Dynamic load regulation and control method and system for distributed new energy access power grid

Abstract

The invention relates to the technical field of power grid load regulation and control, and discloses a dynamic load regulation and control method and system for a distributed new energy access power grid, and the method comprises the steps: obtaining multi-source data of the power grid, carrying out the dynamic alignment and feature extraction of the multi-source data, obtaining a multi-source feature tensor, inputting the multi-source feature tensor into a space-time coding model, and obtaining a space-time coding model; obtaining a space-time coding matrix; inputting the space-time coding matrix into a load prediction model to obtain a load prediction result; and acquiring real-time operation data of the power grid, and dynamically regulating and controlling the load of the power grid according to the real-time operation data and the load prediction result. According to the method, the problem of time granularity inconsistency is solved through multi-source data alignment optimization, short-term fluctuation and long-term trend are effectively separated through a space-time attention template and a space-time coding model, the accuracy of a load prediction result is improved through cross-scale joint modeling, the scheduling response efficiency can be improved, and the load prediction efficiency is improved. And high-timeliness support is provided for intelligent scheduling of a new energy high-permeability power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid load regulation, and in particular to a method and system for dynamically regulating the load of distributed renewable energy access to a power grid. Background Art

[0002] With the large-scale grid connection of distributed renewable energy sources such as wind power and photovoltaic power, the power grid faces many challenges in load regulation. On the one hand, the data sources obtained by the power grid are diverse and the time granularity is different. For example, renewable energy power generation data may be collected at a high frequency in minutes, reflecting the instantaneous fluctuations of photovoltaic or wind power, while meteorological data often records slowly changing characteristics such as temperature and wind speed in hours or days. Load data may be collected at fixed intervals, but sampling is missing or abnormal due to holidays or emergencies. Due to the heterogeneity of multi-source data, it is difficult for the power grid to effectively align and integrate multi-source data when performing load regulation analysis; on the other hand, due to the short-term fluctuations (such as photovoltaic cloud shading) and long-term trends (such as meteorological changes) in the output of renewable energy due to meteorological factors, this short-term and short-term coupling effect leads to the insufficient accuracy of the model results of data analysis and prediction using a single model; in addition, the current power grid regulation mostly relies on offline optimization and rule bases, and there is a delay in command generation, which leads to a sluggish dispatch response and makes it difficult to cope with the rapid fluctuations of renewable energy.

[0003] In view of the problems of multi-source data heterogeneity, multi-time scale feature conflicts and dispatch response hysteresis in the access of distributed renewable energy to the power grid, there is an urgent need for a load dynamic control method suitable for the operation optimization of smart grids with high energy penetration. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method and system for dynamic load control of distributed renewable energy access to the power grid, so as to solve the problems of multi-source data heterogeneity, multi-time scale feature conflicts and scheduling response hysteresis in the existing distributed renewable energy access to the power grid, so as to improve the accuracy of power grid load assessment and the real-time response capability of power grid scheduling.

[0005] In a first aspect, the present invention provides a method for dynamically controlling loads of distributed renewable energy sources connected to a power grid, the method comprising: Acquire multi-source data of distributed renewable energy access to the power grid, and dynamically align and extract features from the multi-source data to obtain a multi-source feature tensor, wherein the multi-source data includes renewable energy power generation data, meteorological data, and load data; Inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding model is constructed based on a deep learning model of a multi-head attention mechanism; Inputting the spatiotemporal coding matrix into a preset load forecasting model to obtain a load forecasting result, wherein the load forecasting model is obtained based on cross-scale joint modeling of a gated recurrent unit and a long short-term memory neural network; Real-time operation data of the distributed renewable energy access to the power grid is obtained, and the load of the distributed renewable energy access to the power grid is dynamically regulated based on the real-time operation data and the load forecast result.

[0006] Furthermore, before the step of inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, the method further includes: Performing speed-layered feature engineering on the multi-source feature tensor to obtain a speed-layered feature matrix; The speed hierarchical feature matrix is ​​input into a preset weight mapping model to obtain a spatiotemporal attention template, and the spatiotemporal attention template is embedded into a preset spatiotemporal coding model. The weight mapping model is constructed based on a multi-layer perceptron.

[0007] Furthermore, the step of performing velocity stratified feature engineering on the multi-source feature tensor to obtain a velocity stratified feature matrix includes: respectively calculating the fluctuation variance of the new energy power generation data, the trend sensitivity of the meteorological data and the stability index of the load data; The fluctuation variance, the trend sensitivity and the stability index are combined to obtain a speed stratified feature matrix.

