actual load demands corresponding to the multiple historical components, multiple algorithms are used to optimize the parameters of the support vector regression model to obtain the power load demand prediction model, including: determining the second fitness function as: the root mean square error between the predicted load demand output by the support vector regression model and the corresponding actual load demand is minimized; based on the second fitness function, multiple historical components, and the actual load demands corresponding to the multiple historical components, multiple algorithms are used to optimize the parameters of the support vector regression model to obtain the power load demand prediction model.

[0067] Optionally, by setting the minimization of the root mean square error (RMSE) as the second fitness function, the specific goal of SVR model parameter optimization is clarified. The root mean square error is a common indicator to measure the difference between the prediction result and the true value. Optimizing this error helps to improve the accuracy of the prediction model. Taking RMSE minimization as the criterion, a variety of optimization algorithms (such as Harris Hawk optimization algorithm, flower pollination optimization algorithm, differential evolution algorithm) can more specifically adjust the parameters of the SVR model, such as the penalty slack variable C and the parameter γ in the kernel function, to reduce the prediction error, thereby improving the prediction accuracy of the prediction model for power load demand. The second fitness function can not only guide the parameter optimization process, but also provide a standard for evaluating model performance. The RMSE value during the optimization process can reflect the changes in the model's prediction ability in real time, which is convenient for monitoring and verifying the optimization effect of the model.

[0068] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation of a method for predicting power load demand, which includes:

[0069] S1, improves the variational mode decomposition algorithm, named ETAHHO, including:

[0070] S11, modify the exploration phase of the variational mode decomposition algorithm as follows:

[0071]

[0072] Among them, X(t+1) is the next iteration position, X(t) is the current eagle position, and X m (t) is the average position of all eagles in the current population, X targeted (t) is the target eagle, X rabbit (t) is the current optimal solution, r1, r2, r3, r4, q are 5 random numbers evenly distributed from 0 to 1, UB is the upper limit of the search space, LB is the lower limit of the search space, N is the population of Harris hawks, X Ni is the set of i hawks selected from N Harris hawks.

[0073] S12, replace the progressive fast dive and soft envelopment behaviors of the variational mode decomposition algorithm with the global pollination and local pollination behaviors of the flower pollination optimization algorithm, and change the original update method to the following formula:

[0074] Y=X(t)+LF(X rabbit (txt))

[0075] Z=X(t)+∈(X j (t)-X k (t))

[0076]

[0077] Among them, X(t) is the current iteration position, LF() is the Levy flight function, ∈ is a random number evenly distributed from 0 to 1, and X j (t), X k (t) are two random Harris hawks, and F() is the fitness function.

[0078] The mutation behavior of the differential evolution algorithm is used to replace the progressive rapid dive and hard encirclement, and the original update method is changed to the following formula:

[0079]

[0080] F=1.2×β t -1

[0081] Among them, X α (t), X β (t) and X γ (t) is randomly selected from the current population. t is an arbitrary number between 0 and 1.

[0082] S13, before the final output result, a chaotic local search is performed on the global optimal solution. The search method is as follows:

[0083] X′ rabbit =(1-λ)X rabbit +λ(X rand1 -Xrand2 )

[0084]

[0085] Among them, X rand1 , X rand2 are two randomly selected Harris Hawks. λ represents the value generated by the piecewise chaos map (Piecewisemap), P = 0.4.

[0086] S2, extracts data from the dataset collected by the grid operator to obtain the power load demand dataset.

[0087] S3, divide the power load demand dataset into a training set and a test set.

[0088] S4, use VMD to decompose the training set into K components (IMF), and use ETAHHO to optimize the parameters of VMD to complete this task. The specific steps are as follows:

[0089] S41, initializing ETAHHO parameters, determining the Lagrangian penalty parameter α and the decomposition mode number K of VMD with the optimization objective;

[0090] S42, determine the fitness function as the minimum envelope entropy (Ee) of the IMF, and the fitness function is constructed as follows:

[0091] fitness=minE e

[0092]

[0093] Wherein, a(j) is obtained by Hilbert demodulation of the original signal and represents the value of the jth element; j=1, 2, ..., N represents the total number of elements; e j represents the proportion of the jth element, E e represents the minimum envelope entropy, and fitness represents the value of the fitness function.

[0094] S43, iterate to the maximum number of iterations to generate the best decomposition solution.

[0095] S5, predict each component using the SVR optimized by the ETAHHO parameters, integrate the results, and build a prediction model. The specific steps are as follows:

[0096] S51, initialize ETAHHO parameters, determine the optimization target as the parameters of SVR, which may include but are not limited to the penalty slack variable C, the parameter γ in the kernel function, and the fitness function is the root mean square error (RMSE).

[0097] S52, predict each component separately, record the current prediction result, and feed it back to ETAHHO to prepare for the next iteration.

[0098] S53, iterate to the maximum number of iterations.

[0099] S54, integrate the prediction results and use them in the test set, and calculate the performance indicators in the test set.

[0100] S6, deploy the trained prediction model in the power operation system.

