Power system operation optimization method and device based on JSO-SVM model and dynamic adjustment

Abstract

The invention relates to an electric power system operation optimization method and device based on a JSO-SVM model and dynamic adjustment, and belongs to the technical field of electric power systems, and the method comprises the following steps: collecting data related to the operation of an electric power system, carrying out the data processing, finding out the historical number of days similar to the target daily load characteristics through calculating the daily feature similarity, and carrying out the calculation of the target daily load characteristics. The related data comprises historical load data, meteorological conditions and user behavior modes of the power system; establishing a JSO-SVM load prediction model based on SVM and JSO algorithms, and performing parameter optimization through model training; setting a target function, and performing real-time load prediction by using a JSO-SVM load prediction model; and according to a load prediction result obtained in real time, applying a dynamic adjustment algorithm to optimize an operation strategy of the power system. According to the method, accurate prediction and real-time dynamic optimal scheduling of the power load are realized, and the stability and the economical efficiency of a power system are improved.

Description

Technical Field

[0001] The invention relates to a method and a device for optimizing power system operation based on a JSO-SVM model and dynamic adjustment, and belongs to the technical field of power systems. Background Art

[0002] With the continuous growth of global energy demand and the intensification of environmental problems, the power system is facing unprecedented challenges. As an important means to alleviate the contradiction between "source and load", the regulation of flexible load resources plays an increasingly critical role in the operation of the power system. Flexible load resources have the characteristics of flexibility, demand elasticity and two-way regulation capabilities, which make them the core tool for economic dispatch and optimal resource allocation of power grids.

[0003] Flexible power loads refer to those loads that can adjust power consumption or shift in different time periods within a certain range. The flexibility of such loads allows them to be dynamically adjusted according to the actual needs of the power system. Through accurate prediction of flexible power loads, power dispatch and resource allocation can be carried out more effectively, thereby improving the operating efficiency and economy of the power grid. For example, large machines in the manufacturing industry can reduce operation during peak power load periods, air conditioning systems in shopping malls can be adjusted according to power demand, and home appliances in households can adjust their operating time during peak power demand periods. Electric vehicle charging stations can choose to charge when power demand is low, while energy storage systems can release power when power load is peak or store power when it is low. The regulation of these flexible loads helps balance the power grid load, improve system stability, and reduce operating costs.

[0004] As a powerful data analysis tool, machine learning can extract potential patterns and rules from a large amount of historical data, providing a new method for predicting power flexible load resources. In particular, in power load forecasting, machine learning algorithms can establish accurate prediction models by comprehensively analyzing historical load data, meteorological conditions, user behavior patterns and other factors. Common machine learning algorithms such as artificial neural networks (ANN) and support vector machines (SVM) are widely used in the field of power load forecasting. ANN can learn complex nonlinear relationships by simulating the connection and information processing of neurons, while SVM is suitable for dealing with small samples, high dimensions, nonlinear problems, and has good generalization ability and robustness. In addition, the research on composite prediction models has also become a hot topic in the field of power flexible load resource forecasting. By combining and optimizing multiple machine learning algorithms, the advantages of each algorithm can be fully utilized to improve the prediction accuracy and stability. For example, meta-heuristic optimization algorithms, such as genetic algorithms, ant colony algorithms, simulated annealing algorithms, etc., are applied to model parameter setting to obtain the best performance prediction model.

[0005] In addition, as flexible load resources are increasingly used in power systems, how to achieve dynamic adjustment and optimal dispatch of loads based on load forecasting has become the key to improving the operating efficiency of power systems. Traditional operation optimization algorithms usually make decisions based on static load forecasting results, and fail to fully consider real-time load changes and the dynamics of the system operating environment, resulting in insufficient effectiveness and flexibility of dispatch strategies. Summary of the invention

[0006] The present invention proposes a method and device for optimizing power system operation based on the JSO-SVM model and dynamic adjustment, which can achieve accurate prediction of power load and real-time dynamic optimization scheduling, and improve the stability and economy of the power system.

[0007] The technical solution adopted by the present invention to solve the technical problem is:

[0008] In a first aspect, an embodiment of the present invention provides a method for optimizing power system operation based on a JSO-SVM model and dynamic adjustment, comprising the following steps:

[0009] Step S1, collecting data related to the operation of the power system, performing data processing, and finding historical days with similar load characteristics to the target day by calculating the daily feature similarity, and constructing a data set, wherein the relevant data includes historical load data of the power system, meteorological conditions, and user behavior patterns;

[0010] Step S2, establishing a JSO-SVM load forecasting model based on SVM and JSO algorithm (Jellyfish Search Optimization), and performing parameter optimization through model training;

[0011] Step S3, setting the objective function and using the JSO-SVM load forecasting model to perform real-time load forecasting;

[0012] Step S4, based on the load forecasting results obtained in real time, a dynamic adjustment algorithm is applied to optimize the operation strategy of the power system.

[0013] As a possible implementation of this embodiment, the data processing includes:

[0014] The normalization method is used to convert data of different dimensions into the same numerical range.

[0015] As a possible implementation of this embodiment, the formula for calculating the daily feature similarity is:

[0016] T(x i ,x j )=αD(x i ,x j)+βR(x i ,x j ) (17)

[0017]

[0018] In the above formula, the sample is x i =(x i1 ,x i2 ,…,x im ), and (0 <x ij <1); T(x i ,x j ) represents the similarity x between samples i ,x j ; D(x i ,x j ) represents the spatial distance similarity; R(x i ,x j ) represents the similarity of the change trend; α and β are weight coefficients; σ k represents the normalized value of the kth feature; m is the total number of features.