[0008] Furthermore, the step of inputting the speed hierarchical feature matrix into a preset weight mapping model to obtain a spatiotemporal attention template includes: Mapping the speed hierarchical feature matrix into attention weights through a weight mapping model, wherein the attention weights include short-term attention weights and long-term attention weights; The attention weights are smoothly optimized through a dynamic programming algorithm to generate a spatiotemporal attention template.

[0009] Furthermore, the step of inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix includes: According to the multi-head attention separation mechanism of the spatiotemporal coding model, feature decomposition is performed on the multi-source feature tensor; Performing spatiotemporal position coding enhancement according to the dynamic load information to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding matrix includes short-term sensitive features and long-term trend features, and the dynamic load information includes historical load information and real-time load information; Among them, the enhanced spatiotemporal position coding is expressed by the following formula: In the formula, t represents the time step, pos represents the position dimension, PE(t,pos) represents the spatiotemporal position encoding, d represents the model dimension, α represents the load dynamic enhancement coefficient, and l t Indicates real-time load information. represents the historical load mean, Represents the historical load standard deviation.

[0010] Furthermore, the step of inputting the space-time coding matrix into a preset load forecasting model to obtain a load forecasting result includes: Inputting the spatiotemporal coding matrix into a preset load forecasting model, wherein the load forecasting model includes a gated recurrent unit, a long short-term memory neural network, and a time series load forecasting module, wherein the gated recurrent unit and the long short-term memory neural network are connected in parallel and then connected in series with the time series load forecasting module after being connected in parallel; Performing short-term feature extraction on the spatiotemporal coding matrix through a gated recurrent unit to obtain short-term fluctuation features; Extracting long-term features from the spatiotemporal coding matrix through a long short-term memory neural network to obtain long-term trend features; Performing weighted fusion on the short-term fluctuation feature and the long-term trend feature to obtain a fusion feature matrix; The fused feature matrix is ​​input into a time series load forecasting module to obtain a load forecasting result, which includes a short-term forecasting result and a long-term forecasting result.

[0011] Furthermore, the step of dynamically regulating the load of the distributed renewable energy connected to the power grid according to the real-time operation data and the load forecast result includes: According to the real-time operation data and the fusion feature matrix, minute-level response scheduling is performed on the load of the distributed renewable energy access to the power grid; According to the fusion characteristic matrix, the load of the distributed renewable energy connected to the power grid is optimally scheduled at the hourly level.

[0012] Furthermore, the step of performing minute-level response scheduling on the load of the distributed renewable energy access to the power grid according to the real-time operation data and the fusion characteristic matrix includes: Judging whether an abnormal event occurs when the distributed renewable energy is connected to the power grid based on the real-time operation data and the short-term fluctuation characteristics in the fusion feature matrix; According to the event type of the abnormal event, execute the corresponding minute-level response scheduling strategy; Among them, the event types include sudden drop in new energy output, sudden load change and line overload, and the minute-level response scheduling strategies corresponding to the event types are starting backup power supply, load transfer and dynamic load shedding.

[0013] Furthermore, the step of performing hourly optimization scheduling of the load of the distributed renewable energy connected to the power grid according to the fusion characteristic matrix includes: According to the long-term trend characteristics in the fusion feature matrix, a multi-objective optimization scheduling model is established with minimization of comprehensive operating costs and maximization of new energy consumption as objective functions; The multi-objective optimization scheduling model is solved to obtain an hourly optimization scheduling strategy.

[0014] In a second aspect, the present invention provides a load dynamic control system for accessing a distributed renewable energy source to a power grid, the system comprising: A data preprocessing module is used to obtain multi-source data of distributed renewable energy access to the power grid, and dynamically align and extract features from the multi-source data to obtain a multi-source feature tensor, wherein the multi-source data includes renewable energy power generation data, meteorological data, and load data; A spatiotemporal coding module, used for inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding model is constructed based on a deep learning model of a multi-head attention mechanism; A load forecasting module, used for inputting the spatiotemporal coding matrix into a preset load forecasting model to obtain a load forecasting result, wherein the load forecasting model is obtained based on cross-scale joint modeling of a gated recurrent unit and a long short-term memory neural network; The dynamic control module is used to obtain real-time operation data of distributed renewable energy access to the power grid, and dynamically control the load of distributed renewable energy access to the power grid based on the real-time operation data and the load forecast results.

