Power load prediction method and device
By building a multi-model architecture and optimizing preset weights using optimization algorithms, the problem of large errors in existing power load prediction methods is solved, and the accuracy and reliability of prediction are improved.
Patent Information
- Application Number
- CN202411079470.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-12
AI Technical Summary
The existing power load prediction methods have large errors, especially because there are many factors affecting load changes and the process is complicated, making it difficult to improve the prediction accuracy.
Build a power load prediction model architecture, consisting of multiple prediction models linearly, each prediction model is equipped with preset weights, and the preset weights are optimized through multiple optimization algorithms (such as Golden Leopard Optimization Algorithm, Sparrow Algorithm and Bourne Sea Scavenger Group Algorithm). The optimized result data set is obtained and the final power load prediction model is determined based on the prediction accuracy and the pre-constructed weight setting rules.
Through various optimization algorithms, the preset weights are optimized, the accuracy of power load prediction is improved, the error is reduced, and more reliable load prediction results are provided.
Smart Images

Figure CN118944084A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of load forecasting, and in particular to a method and device for forecasting electric load. Background Art
[0002] Power load forecasting is an important part of power grid dispatching. Accurate and reliable load forecasting is of great significance to the safe and economical operation of the power system.
[0003] According to the different forecasting periods, it is mainly divided into short-term load forecasting and medium- and long-term load forecasting. Compared with short-term load forecasting, medium- and long-term load forecasting is more difficult to predict because it will be affected by many uncertain factors. The short-term forecasting methods of power grids can be roughly divided into two categories: traditional forecasting methods and artificial intelligence forecasting methods. Traditional forecasting methods mainly include time series method, regression analysis method, state space method, etc., which have the advantages of simple algorithm, fast speed and easy implementation. Artificial intelligence forecasting methods mainly use expert system theory, fuzzy theory and neural network theory to perform short-term load forecasting.
[0004] However, traditional prediction methods have the problem of large errors. Although the accuracy of artificial intelligence prediction methods is constantly improving with the continuous development of artificial intelligence, the errors are also large due to the many factors affecting load changes and the complex process. Summary of the invention
[0005] The embodiment of the present invention provides a method and device for predicting power load to solve the problem of large errors in existing prediction methods.
[0006] In a first aspect, an embodiment of the present invention provides a method for predicting power load, comprising:
[0007] Constructing a power load forecasting model framework, wherein the power load forecasting model framework is linearly composed of multiple forecasting models, and each forecasting model has a corresponding preset weight;
[0008] Based on multiple optimization algorithms, the preset weights of each prediction model in the power load prediction model architecture are optimized to obtain an optimization result data set; wherein the optimization result data set includes multiple groups of results, each group of results corresponds to an optimized power load prediction model, and each group of results includes the optimization weight of each prediction model and the prediction accuracy of the optimized power load prediction model;
[0009] Determine the final power load forecasting model based on the forecast accuracy corresponding to each optimized power load forecasting model in the optimization result data set and the pre-built weight setting rules;
[0010] The power load in the target area and target period is predicted based on the final power load forecasting model.
[0011] In a possible implementation, the preset weights of each prediction model in the power load prediction model architecture are optimized based on multiple optimization algorithms to obtain an optimization result data set, including:
[0012] Initializing a seed cluster of preset weights for each prediction model in the power load prediction model framework;
[0013] Calculate the fitness value of each prediction model based on the clusters, and calculate the average fitness value of the clusters;
[0014] According to the fitness value of each prediction model and the average fitness value of the population, the preset weight of each prediction model in the power load prediction model architecture is optimized based on multiple optimization algorithms to obtain an optimization result data set.
[0015] In a possible implementation, the multiple optimization algorithms include a golden leopard optimization algorithm, a sparrow algorithm, and a salp swarm algorithm.