[0019] As a possible implementation of this embodiment, step S2, establishing a JSO-SVM load forecasting model based on the SVM and JSO algorithms, and performing parameter optimization through model training, includes:

[0020] Define an ε-insensitive band that allows prediction errors to be within the ε range without penalty:

[0021]

[0022] Among them, y i is the predicted value, is the reference value, and ε is the insensitive band;

[0023] Select the kernel function based on the characteristics of the data:

[0024] K(x i ,x j )=φ(x i )·φ(x j ) (twenty one)

[0025] Among them, x i represents the input feature vector of the i-th sample, x j Represents the input feature vector of the jth sample;

[0026] The optimal model is solved by minimizing the following objective function:

[0027]

[0028] Among them, w is the model weight vector, C is the penalty coefficient, L ò is the ε-insensitive loss function;

[0029] The SVM model is trained using the similar day feature vectors, and the jellyfish algorithm is used to find the optimal weight vector w.

[0030] As a possible implementation of this embodiment, the method of finding the optimal weight vector by using the jellyfish algorithm includes:

[0031] ① Generation of initial population: According to the algorithm principle of JSO, a group of jellyfish individuals are randomly generated. Each individual consists of two dimensions, which represent the values ​​of C and g parameters respectively. The initial population is expressed by the following formula:

[0032] X i =(C i ,g i ),i=1,2,...,N (23)

[0033] Where N is the size of the population, C i and g i It is initialized randomly in the set search space;

[0034] ②Fitness calculation: The fitness value of each individual is evaluated based on the classification performance of the SVM model on the validation set. Assuming that the cross-validation technique is used, the fitness function is expressed as the classification accuracy of the model:

[0035] Fitness(X i )=Accuracy(C i ,g i ) (twenty four);

[0036] ③ Population update: In each iteration, JSO will update the population through the following strategies:

[0037] 1) Ocean current movement: If the control parameter C(t) ≥ 0.5, the jellyfish individual updates its position based on the fitness value of the entire population:

[0038]

[0039] Among them, t represents the current number of iterations, Max_iter represents the maximum number of iterations,

[0040] The formula for updating the position is as follows:

[0041]

[0042] Among them, X best is the optimal solution in the current population, and C(t) is the control parameter to ensure that the jellyfish moves smoothly in the ocean current;

[0043] 2) Jellyfish swarm movement: When C(t)<0.5, the swarm enters the local search mode, and the jellyfish explore the search space through passive or active movement. The update formula is as follows:

[0044] Passive exercise:

[0045]

[0046] Active Movement:

[0047]

[0048] Among them, X j is another jellyfish individual randomly selected to simulate the interaction between jellyfish;

[0049] ④The iterative process of JSO stops when the maximum number of iterations is reached or the population converges, and the C and g values ​​finally returned are the optimized support vector machine model parameters.

[0050] As a possible implementation of this embodiment, the objective function is:

[0051] minC=C PCC +C FL +C MR +C DERG (29)

[0052] Where C is the minimum total operating cost in the entire dispatching cycle of the distribution network; C PCC The cost of purchasing electricity from the upper grid; C FL is the compensation cost of flexible load dispatch; MR Distributed power grid dispatching cost; C DERG Penalize the cost of curtailing renewable energy.

[0053] As a possible implementation of this embodiment, the application of a dynamic adjustment algorithm to optimize the operation strategy of the power system includes:

[0054] Transform constraints into explicit equivalence classes for judging the feasibility of each particle;

[0055] Use linear decreasing or adaptive adjustment strategy to dynamically adjust the inertia weight ω:

[0056]

[0057] Where iter is the number of iterations;

[0058] Dynamically adjust the acceleration constants c1 and c2 according to the convergence of the particle group:

[0059]

[0060] Dynamically adjust the maximum value v of particle speed max :

[0061]

[0062] In a second aspect, an embodiment of the present invention provides a power system operation optimization device based on a JSO-SVM model and dynamic adjustment, comprising:

[0063] A data collection module is used to collect data related to the operation of the power system, perform data processing, find historical days with similar load characteristics to the target day by calculating the similarity of daily characteristics, and construct a data set. The relevant data includes historical load data of the power system, meteorological conditions, and user behavior patterns;

[0064] Model building module, used to build JSO-SVM load forecasting model based on SVM and JSO algorithms, and optimize parameters through model training;

[0065] The load forecasting module is used to set the objective function and perform real-time load forecasting using the JSO-SVM load forecasting model;

[0066] The strategy optimization module is used to optimize the operation strategy of the power system by applying a dynamic adjustment algorithm based on the load forecast results obtained in real time.

[0067] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory through the bus, and the processor executes the machine-readable instructions to perform any step of the above-mentioned power system operation optimization method based on the JSO-SVM model and dynamic adjustment.

[0068] In a fourth aspect, an embodiment of the present invention provides a storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned power system operation optimization methods based on the JSO-SVM model and dynamic adjustment are executed.

[0069] The beneficial effects of the technical solution of the embodiment of the present invention are as follows:

[0070] The present invention adopts the SVM prediction model, uses the JSO algorithm to obtain the optimal parameters, constructs the JSO-SVM load prediction model, and combines the dynamic adjustment algorithm to achieve the optimized operation of the power system. It combines the high-precision prediction ability of the JSO-SVM model and the flexibility of the dynamic adjustment algorithm, and can achieve accurate prediction of power load and real-time dynamic optimization scheduling. The present invention effectively reduces the peak-to-average ratio and operating cost of the power system, improves the overall performance and economy of the system, provides support for the stable operation of the power system, and promotes the development of smart grids.