[0015] The present invention provides a method and system for dynamically controlling the load of distributed renewable energy access to a power grid. The present invention can solve the problem of inconsistent time granularity and reduce feature loss and noise interference through multi-source data alignment optimization. It can effectively separate short-term fluctuations from long-term trends through adaptive spatiotemporal attention templates and spatiotemporal coding models. It can enhance the ability of time series modeling by combining gated recurrent units with long short-term memory neural networks, realize the fusion of short-term features and long-term features, improve the accuracy of load forecasting results, and improve the efficiency of dispatch response through short-term response dispatch and long-term optimization dispatch, so as to provide high-timeliness support for the intelligent dispatch of renewable energy high-penetration power grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of a method for dynamically controlling loads of distributed renewable energy sources connected to a power grid according to an embodiment of the present invention; Figure 2 It is a structural schematic diagram of a load dynamic control system for connecting distributed renewable energy to a power grid in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 The first embodiment of the present invention provides a method for dynamically controlling the load of a distributed renewable energy source connected to a power grid, which includes steps S10 to S40: Step S10, obtaining multi-source data of distributed renewable energy access to the power grid, and dynamically aligning and extracting features from the multi-source data to obtain a multi-source feature tensor, wherein the multi-source data includes renewable energy power generation data, meteorological data, and load data; Step S20, inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding model is constructed based on a deep learning model of a multi-head attention mechanism; Step S30, inputting the spatiotemporal coding matrix into a preset load forecasting model to obtain a load forecasting result, wherein the load forecasting model is obtained based on cross-scale joint modeling of a gated recurrent unit and a long short-term memory neural network; Step S40, obtaining real-time operation data of the distributed renewable energy access to the power grid, and dynamically regulating the load of the distributed renewable energy access to the power grid according to the real-time operation data and the load forecast result.

[0019] The present invention provides a method for dynamically regulating the load of a power grid for large-scale grid connection of distributed renewable energy sources. When performing grid regulation, it is necessary to analyze various data of the power grid. Since the data sources of the power grid are diverse, such as renewable energy output data, meteorological data, and load data, and there are differences in time scales and sampling frequencies among the multi-source data, traditional methods are difficult to effectively align and fuse. Therefore, this embodiment provides a method for dynamically aligning and extracting features from the multi-source data of the power grid, so as to unify the time granularity of the multi-source data and extract cross-scale features.

[0020] In this embodiment, multi-source data is obtained in units of one day, and the multi-source data includes renewable energy power generation data, meteorological data and load data, among which renewable energy power generation data is minute-level data, including photovoltaic and wind power power series, meteorological data is hour-level data, including temperature, wind speed and irradiance, and load data is data collected at fixed intervals, such as 15-minute intervals, and load data may be missing or abnormal due to holidays or emergencies. First, dynamic segmented statistics are performed on the multi-source data. When performing segmented statistics, segmented statistics are performed on renewable energy power generation data and meteorological data according to the sliding window mechanism, and linear interpolation is used to fill missing values ​​for load data, and 3σ principle and sliding window filtering are used to remove abnormal values. Specifically, a sliding window is used for renewable energy power generation data, for example, the window is 60 minutes and the step size is 15 minutes, the mean of the renewable energy power generation data is calculated, and the average output level within the hourly window is obtained; a sliding window is used for meteorological data, for example, the window is 3 hours and the step size is 1 hour, the trend slope is calculated, and the long-term change trend of meteorological conditions is quantified. The cubic spline interpolation method is used to fill in the missing values, so as to keep the second-order derivative of the load curve continuous and avoid the "sawtooth" effect of linear interpolation. The ADF test is passed to calculate the stability index of the load data.

[0021] At the same time, these multi-source data are aligned on the time axis to unify data of different frequencies to the same time granularity. Taking the time granularity of minutes as an example, since the renewable energy power generation data is at the minute level, it remains unchanged. The meteorological data and load data are upsampled, and the meteorological data and load data are filled with data through linear interpolation to make the data reach the time granularity of minutes. Of course, the multi-source data can also be aligned at the hourly level. In this case, the meteorological data remains unchanged, and the renewable energy power generation data and load data are downsampled, such as the mean statistics according to the hourly window. After dynamic alignment and feature extraction of multi-source data, the aligned multi-source feature tensor is obtained. The multi-source feature tensor is three-dimensional data, and its format is (time step × data source × feature). The features here include the mean of renewable energy power generation data, the trend slope of meteorological data, and the stability index of load data.