[0016] In a possible implementation, the preset weights of each prediction model in the power load prediction model architecture are optimized based on multiple optimization algorithms to obtain an optimization result data set, including:
[0017] Based on the Golden Leopard optimization algorithm, the preset weights of each prediction model are optimized and updated; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the first result; wherein the first result includes the optimization weights of each prediction model and the prediction accuracy of the optimized power load prediction model;
[0018] The preset weights of each prediction model are optimized and updated based on the sparrow algorithm; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the second result; wherein the second result includes the optimized weights of each prediction model and the prediction accuracy of the optimized power load prediction model;
[0019] Based on the salp swarm algorithm, the preset weights of each prediction model are optimized and updated; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the third result; wherein the third result includes the optimized weights of each prediction model and the prediction accuracy of the optimized power load prediction model.
[0020] In a possible implementation, when it is determined that the iteration stop condition is not met, the method further includes:
[0021] The fitness value of each prediction model is recalculated again according to the cluster, and the average fitness value of the cluster is calculated, and a new round of optimization search is cycled.
[0022] In one possible implementation, the pre-built weight setting rule is:
[0023] Eliminate the results whose prediction accuracy is less than a first preset threshold value from the optimization result data set to obtain a result data set after elimination;
[0024] Determine the optimal weight of each prediction model based on the optimized weight of each prediction model for each group of results in the eliminated result data set;
[0025] Based on the prediction accuracy corresponding to each optimized power load forecasting model in the optimization result data set and the pre-built weight setting rules, the final power load forecasting model is determined, including:
[0026] Based on the prediction accuracy corresponding to each optimized power load prediction model in the optimization result data set and the pre-built weight setting rules, the optimal weight of each prediction model is determined;
[0027] Based on the optimal weight of each prediction model, the final power load prediction model is determined.
[0028] In a possible implementation, based on the optimized weight of each prediction model for each group of results in the eliminated result data set, determining the optimal weight of each prediction model includes:
[0029] Based on the average of the optimized weights of all target prediction models for each group of results in the eliminated result data set, the optimal weights of all target prediction models are determined; wherein the target prediction model is any prediction model for each group of results.
[0030] In one possible implementation, the multiple prediction models include a convolutional neural network model, a differential autoregressive moving average model, and a long short-term memory network model.
[0031] In a possible implementation, each of the multiple prediction models is constructed based on date type, temperature, humidity, rainfall, and wind speed.
[0032] In a second aspect, an embodiment of the present invention provides a device for predicting power load, comprising:
[0033] A construction module is used to construct a power load forecasting model architecture, wherein the power load forecasting model architecture is linearly composed of multiple forecasting models, and each forecasting model has a corresponding preset weight;
[0034] An optimization module, used to optimize the preset weights of each prediction model in the power load prediction model architecture based on an optimization algorithm to obtain an optimization result data set; wherein the optimization result data set includes multiple groups of results, each group of results corresponds to an optimized power load prediction model, and each group of results includes the optimization weights of each prediction model and the prediction accuracy of the optimized power load prediction model;
[0035] A determination module, used to determine a final power load forecasting model based on the prediction accuracy corresponding to each optimized power load forecasting model in the optimization result data set and a pre-built weight setting rule;
[0036] The prediction module is used to predict the power load in the target area and target period based on the final power load prediction model.
[0037] The embodiment of the present invention provides a method and device for predicting electric load. First, a power load prediction model framework is constructed. Then, based on a plurality of optimization algorithms, the preset weights of each prediction model in the power load prediction model framework are optimized to obtain an optimization result data set. Next, based on the prediction accuracy corresponding to each optimized power load prediction model in the optimization result data set and the pre-constructed weight setting rules, the final power load prediction model is determined. Finally, the power load of the target area and target period is predicted based on the final power load prediction model. The present invention optimizes the preset weights of each prediction model in the power load prediction model framework by using a plurality of optimization algorithms, so as to obtain multiple optimization results. Then, based on the pre-constructed weight setting rules, the optimal weights of each prediction model can be determined. Finally, based on the obtained optimal weights, a power load prediction model that can accurately predict the power load can be constructed. Furthermore, the use of the power load prediction model provided by the present invention can provide the accuracy of load prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying creative work.
[0039] Figure 1 is a flow chart of an implementation of a method for predicting power load provided by an embodiment of the present invention;
[0040] Figure 2 It is a structural schematic diagram of the power load prediction device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0041] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.
[0042] In order to make the purpose, technical solutions and advantages of the present invention more clear, specific embodiments will be described below in conjunction with the accompanying drawings.