[0071] The present invention improves the stability and economy of the power system and has the following characteristics compared with the prior art:

[0072] (1) Improve load forecasting accuracy: By building a power load forecasting model based on machine learning and integrating historical data, meteorological variables, and user behavior patterns, high-precision forecasting of power flexible loads can be achieved. The JSO-SVM algorithm is compared with the artificial neural network algorithm (ANN) to optimize the model parameters and thus improve the load forecasting accuracy.

[0073] (2) Optimizing power system operation: Based on load forecasting, a dynamic adjustment optimization algorithm is proposed to adjust the power system operation strategy according to the real-time load forecast results. This algorithm aims to adapt to changes in the power system through real-time data and improve the flexibility and effectiveness of the dispatch strategy.

[0074] (3) Reduce the system peak-to-average ratio and operating costs: By dynamically adjusting the optimization algorithm, the peak-to-average ratio and operating costs of the power system can be significantly reduced, thereby improving the overall economy and operating efficiency of the system. This provides effective support for the operation and management of the power system and promotes the development of smart grid infrastructure. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 is a flow chart of a method for optimizing power system operation based on a JSO-SVM model and dynamic adjustment according to an exemplary embodiment;

[0076] Figure 2 is a structural schematic diagram of a power system operation optimization device based on a JSO-SVM model and dynamic adjustment according to an exemplary embodiment;

[0077] Figure 3 A specific implementation flow chart of the present invention for optimizing power system operation;

[0078] Figure 4 It is a schematic diagram of SVM;

[0079] Figure 5 It is a schematic diagram of the root mean square error of JSO-SVM and ANN during training;

[0080] Figure 6 It is a schematic diagram of the prediction accuracy of JSO-SVM and ANN;

[0081] Figure 7 It is a schematic diagram comparing the peak-to-average ratio (PAPR) before and after optimization;

[0082] Figure 8 It is a schematic diagram comparing the convergence curves before and after optimization. DETAILED DESCRIPTION

[0083] In order to more clearly illustrate the technical features of the solution of the present invention, the present invention is described in detail below through specific implementation methods and in conjunction with the accompanying drawings.

[0084] like Figure 1 As shown, an embodiment of the present invention provides a method for optimizing power system operation based on a JSO-SVM model and dynamic adjustment, comprising the following steps:

[0085] Step S1, collecting data related to the operation of the power system, performing data processing, and finding historical days with similar load characteristics to the target day by calculating the daily feature similarity, and constructing a data set, wherein the relevant data includes historical load data of the power system, meteorological conditions, and user behavior patterns;

[0086] Step S2, establishing a JSO-SVM load forecasting model based on SVM and JSO algorithms, and performing parameter optimization through model training;

[0087] Step S3, setting the objective function and using the JSO-SVM load forecasting model to perform real-time load forecasting;

[0088] Step S4, based on the load forecasting results obtained in real time, a dynamic adjustment algorithm is applied to optimize the operation strategy of the power system.

[0089] As a possible implementation of this embodiment, the data processing includes:

[0090] The normalization method is used to convert data of different dimensions into the same numerical range.

[0091] As a possible implementation of this embodiment, the formula for calculating the daily feature similarity is:

[0092] T(x i ,x j )=αD(x i ,x j )+βR(x i ,x j ) (33)

[0093]

[0094] In the above formula, the sample is x i =(x i1 ,x i2 ,…,x im ), and (0 <x ij <1); T(x i ,x j ) represents the similarity x between samples i ,x j ; D(x i ,x j ) represents the spatial distance similarity; R(x i ,x j ) represents the similarity of the change trend; α and β are weight coefficients; σ k represents the normalized value of the kth feature; m is the total number of features.

[0095] As a possible implementation of this embodiment, step S2, establishing a JSO-SVM load forecasting model based on the SVM and JSO algorithms, and performing parameter optimization through model training, includes:

[0096] Define an ε-insensitive band that allows prediction errors to be within the ε range without penalty:

[0097]

[0098] Among them, y i is the predicted value, is the reference value, and ε is the insensitive band;

[0099] Select the kernel function based on the characteristics of the data:

[0100] K(x i ,x j )=φ(x i )·φ(x j ) (37)

[0101] Among them, x i represents the input feature vector of the i-th sample, x j Represents the input feature vector of the jth sample;

[0102] The optimal model is solved by minimizing the following objective function:

[0103]

[0104] Among them, w is the model weight vector, C is the penalty coefficient, L ò is the ε-insensitive loss function;

[0105] The SVM model is trained using the similar day feature vectors, and the jellyfish algorithm is used to find the optimal weight vector w.