[0101] In this embodiment, the variational modal decomposition algorithm is first improved, and the variational modal decomposition parameters are optimized using the improved variational modal decomposition algorithm so that the input load data is decomposed into multiple components. The support vector regression is optimized using the improved variational modal decomposition algorithm, and predictions are performed separately, and the results are integrated. The trained model is tested in the test set and the evaluation index is obtained. After reaching the standard, the model is deployed in the power operation system. The method of this embodiment has good prediction performance and good practical application value.

[0102] In this embodiment, a power load demand forecasting device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the terms "module" and "device" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0103] According to an embodiment of the present invention, there is also provided an embodiment of a device for implementing the above-mentioned power load demand prediction method. Figure 2 is a schematic diagram of the structure of a power load demand prediction device according to an embodiment of the present invention. Figure 2 As shown, the above-mentioned power load demand prediction device includes: a data acquisition module 200, a variational mode decomposition module 202, a first prediction module 204, and a second prediction module 206, wherein:

[0104] The data acquisition module 200 is used to acquire the power load demand data of the target area in the current sampling period;

[0105] The variational mode decomposition module 202 is used to perform variational mode decomposition on the power load demand data to obtain multiple target components;

[0106] A first prediction module 204 is used to obtain a plurality of predicted load demands based on the plurality of target components by using a power load demand prediction model, wherein the plurality of predicted load demands correspond to the plurality of target components one by one;

[0107] The second prediction module 206 is used to obtain the power load demand prediction result of the target area based on the multiple predicted load demands.

[0108] In an embodiment of the present invention, a data acquisition module 200 is set up to obtain the power load demand data of the target area in the current sampling period; a variational mode decomposition module 202 is used to perform variational mode decomposition on the power load demand data to obtain multiple target components; a first prediction module 204 is used to obtain multiple predicted load demands based on multiple target components using a power load demand prediction model, wherein the multiple predicted load demands correspond one to one with the multiple target components; a second prediction module 206 achieves the purpose of accurately predicting the power load demand based on the multiple target components obtained by performing variational mode decomposition on the power load demand data and using a power load demand prediction model, thereby achieving the technical effect of improving the accuracy of power load demand prediction, and further solving the technical problem of low prediction accuracy in the power load prediction method in the related art.

[0109] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0110] It should be noted that the data acquisition module 200, the variational mode decomposition module 202, the first prediction module 204, and the second prediction module 206 correspond to steps S102 to S108 in the embodiment, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments. It should be noted that the modules as part of the device can be run in a computer terminal.

[0111] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0112] The above-mentioned power load demand prediction device may also include a processor and a memory. The above-mentioned data acquisition module 200, variational mode decomposition module 202, first prediction module 204, second prediction module 206, etc. are all stored in the memory as program modules, and the processor executes the above-mentioned program modules stored in the memory to realize corresponding functions.

[0113] The processor includes a kernel, which retrieves the corresponding program module from the memory. The kernel may be one or more. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory includes at least one memory chip.

[0114] According to an embodiment of the present application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute any of the above-mentioned power load demand prediction methods.

[0115] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group, and the non-volatile storage medium includes a stored program.

[0116] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: obtain the power load demand data of the target area in the current sampling period; perform variational mode decomposition on the power load demand data to obtain multiple target components; based on the multiple target components, use a power load demand prediction model to obtain multiple predicted load demands, wherein the multiple predicted load demands correspond one-to-one to the multiple target components; based on the multiple predicted load demands, obtain the power load demand prediction result of the target area.

[0117] According to an embodiment of the present application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein when the program is run, any of the above-mentioned power load demand prediction methods is executed.

[0118] According to an embodiment of the present application, an embodiment of a computer program product is also provided, which, when executed on a data processing device, is suitable for executing a program that initializes any of the above-mentioned power load demand forecasting method steps.

[0119] Optionally, the above-mentioned computer program product, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: obtaining the power load demand data of the target area in the current sampling period; performing variational mode decomposition on the power load demand data to obtain multiple target components; based on the multiple target components, using a power load demand prediction model to obtain multiple predicted load demands, wherein the multiple predicted load demands correspond one-to-one to the multiple target components; based on the multiple predicted load demands, obtaining the power load demand prediction result of the target area.

[0120] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: obtaining power load demand data of a target area in a current sampling period; performing variational mode decomposition on the power load demand data to obtain multiple target components; based on the multiple target components, using a power load demand prediction model to obtain multiple predicted load demands, wherein the multiple predicted load demands correspond one-to-one to the multiple target components; based on the multiple predicted load demands, obtaining a power load demand prediction result for the target area.

[0121] The above sequence of the embodiments of the present invention is for description only and does not represent the superiority or inferiority of the embodiments.

[0122] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0123] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the above modules can be a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, modules or indirect coupling or communication connection of modules, which can be electrical or other forms.

[0124] The modules described above as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0125] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of software functional modules.

[0126] If the above-mentioned integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a non-volatile storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present invention. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0127] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting power load demand, characterized in that: include: Obtain the power load demand data of the target area during the current sampling period; Performing variational mode decomposition on the power load demand data to obtain multiple target components; Based on the multiple target components, a power load demand forecasting model is used to obtain multiple predicted load demands, wherein the multiple predicted load demands correspond to the multiple target components one by one; Based on the multiple predicted load demands, the power load demand prediction result of the target area is obtained.