[0043] Figure 1 The implementation flow chart of the power load prediction method provided by the embodiment of the present invention is detailed as follows:
[0044] S110. Construct a power load prediction model architecture.
[0045] As power load forecasting has become a hot topic in current research, it is currently mainly based on a single forecasting model for forecasting. However, since power load is easily affected by multiple factors, a single forecasting model may have certain errors in forecasting. In order to improve the accuracy of the forecast, it is necessary to use multiple models for forecasting at the same time. Therefore, the present invention constructs a power load forecasting model architecture.
[0046] The power load forecasting model architecture is linearly composed of multiple forecasting models, and each forecasting model has a corresponding preset weight.
[0047] In some embodiments, the multiple prediction models include Convolutional Neural Networks (CNN), Autoregressive Integrated Moving Average Model (ARIMA), and Long Short-Term Memory (LSTM).
[0048] Since the prediction model is affected by various factors, it is necessary to consider multiple factors when building the model, such as date type, meteorological conditions such as temperature, humidity, rainfall, and wind speed, etc. Based on the historical date type and meteorological conditions, as well as the historical power load of the corresponding date, CNN, ARIMA, and LSTM can be constructed.
[0049] The constructed power load forecasting model architecture is:
[0050]
[0051] in, is the preset weight of CNN, y i1 is the predicted value of CNN, β is the preset weight of ARIMA, y i2 is the predicted value of ARIMA, γ is the preset weight of LSTM, and y i3 is the predicted value of LSTM.
[0052] S120, optimizing the preset weights of each prediction model in the power load prediction model architecture based on multiple optimization algorithms to obtain an optimization result data set.
[0053] The optimization result data set includes multiple groups of results, each group of results corresponds to an optimized power load forecasting model, and each group of results includes the optimization weight of each forecasting model and the forecasting accuracy of the optimized power load forecasting model.
[0054] In some embodiments, the optimization process is specifically as follows:
[0055] First, a seed cluster of preset weights for each prediction model in the power load prediction model framework is initialized.
[0056] Then, the fitness value of each prediction model is calculated based on the clusters, and the average fitness value of the clusters is calculated.
[0057] Finally, according to the fitness value of each prediction model and the average fitness value of the population, the preset weight of each prediction model in the power load prediction model architecture is optimized based on the optimization algorithm to obtain the optimization result data set.
[0058] In some embodiments, the optimization algorithms include the Golden Leopard optimization algorithm, the Sparrow algorithm, and the Salp Swarm algorithm.
[0059] The present invention uses the golden jackal algorithm to optimize the regularization coefficient C and the kernel function parameter S. The fitness function is designed as the error rate of the training set and the test set. The decision of population position update is made by assigning different weights to the golden jackal individuals.
[0060] The sparrow algorithm divides the sparrow population into two categories: finders and joiners. The finder is responsible for finding food for the entire population and providing foraging directions for the joiner, so the foraging search range of the finder is larger than that of the joiner. In each iteration, the finder iterates according to a specific formula to find a better food location. The joiner updates its position based on the finder's position and its own fitness value to get closer to the food source.
[0061] The salp swarm algorithm is inspired by the predation behavior of a deep-sea microorganism, the salp. It is simple to implement, converges quickly, and is easy to implement on a computer. The algorithm uses a chain-like swarm model.
[0062] In this embodiment, when the Golden Leopard optimization algorithm is used for optimization, the preset weights of each prediction model are optimized and updated based on the Golden Leopard optimization algorithm; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the first result. The first result includes the optimization weights of each prediction model and the prediction accuracy of the optimized power load prediction model.
[0063] In this embodiment, when the sparrow algorithm is used for optimization, the preset weight of each prediction model is optimized and updated based on the sparrow algorithm; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the second result. The second result includes the optimization weight of each prediction model and the prediction accuracy of the optimized power load prediction model.
[0064] In this embodiment, when the sparrow algorithm is used for optimization, the preset weight of each prediction model is optimized and updated based on the salp swarm algorithm; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the third result. The third result includes the optimization weight of each prediction model and the prediction accuracy of the optimized power load prediction model.