[0022] After obtaining the multi-source feature tensor, it is input into the pre-trained spatiotemporal coding model to perform cross-scale separation of short-term features and long-term features and encode the features. The specific steps include: According to the multi-head attention separation mechanism of the spatiotemporal coding model, feature decomposition is performed on the multi-source feature tensor; The spatiotemporal position coding enhancement is performed according to the dynamic load information to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding matrix includes short-term sensitive features and long-term trend features, and the dynamic load information includes historical load information and real-time load information.

[0023] In this embodiment, the spatiotemporal coding model is constructed based on the deep learning model Transformer of the multi-head attention mechanism, wherein the number of short-term attention heads and the number of long-term attention heads are the same, for example, both are 4 heads, and the multi-source feature tensors are feature split and spliced ​​through multi-head attention, and the corresponding position coding matrix is ​​output. Feature splitting refers to splitting the multi-source feature tensor into short-term features and long-term features. Short-term features are related to new energy power generation data and load data, such as new energy fluctuation variance, load instantaneous value, etc., while meteorological data and load data are related to long-term features, such as meteorological trend slope, load mean, load stability index, etc. Specific features can be flexibly selected according to actual conditions when constructing the model, and are not overly limited here. Through the multi-head attention separation mechanism, short-term fluctuations and long-term trends of power grid load can be captured simultaneously.

[0024] Since conventional position coding only relies on a fixed sine function, when distributed renewable energy is connected to the grid, the change in its actual load will also affect the fluctuation of renewable energy output. In order to improve the model's ability to capture multi-time scale characteristics of the grid, in this embodiment, the traditional position coding is integrated with dynamic load information to achieve spatial position coding enhancement. The enhanced spatiotemporal position coding formula is: Wherein, t represents the time step, i.e., the absolute position of the current moment in the sequence; pos represents the position dimension, i.e., the dimension index of the position encoding; d represents the model dimension, i.e., the hidden layer dimension of the model; PE(t,pos) represents the spatiotemporal position encoding, α represents the load dynamic enhancement coefficient, which is used to adjust the contribution weight of the load information in the position encoding to avoid the load feature dominating the original position information. Preferably, α is set to 0.3; l t Indicates real-time load information, that is, the total load value of the power grid at the current moment. Represents the historical load mean, which is used for standardization processing to eliminate dimensional differences and ensure compatibility of power grid data in different regions. Represents the historical load standard deviation, which is used to reflect the intensity of load fluctuation.

[0025] In the above formula, the sin function part is the traditional position coding, and the second half is the load dynamic enhancement term. It can be seen that the load dynamic enhancement term is the product of the load dynamic enhancement coefficient and the standardized load deviation. Therefore, the load dynamic enhancement term actually integrates the standardized load deviation into the position coding, so that the model can distinguish between load peak and valley periods and capture abnormal fluctuations. For example, during peak hours, , at this time the encoding value increases, strengthening the model's attention to the high load state. , that is, when there are outliers that do not meet the 3σ principle, the coded value will deviate significantly from the baseline, and the anomaly detection mechanism can be triggered at this time. In this embodiment, the minute-level time series dependency (such as fluctuations in new energy output) is captured through the sinusoidal term, and the hour-level trend is implied through the load enhancement term, that is, the load cycle law is reflected by the mean and standard deviation of the historical load, thereby realizing the multi-time scale fusion of high-frequency features and low-frequency features. The spatiotemporal position coding enhancement design of this embodiment significantly improves the spatiotemporal modeling capability of the Transformer model in complex power grid scenarios through the organic combination of static position signals and dynamic load characteristics, and provides theoretical guarantee and technical support for the precise scheduling of power grids with high penetration of new energy.

[0026] In order to eliminate abnormal fluctuations and provide a reliability reference for the attention mechanism of the Transformer model, in a preferred embodiment, before inputting the multi-source feature tensor into the Transformer model, the present invention generates a spatiotemporal attention template by performing hierarchical feature engineering on the multi-source feature tensor, and embeds the spatiotemporal attention template into the Transformer model, thereby improving the attention of the Transformer model to short-term fluctuations of the power grid. The specific steps include: Performing speed-layered feature engineering on the multi-source feature tensor to obtain a speed-layered feature matrix; The speed hierarchical feature matrix is ​​input into a preset weight mapping model to obtain a spatiotemporal attention template, and the spatiotemporal attention template is embedded into a preset spatiotemporal coding model. The weight mapping model is constructed based on a multi-layer perceptron.

