Park integrated energy system multi-element load joint prediction model and optimization method

Abstract

The invention relates to the technical field of energy system management and optimization, in particular to multi-element load prediction and optimization for a park integrated energy system. A park comprehensive energy system relates to conversion, storage and distribution of various energy types, and complex coupling relations exist among various loads. A traditional load prediction method usually only aims at a single load type and ignores the coupling characteristic between loads, so that the prediction precision is limited. In order to solve the problem, the invention provides a multi-element load joint prediction model and an optimization method for a park integrated energy system. According to the method, a coupling relation between a load and various factors is mined by using a maximum information coefficient (MIC), and combined prediction of the multi-element load is realized based on a time convolution network (TCN). Meanwhile, an improved particle swarm optimization (MPSO) algorithm is introduced to optimize model parameters, so that the prediction precision and the model generalization ability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy system management and optimization, and in particular to a multi-load joint prediction model and optimization method applied to a park integrated energy system. Background Art

[0002] With the rapid development of energy internet, the park integrated energy system, as an important part of energy internet, has great significance for improving energy utilization efficiency and reducing energy costs due to its efficient operation and optimized scheduling. However, the park integrated energy system involves the conversion, storage and distribution of multiple energy types (such as electricity, cold, heat, etc.), and there are complex coupling relationships between various loads, making load forecasting a challenging task. Traditional load forecasting methods often only predict a single load type, ignoring the coupling characteristics between loads, resulting in limited prediction accuracy. Summary of the invention

[0003] To solve the above problems, the present invention proposes a joint prediction model and optimization method for multiple loads of the park integrated energy system. The method fully considers the coupling relationship between multiple loads, uses the maximum information coefficient (MIC) to mine the coupling relationship between loads and various factors, and realizes the joint prediction of multiple loads based on the time convolution network (TCN). At the same time, the improved particle swarm optimization algorithm (MPSO) is introduced to optimize the model parameters to improve the prediction accuracy and model generalization ability.

[0004] Data preprocessing: Preprocess the historical load data, meteorological data and other input features of the park’s integrated energy system, including deleting outliers, filling missing values, and normalizing to reduce input noise and improve model training efficiency.

[0005] Coupling relationship mining: The maximum information coefficient (MIC) is used to analyze the coupling relationship between load and various factors, and the important features that affect load fluctuations are screened out as model input.

[0006] Model construction: Construct a multi-load joint forecasting model based on the temporal convolutional network (TCN). The TCN layer adopts dilated causal convolution and residual block structure, which can process time series data in parallel, capture long-term dependencies, and ensure that the model output is consistent with the input sequence length. On the basis of the TCN layer, the multi-task learning (MTL) algorithm is added to regard the multi-load forecasting as multiple subtasks to achieve multi-task parallel forecasting. At the same time, the attention mechanism (AM) is used to give each subtask different weights in the shared layer to improve the generalization ability of the model. The attention mechanism (AM) can focus the attention of the neural network on effective feature information, avoid wasting too much time in noise, and thus improve the efficiency of the neural network. In the AM algorithm, the calculation formulas for the attention weight matrix a and the feature vector v are:

[0007] z t,k =u s ·tan(w s h t +b)

[0008]

[0009] Among them, zt,k is the metric of feature importance; tan is the activation function; w s , b and u s are the randomly initialized attention weight matrix, bias matrix and time series matrix respectively; h t is the hidden state corresponding to the tth element; a t ,k is the attention probability distribution value of the hidden layer output; H is the final output value.

[0010] Parameter optimization: The modified particle swarm optimization algorithm (MPSO) is introduced to optimize the parameters of the TCN model. In order to solve the problem that the traditional particle swarm optimization algorithm (PSO) is prone to premature convergence, the average particle distance method is introduced to improve the probability of searching for the global optimum, and the current state of the population is judged according to the overall change of particle fitness. The calculation formula of the modified particle swarm optimization algorithm (MPSO) is as follows:

[0011]

[0012] in:

[0013]

[0014] Where D(t) is the discreteness of the particle distribution in the population; L is the maximum diagonal length of the search space; M is the number of particles; n is the dimension of the solution; x id is the d-dimensional coordinate of the ith particle; x d is the average value of the d-dimensional coordinates of all particles; f i is the current fitness of particle i, is the average fitness of the population; f is the limit σ 2 The normalized scale factor of .