[0065] In addition, when it is determined that the iteration stop condition is not met, it is necessary to recalculate the fitness value of each prediction model according to the species cluster, calculate the average fitness value of the species cluster, and cycle a new round of optimization search.
[0066] Based on the optimization of the above three methods, the optimization result data set includes the first result, the second result and the third result. Each result includes the optimization weight of each prediction model and the prediction accuracy of the optimized power load prediction model.
[0067] S130: Determine a final power load forecasting model based on the forecast accuracy corresponding to each optimized power load forecasting model in the optimization result data set and a pre-constructed weight setting rule.
[0068] In some embodiments, since the optimization result data set includes different values obtained using different optimization algorithms, it is necessary to determine the optimal result based on multiple groups of results.
[0069] In this embodiment, the pre-built weight setting rule is:
[0070] First, the results whose prediction accuracy is less than the first preset threshold in the optimization result data set are eliminated to obtain a result data set after elimination. The first preset threshold here can be set according to the needs and is not limited here.
[0071] Then, based on the optimized weight of each prediction model for each group of results in the eliminated result data set, the optimal weight of each prediction model is determined.
[0072] Next, based on the prediction accuracy corresponding to each optimized power load prediction model in the optimization result data set and the pre-constructed weight setting rules, the optimal weight of each prediction model is determined.
[0073] The optimal weight can be determined based on the average value of the weight of each prediction model. For example, the weights of the three optimization algorithms CNN, ARIMA and LSTM obtained by the Golden Leopard optimization algorithm are and The weights of the three optimization algorithms CNN, ARIMA and LSTM obtained by the sparrow algorithm are β1, β2, β3 respectively, and the weights of the three optimization algorithms CNN, ARIMA and LSTM obtained by the salp swarm algorithm are γ1, γ2, γ3 respectively.
[0074] If the prediction accuracy of the three optimization algorithms is greater than the first preset threshold, the weight of CNN is
[0075] The weights of ARIMA are The weight of LSTM is
[0076] Finally, the final power load forecasting model can be determined based on the optimal weight of each forecasting model.
[0077] In addition, when the prediction accuracy obtained by the selected optimization algorithms is less than the first preset threshold, the Gray Wolf Optimization Algorithm is required to be optimized. The prediction accuracy obtained by the Gray Wolf Optimization Algorithm has been determined. If the prediction accuracy obtained by the Gray Wolf Optimization Algorithm is still less than the first preset threshold, the optimal weight is determined based on the average value of the weights of each prediction model of all optimization algorithms.
[0078] S140: Predict the power load in the target area and target period based on the final power load prediction model.
[0079] Before making a prediction, you need to first obtain relevant data based on the relevant data required by each prediction model.
[0080] Then, the acquired relevant data is input into the final power load forecasting model, and the prediction result of the power load in the target area and the target period can be output.
[0081] The embodiment of the present invention provides a method and device for predicting electric load. First, a power load prediction model framework is constructed. Then, based on a plurality of optimization algorithms, the preset weights of each prediction model in the power load prediction model framework are optimized to obtain an optimization result data set. Next, based on the prediction accuracy corresponding to each optimized power load prediction model in the optimization result data set and the pre-constructed weight setting rules, the final power load prediction model is determined. Finally, the power load of the target area and target period is predicted based on the final power load prediction model. The present invention optimizes the preset weights of each prediction model in the power load prediction model framework by using a plurality of optimization algorithms, so as to obtain multiple optimization results. Then, based on the pre-constructed weight setting rules, the optimal weights of each prediction model can be determined. Finally, based on the obtained optimal weights, a power load prediction model that can accurately predict the power load can be constructed. Furthermore, the use of the power load prediction model provided by the present invention can provide the accuracy of load prediction.
[0082] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
[0083] The following is an embodiment of the device of the present invention. For details not described in detail therein, reference may be made to the corresponding method embodiment described above.