[0106] As a possible implementation of this embodiment, the method of finding the optimal weight vector by using the jellyfish algorithm includes:

[0107] ① Generation of initial population: According to the algorithm principle of JSO, a group of jellyfish individuals are randomly generated. Each individual consists of two dimensions, which represent the values ​​of C and g parameters respectively. The initial population is expressed by the following formula:

[0108] X i =(C i ,g i ),i=1,2,...,N (39)

[0109] Where N is the size of the population, C i and g i It is initialized randomly in the set search space;

[0110] ②Fitness calculation: The fitness value of each individual is evaluated based on the classification performance of the SVM model on the validation set. Assuming that the cross-validation technique is used, the fitness function is expressed as the classification accuracy of the model:

[0111] Fitness(X i )=Accuracy(C i ,g i ) (40);

[0112] ③ Population update: In each iteration, JSO will update the population through the following strategies:

[0113] 1) Ocean current movement: If the control parameter C(t) ≥ 0.5, the jellyfish individual updates its position based on the fitness value of the entire population:

[0114]

[0115] Among them, t represents the current number of iterations, Max_iter represents the maximum number of iterations,

[0116] The formula for updating the position is as follows:

[0117]

[0118] Among them, X best is the optimal solution in the current population, and C(t) is the control parameter to ensure that the jellyfish moves smoothly in the ocean current;

[0119] 2) Jellyfish swarm movement: When C(t)<0.5, the swarm enters the local search mode, and the jellyfish explore the search space through passive or active movement. The update formula is as follows:

[0120] Passive exercise:

[0121]

[0122] Active Movement:

[0123]

[0124] Among them, X j is another jellyfish individual randomly selected to simulate the interaction between jellyfish;

[0125] ④The iterative process of JSO stops when the maximum number of iterations is reached or the population converges, and the C and g values ​​finally returned are the optimized support vector machine model parameters.

[0126] As a possible implementation of this embodiment, the objective function is:

[0127] minC=C PCC +C FL +C MR +C DERG (45)

[0128] Where C is the minimum total operating cost in the entire dispatching cycle of the distribution network; C PCC The cost of purchasing electricity from the upper grid; C FL is the compensation cost of flexible load dispatch; MR Distributed power grid dispatching cost; C DERG Penalize the cost of curtailing renewable energy.

[0129] As a possible implementation of this embodiment, the application of a dynamic adjustment algorithm to optimize the operation strategy of the power system includes:

[0130] Transform constraints into explicit equivalence classes for judging the feasibility of each particle;

[0131] Use linear decreasing or adaptive adjustment strategy to dynamically adjust the inertia weight ω:

[0132]

[0133] Where iter is the number of iterations;

[0134] Dynamically adjust the acceleration constants c1 and c2 according to the convergence of the particle group:

[0135]

[0136] Dynamically adjust the maximum value v of particle speed max :

[0137]

[0138] like Figure 2 As shown, an embodiment of the present invention provides a power system operation optimization device based on a JSO-SVM model and dynamic adjustment, comprising:

[0139] A data collection module is used to collect data related to the operation of the power system, perform data processing, find historical days with similar load characteristics to the target day by calculating the similarity of daily characteristics, and construct a data set. The relevant data includes historical load data of the power system, meteorological conditions, and user behavior patterns;

[0140] Model building module, used to build JSO-SVM load forecasting model based on SVM and JSO algorithms, and optimize parameters through model training;

[0141] The load forecasting module is used to set the objective function and perform real-time load forecasting using the JSO-SVM load forecasting model;

[0142] The strategy optimization module is used to optimize the operation strategy of the power system by applying a dynamic adjustment algorithm based on the load forecast results obtained in real time.

[0143] like Figure 3 As shown, when the present invention optimizes the operation of the power system, first, by collecting and organizing the historical load data, meteorological conditions, user behavior patterns and other information of the power system, a high-precision power load forecasting model is constructed using the SVM algorithm. The model can effectively capture the complex nonlinear relationship of load changes, and adjust the model parameters through the jellyfish swarm algorithm during the training process to improve the prediction accuracy. Then, according to the load forecast results obtained in real time, the dynamic adjustment algorithm is applied to optimize the operation strategy of the power system. The algorithm can adjust the distribution of power supply and load in real time, consider the flexibility and adjustment cost of the load, optimize resource allocation, and balance the supply and demand relationship of the power system. By setting reasonable scheduling rules, the algorithm can reduce the peak-to-average ratio and operating cost of the system and improve the overall operating efficiency of the power grid. Finally, the optimization results are applied to the actual power system for real-time monitoring and adjustment to ensure the stable operation of the power system and maximize the economic benefits. The example results show that this method significantly improves the accuracy of load forecasting and effectively optimizes the operating performance of the power system, providing innovative technical support for the development of smart grids.

[0144] The specific implementation process of the present invention for optimizing the operation of the power system includes the following steps.

[0145] Step 1: Data processing and similar day calculation.

[0146] (1) Data processing

[0147] Climate conditions have a significant impact on people's electricity consumption behavior, so factors such as climate conditions need to be considered in power load forecasting. Since there are large numerical differences between load data and various influencing factor data (such as temperature, humidity, wind speed, etc.), direct modeling or analysis may cause the role of certain influencing factors to be ignored or exaggerated. To address this problem, the present invention adopts a normalization method to convert data of different dimensions into the same numerical range for easy comparison and analysis. The normalization formula is as follows:

[0148]

[0149] In the formula, y i is the result after data normalization; x i is the value before normalization; x min is the minimum value in the attribute sample data; x max is the maximum value in the attribute sample data.

[0150] (2) Similar day calculation (i.e., calculating the similarity of daily features).