[0015] When σ 2 The smaller the particle size, the smaller the discreteness of the particles. When the number of iterations increases, D(t) and σ 2 Get smaller and smaller, when D(t) and σ 2 When it is less than a certain value, it is considered to enter the late stage of search. Therefore, the optimal position of the current particle swarm is retained to guide the particles out of the local optimum.

[0016] The MPSO algorithm improves the probability of searching for the global optimum by introducing the average particle distance method, avoids premature convergence, and thus improves the prediction accuracy of the model.

[0017] Performance evaluation: Evaluation indicators such as root mean square error (RMSE) and mean absolute percentage error (MAPE) are used to evaluate the model performance to ensure that the model has high prediction accuracy and generalization ability.

[0018] Improve prediction accuracy: The present invention fully considers the coupling relationship between multiple loads, uses MIC to mine the coupling relationship, and realizes joint prediction of multiple loads based on TCN, which effectively improves the prediction accuracy.

[0019] Improve model generalization ability: By introducing multi-task learning and attention mechanism, the model can better adapt to changes in different load types and improve the model generalization ability.

[0020] Optimization parameter selection: The MPSO algorithm is used to optimize the model parameters, which avoids the problem of premature convergence of traditional optimization algorithms and improves the probability of the model searching for the global optimum.

[0021] Wide application: The present invention can be applied to load forecasting in various parks across the country, helping the parks to achieve faster and more accurate load forecasting and providing strong support for the optimized operation of the park's comprehensive energy system. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of an energy storage system for a seismic station according to an embodiment of the present invention;

[0023] Figure 2 Schematic diagram of a fault warning module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0025] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0026] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0027] like Figure 1 As shown in the figure, the multi-load joint prediction model of the park integrated energy system can accurately predict the PIES short-term electric load, cooling load and heating load at the same time, and use the maximum information coefficient to explore the coupling relationship between the load and various factors; use the time convolution network to predict the cooling, heating and electric loads; use the improved particle swarm optimization algorithm to select the optimal parameters of the time convolution network; add a multi-task learning algorithm, use the attention mechanism to give each subtask different weights in the shared layer, improve the generalization ability of the prediction model, and realize the joint prediction of multiple loads.

[0028] like Figure 2 As shown in the figure, the overall framework of the multivariate load forecasting model (PTA-MTL) of the park integrated energy system includes data preprocessing module, load forecasting module, performance evaluation module and parameter optimization module.

[0029] Data preprocessing module: First, outliers are deleted and missing values ​​are filled. Second, MIC is used to screen important factors as input features of the model. Finally, in order to avoid unnecessary errors caused by large differences in the dimensions of feature data, the data is normalized. The calculation formula is as follows:

[0030]

[0031] Among them, X′ is the normalized value; X is the original data; X max , X min are the maximum and minimum values ​​of the data respectively.

[0032] Load forecasting module: This module includes input layer, TCN layer (shared layer) and output layer. The TCN layer (shared layer) uses a hard parameter sharing mechanism to learn and share the relevant information of each task, and uses MPSO to optimize its parameters.

[0033] Performance evaluation module: Root mean square error (RMSE) and mean absolute percentage error (MAPE) are used as model evaluation indicators. The calculation process is as follows:

[0034]

[0035] Where n is the number of data; i represents the i-th data; Y i ,y i are the predicted value and the true value of the data respectively.

[0036] Parameter optimization module: The average particle distance method is introduced to improve the probability of searching for the global optimum, and the current state of the population is judged according to the overall change of particle fitness. The calculation formula of the modified particle swarm optimization algorithm (MPSO) is as follows:

[0037]

[0038] in:

[0039]

[0040] Where D(t) is the discreteness of the particle distribution in the population; L is the maximum diagonal length of the search space; M is the number of particles; n is the dimension of the solution; x id is the d-dimensional coordinate of the ith particle; x d is the average value of the d-dimensional coordinates of all particles; f i is the current fitness of particle i, is the average fitness of the population; f is the limit σ 2 The normalized scale factor of 2 The smaller the particle size, the smaller the discreteness of the particles. When the number of iterations increases, D(t) and σ 2 Get smaller and smaller, when D(t) and σ2 When it is less than a certain value, it is considered to enter the late stage of search. Therefore, the optimal position of the current particle swarm is retained to guide the particles out of the local optimum.