2. The method according to claim 1, characterized in that The power load demand data is subjected to variational mode decomposition to obtain a plurality of target components, including: Using a target variational modal decomposition model to perform variational modal decomposition on the power load demand data to obtain the multiple target components; The target variational modal decomposition model is obtained by optimizing the parameters of the initial variational modal decomposition model based on multiple algorithms.

3. The method according to claim 2, characterized in that Before performing variational mode decomposition on the power load demand data using the target variational mode decomposition model to obtain the multiple target components, the method further includes: Acquire the power load demand data of the target area in the historical sampling period to obtain a power load demand data set; Based on the power load demand data set, the multiple algorithms are used to perform parameter optimization on the initial variational modal decomposition model to obtain the target variational modal decomposition model, wherein the multiple algorithms include at least: Harris Hawk optimization algorithm, flower pollination optimization algorithm, and differential evolution algorithm.

4. The method according to claim 3, characterized in that The method of optimizing parameters of the initial variational modal decomposition model based on the power load demand data set using the multiple algorithms to obtain the target variational modal decomposition model includes: Initializing a group of individuals representing variational mode decomposition model parameters in the first search space as the current population, wherein the variational mode decomposition model parameters at least include: a Lagrangian penalty parameter and a decomposition mode number; Based on the power load demand data set and the current population, using the Harris Hawk optimization algorithm to find an optimal solution in the first search space; In the process of searching for the optimal solution in the first search space, the positions of the individuals in the current population are updated based on the flower pollination optimization algorithm, and the individuals in the current population are updated based on the differential evolution algorithm; Repeat the above operation until the preset first termination condition is reached; Based on the fact that the current population includes individuals when the preset first termination condition is reached, the target variational mode decomposition model is determined.

5. The method according to claim 3, characterized in that: The method of optimizing parameters of the initial variational modal decomposition model based on the power load demand data set using the multiple algorithms to obtain the target variational modal decomposition model includes: Determine the first fitness function as follows: the envelope entropy of the components obtained by variational mode decomposition is minimized, wherein the envelope entropy is used to quantify the complexity of the components obtained by variational mode decomposition; Based on the power load demand data set and the first fitness function, the multiple algorithms are used to perform parameter optimization on the initial variational modal decomposition model to obtain the target variational modal decomposition model.

6. The method according to claim 2, characterized in that Before obtaining a plurality of predicted load demands based on the plurality of target components using a power load demand prediction model, the method further includes: Using the target variational modal decomposition model to perform variational modal decomposition on the power load demand data set to obtain multiple historical components; Based on the multiple historical components and the actual load demands corresponding to the multiple historical components, the multiple algorithms are used to optimize the parameters of the support vector regression model to obtain the power load demand prediction model, wherein the multiple algorithms include at least: Harris Hawk optimization algorithm, flower pollination optimization algorithm, and differential evolution algorithm.

7. The method according to claim 6, characterized in that The method of optimizing the parameters of the support vector regression model using the multiple algorithms based on the multiple historical components and the actual load demands respectively corresponding to the multiple historical components to obtain the power load demand prediction model includes: Initializing a group of individuals representing support vector regression model parameters in the second search space as the target population, wherein the support vector regression model parameters at least include: penalty slack variables and parameters in the kernel function; Based on the multiple historical components, the actual load demands respectively corresponding to the multiple historical components and the target population, using the Harris Eagle optimization algorithm to find an optimal solution in the second search space; In the process of searching for the optimal solution in the second search space, the positions of the individuals in the target population are updated based on the flower pollination optimization algorithm, and the individuals in the target population are updated based on the differential evolution algorithm; Repeat the above operation until the preset second termination condition is reached; The power load demand prediction model is determined based on that the target population includes individuals when the preset second termination condition is reached.

8. The method according to claim 6, characterized in that The method of optimizing the parameters of the support vector regression model using the multiple algorithms based on the multiple historical components and the actual load demands respectively corresponding to the multiple historical components to obtain the power load demand prediction model includes: Determine the second fitness function as follows: the root mean square error between the predicted load demand output by the support vector regression model and the corresponding actual load demand is minimized; Based on the second fitness function, the multiple historical components, and the actual load demands corresponding to the multiple historical components, the multiple algorithms are used to optimize the parameters of the support vector regression model to obtain the power load demand prediction model.

9. A power load demand forecasting device, characterized in that: include: A data acquisition module is used to obtain the power load demand data of the target area in the current sampling period; A variational mode decomposition module, used for performing variational mode decomposition on the power load demand data to obtain multiple target components; A first prediction module is used to obtain a plurality of predicted load demands based on the plurality of target components by using a power load demand prediction model, wherein the plurality of predicted load demands correspond to the plurality of target components one by one; The second prediction module is used to obtain the power load demand prediction result of the target area based on the multiple predicted load demands.

10. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the power load demand forecasting method described in any one of claims 1 to 8.

Images