[0151] In power load forecasting, similar day calculation is an important step, which helps to find historical days with similar load characteristics to the target day, thereby improving the accuracy of the forecast. The calculation of similarity is usually based on the daily feature vector, which may contain multiple factors such as weather, date type, temperature, humidity, etc. Usually there are two types of similarity. One is the similarity measurement based on spatial distance. If two vectors are close in space, they are considered similar. The other is the similarity measurement based on the change trend. This method pays more attention to the relative changes between vector elements rather than the absolute values. In fact, the cosine similarity between the two feature vectors is calculated to measure the angular relationship between the two vectors, reflecting their similarity in trend changes. The present invention combines these two methods and proposes a new method for calculating daily feature similarity. The formula is:

[0152] T(x i ,x j )=αD(x i ,x j )+βR(x i ,x j ) (50)

[0153]

[0154] In the above formula, the sample is x i =(x i1 ,x i2 ,…,x im), and (0 <x ij <1); T(x i ,x j ) represents the similarity x between samples i ,x j ; D(x i ,x j ) represents the spatial distance similarity; R(x i ,x j ) represents the similarity of the change trend; α and β are weight coefficients; σ k represents the normalized value of the kth feature; m is the total number of features.

[0155] Step 2: Establishment of JSO-SVM load forecasting model and parameter optimization.

[0156] (1) Establishment of load forecasting model.

[0157] In power load forecasting, the support vector machine (SVM) model is widely used due to its powerful nonlinear processing capabilities and good generalization capabilities. The results of similar day calculations can be used as input features of the SVM model to provide the model with data that is more relevant to the target day. In load forecasting tasks, SVM models are usually used for regression tasks, namely support vector regression. The core idea of ​​SVM is to achieve data classification or regression prediction by finding an optimal hyperplane to maximize the interval between different categories, such as Figure 4 As shown, the specific principles are as follows.

[0158] a) Define an ε-insensitive band: SVR defines an ε-insensitive band, allowing prediction errors to be unpunished within the ε range. This can reduce the sensitivity of the model to noisy data.

[0159]

[0160] Among them, y i is the predicted value, is the reference value, and ε is the insensitive band.

[0161] b) Kernel function selection: Select a suitable kernel function, such as linear kernel, polynomial kernel or RBF kernel, according to the characteristics of the data to handle nonlinear relationships.

[0162] K(x i ,x j )=φ(x i )·φ(x j ) (54)

[0163] Among them, x i represents the input feature vector of the i-th sample, x j Represents the input feature vector of the jth sample.

[0164] c) Minimize the objective function: SVR solves the optimal model by minimizing the following objective function:

[0165]

[0166] Among them, w is the model weight vector, C is the penalty coefficient, L ò is an ε-insensitive loss function.

[0167] During the training process, the SVM model is trained using the feature vectors of similar days. The goal is to find the optimal weight vector w so that the model performs best on the training set. The generalization ability of the model is evaluated using methods such as cross-validation, and the model parameters (such as kernel function type g, C value, ε value, etc.) are adjusted to ensure that the model has good prediction performance on the test set.

[0168] (2) Optimizing the parameters of the jellyfish algorithm.

[0169] Jellyfish Search Optimization (JSO) is a meta-heuristic algorithm that searches for the global optimal solution by simulating the movement of jellyfish in the ocean. The algorithm flow is as follows: Figure 3 As shown in the parameter selection section. It combines the two strategies of jellyfish current movement and group behavior to update the solution in the solution space in different ways. In the parameter tuning of support vector machine (SVM), JSO can be used to automatically search for suitable penalty parameters C and kernel parameters g to improve the classification performance of the model. The present invention will elaborate on the principle of applying JSO algorithm to SVM parameter tuning.

[0170] JSO explores the search space of C and g in a population manner. Each jellyfish individual represents a candidate solution for C and g. During the optimization process, the fitness value of the jellyfish is determined by the classification accuracy of the SVM model on the validation set.

[0171] ① Generation of initial population.

[0172] According to the algorithm principle of JSO, a group of jellyfish individuals are first randomly generated, each of which consists of two dimensions, representing the values ​​of C and g parameters respectively. The initial population can be expressed by the following formula:

[0173] X i =(C i ,g i ),i=1,2,...,N (56)

[0174] Where N is the size of the population, C i and g i It is initialized randomly in the set search space.

[0175] ②Fitness calculation.

[0176] The fitness value of each individual is evaluated based on the classification performance of the SVM model on the validation set. Assuming that the cross-validation technique is used, the fitness function can be expressed as the classification accuracy of the model:

[0177] Fitness(X i )=Accuracy(C i ,g i ) (57)

[0178] ③ Population update.

[0179] In each iteration, JSO will update the population by the following strategy:

[0180] 1) Ocean current movement: If the control parameter C(t) ≥ 0.5, the jellyfish individual updates its position based on the fitness value of the entire population. That is, the population will update along the direction of the current optimal solution to achieve global search.

[0181]

[0182] Where t represents the current number of iterations, and Max_iter represents the maximum number of iterations.

[0183] The update formula is as follows:

[0184]

[0185] Among them, X best is the optimal solution in the current population, and C(t) is the control parameter to ensure that the jellyfish moves smoothly in the ocean current.

[0186] 2) Jellyfish swarm movement: When C(t)<0.5, the swarm enters the local search mode, and the jellyfish explore the search space through passive or active movement. The update formula is as follows:

[0187] Passive exercise:

[0188]

[0189] Active Movement:

[0190]

[0191] Among them, X j is another jellyfish individual randomly selected to simulate the interaction between jellyfish.

[0192] ④Stop condition.

[0193] The iterative process of JSO will stop when the maximum number of iterations is reached or the population converges. Finally, the C and g values ​​returned by the algorithm are the optimized support vector machine model parameters.

[0194] The jellyfish optimization algorithm provides an effective mechanism combining global search and local search by simulating the movement and group interaction of jellyfish in ocean currents. When applied to SVM parameter tuning, JSO can automatically search for the optimal penalty parameter C and kernel parameter g, improving the classification performance of the model. Compared with traditional grid search or random search, JSO has higher search efficiency and accuracy in dealing with high-dimensional search space and complex hyperparameter tuning problems.