[0027] In this embodiment, firstly, the multi-source feature tensor is subjected to speed-stratified feature engineering. For the renewable energy power generation data, the fluctuation variance is calculated according to the minimum value and hourly mean of the hourly renewable energy power generation data. According to the trend slope of the meteorological data, the sigmoid function is used to calculate the trend sensitivity: In the formula, represents the trend sensitivity at time t, represents the trend slope at time t, represents the slope threshold, k represents the sensitivity scaling factor, and t represents the time step, which can also be understood as the current moment.

[0028] For load data, the ADF test is used to calculate its stability index. If the index exists in the multi-source feature tensor, it can be directly extracted and used. Finally, the fluctuation variance, trend sensitivity and stability index are combined to obtain the speed stratified feature matrix.

[0029] Then the speed hierarchical feature matrix is ​​input into the preset weight mapping model to obtain the spatiotemporal attention template. The specific steps include: Mapping the speed hierarchical feature matrix into attention weights through a weight mapping model, wherein the attention weights include short-term attention weights and long-term attention weights; The attention weights are smoothly optimized through a dynamic programming algorithm to generate a spatiotemporal attention template.

[0030] In this embodiment, the weight mapping model is constructed based on a multi-layer perceptron MLP, and the weight mapping model is trained through historical data. The weight mapping model performs weight mapping on the input speed hierarchical feature matrix, thereby outputting a short-term attention weight sequence in the Transformer model, wherein the sum of the short-term attention weight and the long-term attention weight is 1. Therefore, the long-term attention weight sequence can be directly obtained through the short-term attention weight sequence.

[0031] In order to improve the stability of the weight sequence and provide the reliability of subsequent decision analysis, in this embodiment, the weight sequence is smoothed and optimized by a dynamic programming algorithm, with the objective function being the minimization of the difference in weights of adjacent time steps and the degree of deviation between the smoothed weights and the original estimated weights: In the formula, represents the short-term attention weight after smoothing at t time steps, Represents the original estimated weight output by the model, T represents the time window length, if the step length is 1 hour, then T=60, and λ represents the smoothing coefficient.

[0032] The above formula includes a smoothing term and a fidelity term, where the smoothing term is: , the fidelity term is: .

[0033] The purpose of the smoothing term is to minimize the difference in weights between adjacent time steps, avoid drastic jumps, thereby suppressing noise interference and improving the model's robustness to short-term fluctuations. The role of the fidelity term is to limit the degree of deviation between the smoothed weights and the original estimated weights, and prevent excessive smoothing from causing the model to lose key fluctuation information. The smoothing coefficient is used as a balance weight to control the smoothing term and the fidelity term. Set the initial value of the weight The original estimated weights for the first output of the MLP model , to avoid initial jumps, and then dynamically program and iteratively solve the objective function to obtain a smooth short-term attention weight sequence, and based on the constraint relationship between the short-term attention weight and the long-term attention weight, obtain the long-term attention weight sequence, and finally output the spatiotemporal attention template containing the short-term attention weight sequence and the long-term attention weight sequence.

[0034] This embodiment quantifies the speed of change at different time scales through speed-layered feature engineering, thereby adjusting the attention weight in the new energy scenario, and embedding the generated spatiotemporal attention template into the Transformer model, which can effectively improve the Transformer model's attention focus accuracy for key periods (such as a sudden drop in new energy output), thereby providing more accurate data support for subsequent load forecasting.

[0035] In the above embodiment, the spatiotemporal coding matrix output by the Transformer model actually contains short-term features and long-term features. Assuming that the spatiotemporal coding matrix is ​​256 dimensions, the first 128 dimensions are short-term features and the last 128 dimensions are long-term features. In order to overcome the problem of low prediction accuracy of a single model caused by the coupling effect of data at different time scales, in a preferred embodiment, the present invention constructs a load forecasting model through a cross-scale joint modeling method of a gated recurrent unit GRU and a long short-term memory neural network LSTM to achieve coordinated processing of short-term features and long-term features.