[0041] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0042] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A multi-load joint prediction model for a park integrated energy system, characterized in that: include: Data preprocessing module: used to preprocess the historical load data, meteorological data and other relevant input features of the park's integrated energy system. The preprocessing steps include deleting outliers, filling missing values, normalization, etc. to reduce input noise and improve model training efficiency; the data preprocessing module further includes data smoothing, detrending or seasonal adjustment to enhance data stability and consistency; Coupling relationship mining module: Use the maximum information coefficient (MIC) to analyze the coupling relationship between load and various factors, and select important features that affect load fluctuations as model input; The coupling relationship mining module also uses correlation analysis, principal component analysis, grey correlation analysis or other statistical methods to comprehensively evaluate the importance and relevance of features; Load forecasting module: The temporal convolutional network (TCN) is used to realize the joint forecasting of multiple loads. The TCN layer adopts the dilated causal convolution and residual block structure, which can process time series data in parallel and capture long-term dependencies. On the basis of the TCN layer, a multi-task learning (MTL) algorithm is introduced to regard multivariate load forecasting as multiple subtasks to achieve multi-task parallel forecasting, and use the attention mechanism (AM) to give each subtask different weights in the shared layer; the multi-task learning algorithm adopts hard parameter sharing, soft parameter sharing or hybrid parameter sharing to achieve information sharing and collaborative optimization between different subtasks; the attention mechanism adopts self-attention mechanism, multi-head attention mechanism, cross attention mechanism or other forms of attention mechanism to dynamically adjust the weights of different subtasks in the shared layer; Parameter optimization module: The improved particle swarm optimization algorithm (MPSO) is used to optimize the parameters of the TCN model. The MPSO algorithm improves the probability of searching for the global optimum and avoids premature convergence by introducing the average particle distance, dynamically adjusting the inertia weight, and increasing the global search strategy. The parameter optimization module also includes adjusting the parameters of the MPSO algorithm, such as adjusting the number of particles, the number of iterations, the search space range, etc., to further improve the optimization performance and search efficiency of the algorithm. Performance evaluation module: The model performance is evaluated using evaluation indicators such as root mean square error (RMSE), mean absolute percentage error (MAPE), and mean square error (MSE) to ensure that the model has high prediction accuracy and generalization ability; the performance evaluation module further includes the calculation and analysis of other evaluation indicators, such as the stability, robustness, and reliability of the prediction results, to comprehensively evaluate the performance of the model.

2. The multi-load joint prediction model for the park integrated energy system according to claim 1 is characterized in that: The data preprocessing module also includes seasonal decomposition and trend analysis of the input data to separate seasonal components and trend components, thereby improving the model's ability to predict seasonal fluctuations and trend changes.

3. The multi-load joint prediction model for the park integrated energy system according to claim 1 is characterized in that: The coupling relationship mining module also uses deep learning technology, such as convolutional neural network (CNN), recurrent neural network (RNN) or graph neural network (GNN), to model and analyze the complex relationship between load and various factors.

4. The multi-load joint prediction model for the park integrated energy system according to claim 1 is characterized in that: The load forecasting module also includes ensemble learning methods, such as random forests, gradient boosted trees, or extreme gradient boosting (XGBoost), to further improve the prediction accuracy and robustness of the model.

5. The multi-load joint prediction model for the park integrated energy system according to claim 1 is characterized in that: The parameter optimization module also includes the application of other optimization algorithms, such as genetic algorithm (GA), differential evolution algorithm (DE) or simulated annealing algorithm (SA), to compare the impact of different optimization algorithms on model performance and select the best optimization algorithm.

6. The multi-load joint prediction model for the park integrated energy system according to claim 1 is characterized in that: The performance evaluation module also includes evaluating the performance of the model at different time scales, different load types or different application scenarios to verify the applicability and versatility of the model.

7. A method for load forecasting using the multi-load joint forecasting model for a park integrated energy system according to any one of claims 1 to 6, characterized in that: include: a) Input the historical load data, meteorological data and other relevant input features of the park's integrated energy system into the data preprocessing module for preprocessing; b) Use the coupling relationship mining module to mine the coupling relationship between load and various factors, and screen out the important features that affect load fluctuations; c) Input the filtered features into the load forecasting module for multi-element load joint forecasting; d) Optimizing the parameters of the load forecasting module using the parameter optimization module; e) Use the performance evaluation module to evaluate the prediction results, and adjust and optimize the model based on the evaluation results; f) Apply the optimized model to the actual load forecast of the park’s integrated energy system to achieve accurate and efficient energy scheduling and management.

8. A park integrated energy system, characterized in that: It comprises a multi-load joint prediction model of a comprehensive energy system of a park as described in any one of claims 1 to 6, and an energy dispatching and management system based on the model, which is used to realize intelligent dispatching and optimized management of various energy types in the park.

Images