[0195] Step 3: Objective function setting and dynamic adjustment of optimization strategy

[0196] (1) Objective function setting

[0197] Load forecasting predicts future electricity demand by analyzing historical data and influencing factors, providing basic data for operation optimization. Operation optimization relies on these forecast results to formulate power generation plans, demand response strategies, and energy storage scheduling to ensure the stability and economy of the power system. Both work together to improve the overall efficiency of the power system. Optimization with the goal of grid economic optimization is a complex and multi-factor problem, involving multiple economic cost links in grid operation, mainly including the cost of purchasing electricity from the upper grid, the operation and maintenance costs of each dispatching unit, the investment conversion cost, the dispatching cost of flexible loads, and the penalty cost of abandoning wind and solar power.

[0198] minC=C PCC +C FL +C MR +C DERG (62)

[0199] Where C is the minimum total operating cost in the entire dispatching cycle of the distribution network; C PCC The cost of purchasing electricity from the upper grid; C FL is the compensation cost of flexible load dispatch; MR Distributed power grid dispatching cost; C DERG Penalize the cost of curtailing renewable energy.

[0200] (2) Dynamically adjust optimization strategies

[0201] Particle swarm optimization is an optimization algorithm based on swarm intelligence. Each particle represents a possible solution, and the particle approaches the optimal solution by constantly updating its speed and position. Considering the uncertainty and chance constraints in resource scheduling, the traditional particle swarm optimization (PSO) algorithm may not be able to deal with it effectively. To address this problem, the dynamic adjustment algorithm can be integrated into the particle swarm optimization (PSO) algorithm to establish a dynamic adjustment optimization strategy to better deal with the uncertainty and chance constraints in resource scheduling.

[0202] First, the constraints are converted into clear equivalence classes. These equivalence classes will be used to determine the feasibility of each particle. If the position of a particle does not meet these clear constraints, the particle is considered infeasible and needs to be regenerated. In the traditional PSO algorithm, the particle position and velocity update formula is: The particle position and velocity update formula is as follows:

[0203]

[0204] Where: represents the velocity of the i-th particle at the iteration time t+1; ω represents the influence of the primary velocity; represents the speed of the ith particle at the iteration time t; c1 and c2 are acceleration constants, which control the weight of the particle moving to its own best position and the global best position; rand() represents a random number with a value between [0,1], which increases the randomness of the algorithm; pbest represents the best position found by the ith particle so far. gbest represents the global best position found by all particles in the current iteration. Represents the position of the i-th particle at iteration time t.

[0205] Then, in order to enhance the adaptability of the algorithm, a dynamic adjustment algorithm can be incorporated into it, as follows: d) Dynamically adjust the inertia weight ω: Use a linear decreasing or adaptive adjustment strategy. In the early stage, a larger inertia weight is set to enhance the exploration ability, and in the later stage, the inertia weight is gradually reduced to enhance the convergence. The update formula is:

[0206]

[0207] Where iter is the number of iterations.

[0208] a) Dynamically adjust the acceleration constants c1 and c2: Dynamically adjust c1 and c2 according to the convergence of the particle group. For example, set c1 to be larger in the early stage of the algorithm to enhance the individual exploration ability, and gradually increase c2 as the number of iterations increases to enhance the convergence ability of the group.

[0209]

[0210] b) Dynamic limit of particle speed: In order to prevent excessive search divergence caused by excessive particle speed, the maximum speed v can be dynamically adjusted max , so that the speed limit gradually decreases with the number of iterations.

[0211]

[0212] Step 4: Case analysis.

[0213] (1) Comparative analysis of JSO-SVM load forecasting results.

[0214] In order to verify the effectiveness of the method of the present invention, the load data of a certain place was used to compare the root mean square error (MSE) of support vector machine and artificial neural network (ANN) in predicting power flexible load. The results are as follows: Figure 5 As shown in the figure. By observing the MSE curve in the figure, it can be clearly seen that the MSE curve of the JSO-SVM algorithm is lower than the MSE curve of the ANN algorithm as a whole. This shows that under the same prediction task and data set, the prediction result of the JSO-SVM algorithm is closer to the actual value and has a higher prediction accuracy.

[0215] On the other hand, compared with the traditional neural network algorithm, the prediction accuracy of the present invention is improved by more than 23%. Figure 6 As shown. The improvement in prediction accuracy means that the algorithm proposed in the present invention can more accurately predict the changing trend of flexible power load and provide more reliable data support for power grid dispatching and energy management. This is because the present invention uses JSO-SVM to better capture the nonlinear relationship in the data and improve the training speed and generalization ability of the model. In addition, JSO-SVM has strong robustness to noise and outliers, which enables it to effectively deal with uncertainties and abnormal situations in the data when processing actual power load forecasts. On the whole, the JSO-SVM algorithm, with its excellent generalization ability and adaptability to complex data, shows relatively excellent performance in predicting flexible power loads. This also further verifies the effectiveness and practicality of the method proposed in the present invention in the task of power load forecasting.

[0216] (2) Dynamically adjust algorithm optimization results.