[0084] Figure 2 The following is a schematic diagram showing the structure of a power load prediction device provided by an embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown, which are described in detail as follows:
[0085] like Figure 2 As shown, the power load prediction device 200 includes:
[0086] A construction module 210 is used to construct a power load prediction model framework, wherein the power load prediction model framework is linearly composed of multiple prediction models, and each prediction model has a corresponding preset weight;
[0087] The optimization module 220 is used to optimize the preset weight of each prediction model in the power load prediction model architecture based on the optimization algorithm to obtain an optimization result data set; wherein the optimization result data set includes multiple groups of results, each group of results corresponds to an optimized power load prediction model, and each group of results includes the optimization weight of each prediction model and the prediction accuracy of the optimized power load prediction model;
[0088] A determination module 230, for determining a final power load forecasting model based on the prediction accuracy corresponding to each optimized power load forecasting model in the optimization result data set and a pre-constructed weight setting rule;
[0089] The prediction module 240 is used to predict the power load in the target area and target period based on the final power load prediction model.
[0090] In a possible implementation, the optimization module 220 is used to initialize a seed cluster of preset weights for each prediction model in the power load prediction model architecture;
[0091] Calculate the fitness value of each prediction model based on the clusters, and calculate the average fitness value of the clusters;
[0092] According to the fitness value of each prediction model and the average fitness value of the population, the preset weight of each prediction model in the power load prediction model architecture is optimized based on multiple optimization algorithms to obtain an optimization result data set.
[0093] In a possible implementation, the multiple optimization algorithms include a golden leopard optimization algorithm, a sparrow algorithm, and a salp swarm algorithm.
[0094] In a possible implementation, the optimization module 220 is used to optimize and update the preset weights of each prediction model based on the Golden Leopard optimization algorithm; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the first result; wherein the first result includes the optimization weights of each prediction model and the prediction accuracy of the optimized power load prediction model;
[0095] The preset weights of each prediction model are optimized and updated based on the sparrow algorithm; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the second result; wherein the second result includes the optimized weights of each prediction model and the prediction accuracy of the optimized power load prediction model;
[0096] Based on the salp swarm algorithm, the preset weights of each prediction model are optimized and updated; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the third result; wherein the third result includes the optimized weights of each prediction model and the prediction accuracy of the optimized power load prediction model.
[0097] In a possible implementation, the optimization module 220 is used to recalculate the fitness value of each prediction model according to the clusters, calculate the average fitness value of the clusters, and cycle a new round of optimization search.
[0098] In one possible implementation, the pre-built weight setting rule is:
[0099] Eliminate the results whose prediction accuracy is less than a first preset threshold value from the optimization result data set to obtain a result data set after elimination;
[0100] Determine the optimal weight of each prediction model based on the optimized weight of each prediction model for each group of results in the eliminated result data set;
[0101] A determination module 230, for determining an optimal weight of each prediction model based on the prediction accuracy corresponding to each optimized power load prediction model in the optimization result data set and a pre-established weight setting rule;
[0102] Based on the optimal weight of each prediction model, the final power load prediction model is determined.
[0103] In one possible implementation, the determination module 230 is used to determine the optimal weights of all target prediction models based on the average value of the optimized weights of all target prediction models for each group of results in the eliminated result data set; wherein the target prediction model is any prediction model for each group of results.
[0104] In one possible implementation, the multiple prediction models include a convolutional neural network model, a differential autoregressive moving average model, and a long short-term memory network model.
[0105] In a possible implementation, each of the multiple prediction models is constructed based on date type, temperature, humidity, rainfall, and wind speed.
[0106] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0107] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0108] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned various power load prediction method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium.
[0109] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A method for predicting power load, characterized in that: include: Constructing a power load forecasting model framework, wherein the power load forecasting model framework is linearly composed of multiple forecasting models, and each forecasting model has a corresponding preset weight; Based on multiple optimization algorithms, the preset weights of each prediction model in the power load prediction model architecture are optimized to obtain an optimization result data set; wherein the optimization result data set includes multiple groups of results, each group of results corresponds to an optimized power load prediction model, and each group of results includes the optimization weight of each prediction model and the prediction accuracy of the optimized power load prediction model; Determining a final power load forecasting model based on the prediction accuracy corresponding to each optimized power load forecasting model in the optimization result data set and a pre-constructed weight setting rule; The power load in the target area and target period is predicted based on the final power load forecasting model.