[0217] Figure 7This is a comparison chart of the peak-to-average power ratio (PAPR) of the system before and after optimization using the method of the present invention. By comparing the data before and after optimization, it can be clearly seen that the optimization algorithm has a significant impact on the PAPR of the system. The PAPR of the system before optimization is relatively high, that is, there is a large difference between the peak power and the average power of the system. A higher PAPR may cause large fluctuations in the load of the power grid, which is not conducive to the stable operation of the power grid, and also increases the operating pressure and maintenance cost of the power grid equipment. After optimization using the method of the present invention, the PAPR of the system has been significantly reduced, indicating that the optimization algorithm has effectively balanced the power distribution of the system and narrowed the gap between the peak power and the average power. The optimized system load curve is smoother, which helps to reduce the burden on the power grid and improve the stability and reliability of the power grid. Through accurate prediction and reasonable scheduling, the algorithm can more effectively manage the load of the power grid and reduce the PAPR, thereby achieving the economically optimal operation of the power grid.

[0218] Figure 8 The convergence curves of the optimization iterations were compared to show the difference in convergence speed between the method of the present invention and the traditional neural network algorithm during the optimization iteration process. By comparing the convergence curves of the two, the advantage of the method of the present invention in terms of convergence speed can be clearly seen. In the initial stage of the optimization iteration, the traditional neural network algorithm needs to go through multiple iterations to reach a relatively stable optimization state. The method of the present invention shows a faster convergence speed and can reach the optimal solution within a smaller number of iterations. This fast convergence feature enables the method of the present invention to find solutions faster in practical applications, thereby improving the efficiency of the optimization process. A faster convergence speed not only means that the algorithm can complete the optimization task in a shorter time, but also helps to reduce the consumption of computing resources and cost investment. This is especially important for applications with high real-time requirements or limited computing resources.

[0219] In summary, through comparative experiments, it can be found that compared with the traditional neural network method, the prediction accuracy of the JSO-SVM prediction model proposed in the present invention is significantly improved, which provides more accurate data support for the dispatching decision of the power system, and helps to reduce the energy waste and dispatching costs caused by inaccurate predictions. In addition, the dynamic adjustment optimization algorithm realizes the real-time optimization of the power system. The algorithm can automatically adjust the distribution of power supply and load according to the real-time changes of power load, thereby effectively reducing the peak-to-average ratio and operating cost of the system. The experimental results show that the algorithm significantly improves the overall performance of the power system and enhances the stability and economy of the power grid.

[0220] An electronic device provided by an embodiment of the present invention includes a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the device is running, the processor communicates with the memory through the bus, and the processor executes the machine-readable instructions to perform any step of the above-mentioned power system operation optimization method based on the JSO-SVM model and dynamic adjustment.

[0221] Specifically, the above-mentioned memory and processor can be general-purpose memory and processor, which are not specifically limited here. When the processor runs the computer program stored in the memory, it can execute the above-mentioned power system operation optimization method based on the JSO-SVM model and dynamic adjustment.

[0222] Those skilled in the art will appreciate that the structure of the electronic device does not limit the electronic device and may include more or fewer components than shown in the figure, or combine or split certain components, or arrange the components differently.

[0223] In some embodiments, the electronic device may also include a touch screen that can be used to display a graphical user interface (e.g., a startup interface of an application) and receive user operations on the graphical user interface (e.g., startup operations on an application). The specific touch screen may include a display panel and a touch panel. The display panel may be configured in the form of LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode), etc. The touch panel may collect the user's contact or non-contact operations on or near it, and generate pre-set operation instructions, for example, the user uses any suitable object such as a finger, stylus, or accessories on or near the touch panel. In addition, the touch panel may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation and posture, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into information that the processor can process, and then sends it to the processor, and can receive and execute commands from the processor. In addition, the touch panel can be implemented by various types such as resistive, capacitive, infrared and surface acoustic wave, and any technology developed in the future can also be used to implement the touch panel. Further, the touch panel can cover the display panel, and the user can operate on or near the touch panel covered on the display panel according to the graphical user interface displayed on the display panel. After the touch panel detects the operation on or near it, it is transmitted to the processor to determine the user input, and then the processor provides corresponding visual output on the display panel in response to the user input. In addition, the touch panel and the display panel can be implemented as two independent components or integrated.

[0224] Corresponding to the startup method of the above-mentioned application, an embodiment of the present invention also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned power system operation optimization methods based on the JSO-SVM model and dynamic adjustment are executed.

[0225] The application startup device provided in the embodiment of the present application can be specific hardware on the device or software or firmware installed on the device. The device provided in the embodiment of the present application, its implementation principle and the technical effect produced are the same as those in the aforementioned method embodiment. For the sake of brief description, the parts not mentioned in the device embodiment can refer to the corresponding contents in the aforementioned method embodiment. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above can all refer to the corresponding processes in the aforementioned method embodiment, and will not be repeated here.

[0226] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0227] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or modules can be electrical, mechanical or other forms.

[0228] The modules described 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 network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0229] In addition, each functional module in the embodiments provided in the present application 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.

[0230] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0231] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0232] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0233] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for optimizing power system operation based on JSO-SVM model and dynamic adjustment, characterized in that: The steps include: Step S1, collecting data related to the operation of the power system, performing data processing, and finding historical days with similar load characteristics to the target day by calculating the daily feature similarity, and constructing a data set, wherein the relevant data includes historical load data of the power system, meteorological conditions, and user behavior patterns; Step S2, establishing a JSO-SVM load forecasting model based on SVM and JSO algorithms, and performing parameter optimization through model training; Step S3, setting the objective function and using the JSO-SVM load forecasting model to perform real-time load forecasting; Step S4, based on the load forecasting results obtained in real time, a dynamic adjustment algorithm is applied to optimize the operation strategy of the power system.

2. The power system operation optimization method based on the JSO-SVM model and dynamic adjustment according to claim 1 is characterized in that: The data processing includes: The normalization method is used to convert data of different dimensions into the same numerical range.

3. The power system operation optimization method based on the JSO-SVM model and dynamic adjustment according to claim 1 is characterized in that: The formula for calculating the daily feature similarity is: T(x i ,x j )=αD(x i ,x j )+βR(x i ,x j ) (1) In the above formula, the sample is x i =(x i1 ,x i2 ,…,x im ), and (0 <x ij <1); T(x i ,x j ) represents the similarity x between samples i ,x j ; D(x i ,x j ) represents the spatial distance similarity; R(x i ,x j ) represents the similarity of the change trend; α and β are weight coefficients; σ k represents the normalized value of the kth feature; m is the total number of features.

4. The power system operation optimization method based on the JSO-SVM model and dynamic adjustment according to claim 1 is characterized in that: The step S2, establishing a JSO-SVM load forecasting model based on the SVM and JSO algorithms, and performing parameter optimization through model training, includes: Define an ε-insensitive band that allows prediction errors to be within the ε range without penalty: Among them, y i is the predicted value, is the reference value, and ε is the insensitive band; Select the kernel function based on the characteristics of the data: K(x i ,x j )=φ(x i )·φ(x j ) (5) Among them, x i represents the input feature vector of the i-th sample, x j Represents the input feature vector of the jth sample; The optimal model is solved by minimizing the following objective function: Among them, w is the model weight vector, C is the penalty coefficient, L ò is the ε-insensitive loss function; The SVM model is trained using the similar day feature vectors, and the jellyfish algorithm is used to find the optimal weight vector w.

5. The power system operation optimization method based on the JSO-SVM model and dynamic adjustment according to claim 4 is characterized in that: The method of finding the optimal weight vector by using the jellyfish algorithm includes: ① Generation of initial population: According to the algorithm principle of JSO, a group of jellyfish individuals are randomly generated. Each individual consists of two dimensions, which represent the values ​​of C and g parameters respectively. The initial population is expressed by the following formula: X i =(C i ,g i ),i=1,2,...,N (7) Where N is the size of the population, C i and g i It is initialized randomly in the set search space; ②Fitness calculation: The fitness value of each individual is evaluated based on the classification performance of the SVM model on the validation set. Assuming that the cross-validation technique is used, the fitness function is expressed as the classification accuracy of the model: Fitness(X i )=Accuracy(C i ,g i ) (8); ③ Population update: In each iteration, JSO will update the population through the following strategies: 1) Ocean current movement: If the control parameter C(t) ≥ 0.5, the jellyfish individual updates its position based on the fitness value of the entire population: Among them, t represents the current number of iterations, Max_iter represents the maximum number of iterations, The formula for updating the position is as follows: Among them, X best is the optimal solution in the current population, and C(t) is the control parameter to ensure that the jellyfish moves smoothly in the ocean current; 2) Jellyfish swarm movement: When C(t)<0.5, the swarm enters the local search mode, and the jellyfish explore the search space through passive or active movement. The update formula is as follows: Passive exercise: Active Movement: Among them, X j is another jellyfish individual randomly selected to simulate the interaction between jellyfish; ④The iterative process of JSO stops when the maximum number of iterations is reached or the population converges, and the C and g values ​​finally returned are the optimized support vector machine model parameters.

6. The power system operation optimization method based on the JSO-SVM model and dynamic adjustment according to any one of claims 1 to 5, characterized in that: The objective function is: minC=C PCC +C FL +C MR +C DERG (13) Where C is the minimum total operating cost in the entire dispatching cycle of the distribution network; C PCC The cost of purchasing electricity from the upper power grid; C FL is the compensation cost of flexible load dispatch; MR Distributed power grid dispatching cost; C DERG Penalize the cost of curtailing renewable energy.

7. The power system operation optimization method based on the JSO-SVM model and dynamic adjustment according to claim 6 is characterized in that: The application of the dynamic adjustment algorithm to optimize the operation strategy of the power system includes: Transform constraints into explicit equivalence classes for judging the feasibility of each particle; Use linear decreasing or adaptive adjustment strategy to dynamically adjust the inertia weight ω: Where iter is the number of iterations; Dynamically adjust the acceleration constants c1 and c2 according to the convergence of the particle group: Dynamically adjust the maximum value v of particle speed max :

8. A power system operation optimization device based on JSO-SVM model and dynamic adjustment, characterized in that: include: A data collection module is used to collect data related to the operation of the power system, perform data processing, find historical days with similar load characteristics to the target day by calculating the similarity of daily characteristics, and construct a data set. The relevant data includes historical load data of the power system, meteorological conditions, and user behavior patterns; Model building module, used to build JSO-SVM load forecasting model based on SVM and JSO algorithms, and optimize parameters through model training; The load forecasting module is used to set the objective function and perform real-time load forecasting using the JSO-SVM load forecasting model; The strategy optimization module is used to optimize the operation strategy of the power system by applying a dynamic adjustment algorithm based on the load forecast results obtained in real time.

9. An electronic device, characterized in that: It includes a processor, a memory and a bus, the memory stores machine-readable instructions executable by the processor, when the electronic device is running, the processor and the memory communicate through the bus, and the processor executes the machine-readable instructions to perform the steps of the power system operation optimization method based on the JSO-SVM model and dynamic adjustment as described in any one of claims 1-7.

10. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, executes the steps of the power system operation optimization method based on the JSO-SVM model and dynamic adjustment as described in any one of claims 1 to 7.

Images