2. The method for predicting power load according to claim 1, characterized in that: The method optimizes the preset weights of each prediction model in the power load prediction model architecture based on multiple optimization algorithms to obtain an optimization result data set, including: Initializing a seed cluster of preset weights for each prediction model in the power load prediction model framework; Calculating the fitness value of each prediction model based on the clusters, and calculating the average fitness value of the clusters; According to the fitness value of each prediction model and the average fitness value of the cluster, the preset weight of each prediction model in the power load prediction model architecture is optimized based on multiple optimization algorithms to obtain an optimization result data set.
3. The method for predicting power load according to claim 2, characterized in that: The multiple optimization algorithms include the Golden Leopard optimization algorithm, the Sparrow algorithm and the Salp Swarm algorithm.
4. The method for predicting power load according to claim 3, characterized in that: The method optimizes the preset weights of each prediction model in the power load prediction model architecture based on multiple optimization algorithms to obtain an optimization result data set, including: Based on the Golden Leopard optimization algorithm, the preset weights of each prediction model are optimized and updated; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the first result; wherein the first result includes the optimization weights of each prediction model and the prediction accuracy of the optimized power load prediction model; Optimizing and updating the preset weight of each prediction model based on the sparrow algorithm; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the second result; wherein the second result includes the optimization weight of each prediction model and the prediction accuracy of the optimized power load prediction model; Based on the salp swarm algorithm, the preset weights of each prediction model are optimized and updated; when the iteration stop condition is met, the iteration stops, and the parameters of the current iteration are output as the third result; wherein the third result includes the optimized weights of each prediction model and the prediction accuracy of the optimized power load prediction model.
5. The method for predicting power load according to claim 4, characterized in that: When it is determined that the iteration stop condition is not met, it also includes: The fitness value of each prediction model is recalculated again according to the cluster, and the average fitness value of the cluster is calculated, and a new round of optimization search is cycled.
6. The method for predicting power load according to claim 1, characterized in that: The pre-built weight setting rule is: Eliminate the results whose prediction accuracy is less than a first preset threshold value from the optimization result data set to obtain an eliminated result data set; Determining the optimal weight of each prediction model based on the mean of the optimized weights of each prediction model for each group of results in the eliminated result data set; The method of determining a final power load forecasting model based on the forecast accuracy corresponding to each optimized power load forecasting model in the optimization result data set and a pre-constructed weight setting rule includes: Determining the optimal weight of each prediction model based on the prediction accuracy corresponding to each optimized power load prediction model in the optimization result data set and the pre-constructed weight setting rule; Based on the optimal weight of each prediction model, the final power load prediction model is determined.
7. The method for predicting power load according to claim 6, characterized in that: The step of determining the optimal weight of each prediction model based on the optimized weight of each prediction model for each group of results in the eliminated result data set comprises: Based on the average value of the optimized weights of all target prediction models for each group of results in the eliminated result data set, the optimal weights of all the target prediction models are determined; wherein the target prediction model is any one prediction model for each group of results.
8. The method for predicting power load according to claim 1, characterized in that: The multiple prediction models include a convolutional neural network model, a differential autoregressive moving average model and a long short-term memory network model.
9. The method for predicting power load according to claim 8, characterized in that: Each of the plurality of prediction models is constructed based on date type, temperature, humidity, rainfall, and wind speed.
10. A device for predicting power load, characterized in that: include: A construction module is used to construct a power load prediction model architecture, wherein the power load prediction model architecture is linearly composed of multiple prediction models, and each prediction model has a corresponding preset weight; An optimization module, used to optimize the preset weights of each prediction model in the power load prediction model architecture based on multiple optimization algorithms to obtain an optimization result data set; wherein the optimization result data set includes multiple groups of results, each group of results corresponds to an optimized power load prediction model, and each group of results includes the optimization weights of each prediction model and the prediction accuracy of the optimized power load prediction model; A determination module, used to determine a final power load forecasting model based on the prediction accuracy corresponding to each optimized power load forecasting model in the optimization result data set and a pre-constructed weight setting rule; The prediction module is used to predict the power load in the target area and target period based on the final power load prediction model.
Citation Information
Cited By
Oil film segmentation method based on GCN-IGJO
CN122200029A