[0036] In this embodiment, the GRU and LSTM in the load forecasting model are connected in parallel. When the spatiotemporal coding matrix is ​​input into the trained load forecasting model, the spatiotemporal coding matrix will be input into the GRU and LSTM respectively. The GRU and LSTM extract short-term features and long-term features respectively according to the matrix dimensions. In the GRU, the retention ratio of information is controlled by updating the gate, and the historical memory is reset by resetting the gate to realize short-term feature processing. Its output is short-term fluctuation features at the minute level, such as new energy output variance, load instantaneous change rate, meteorological difference volatility, etc.; in the LSTM, long-term feature processing is realized by updating the cell state, and its output is long-term trend features at the hourly level, such as meteorological trend slope, load daily cycle pattern, new energy output long-term prediction deviation, etc.

[0037] For the short-term fluctuation features of GRU output and the long-term trend features of LSTM output, weighted fusion is performed through the cross-scale fusion gate to obtain the fusion feature matrix. When the features are fused, the weight value can be a preset value or determined by dynamic weight calculation. When dynamic weight is used, the sigmoid function is used to calculate the dynamic weight: In the formula, represents the weight of short-term volatility characteristics, Represents the preset dynamic weight matrix, GRU t Represents short-term fluctuation characteristics, LSTM t Indicates long-term trend characteristics.

[0038] Then, through dynamic weighted fusion, the fused feature matrix is ​​obtained: In the formula, Output t Represents the fused feature matrix, which has the format of (time steps × feature dimensions).

[0039] The fused feature matrix is ​​then input into the time series load forecasting module for load forecasting. The time series load forecasting module can be a fully connected layer or a lightweight time series forecasting module built based on linear regression or shallow neural network to generate the final load forecasting result.

[0040] In a preferred embodiment, the time series load forecasting module includes a short-term forecasting submodule and a long-term forecasting submodule, wherein the short-term forecasting submodule is constructed based on a lightweight time series convolutional network (TCN), and the long-term forecasting submodule is constructed based on an autoregressive integrated moving average (ARIMA). When making a forecast, the fused feature matrix is ​​input into the two submodules respectively, and the submodules extract short-term fluctuation features and long-term trend features through different dimensions of the matrix, thereby realizing load forecasting at different time scales. The cross-scale joint modeling method of this embodiment overcomes the problem that a single model is insufficient in capturing minute-level mutations and hour-level trends, and improves the accuracy of model prediction.

[0041] Based on the load forecast results and combined with the real-time operating data of distributed renewable energy access to the power grid, the load of distributed renewable energy access to the power grid can be dynamically regulated. When regulating, conventional power grid regulation methods can be referred to. The power grid can be regulated based on the predicted load and current operating data to match the power grid operation strategy with the load forecast results.

[0042] In order to improve the response time of power grid control and realize cross-scale dynamic scheduling decisions, in a preferred embodiment, the present invention provides a cross-scale dynamic control method, and the specific steps include: According to the real-time operation data and the fusion feature matrix, minute-level response scheduling is performed on the load of the distributed renewable energy access to the power grid; According to the fusion characteristic matrix, the load of the distributed renewable energy connected to the power grid is optimally scheduled at the hourly level.

[0043] In this embodiment, the dynamic control of the power grid load includes minute-level emergency response and hour-level optimization scheduling, wherein the step of minute-level emergency response includes: Judging whether an abnormal event occurs when the distributed renewable energy is connected to the power grid based on the real-time operation data and the short-term fluctuation characteristics in the fusion feature matrix; According to the event type of the abnormal event, execute the corresponding minute-level response scheduling strategy; Among them, the event types include sudden drop in new energy output, sudden load change and line overload, and the minute-level response scheduling strategies corresponding to the event types are starting backup power supply, load transfer and dynamic load shedding.

[0044] In this embodiment, the data used for the dispatch decision of minute-level emergency response includes the short-term fluctuation characteristics in the fusion feature matrix and the real-time operation data of the power grid, and the real-time operation data includes real-time load data, the battery state of charge SOC of the energy storage device and the real-time load rate of the line. Then, based on the short-term fluctuation characteristics and the real-time operation data of the power grid, it is judged whether the power grid has abnormal events, and the event types include sudden drop in new energy output, sudden load change and line overload.

[0045] Specifically, based on the variance of new energy output in the short-term fluctuation characteristics, determine whether a sudden drop in new energy output occurs. If it occurs and the short-term fluctuation exceeds the long-term benchmark, start the backup power supply, that is, start the energy storage discharge according to the battery state of charge SOC of the energy storage device. Based on the real-time load data and the instantaneous rate of change of load in the short-term fluctuation characteristics, determine whether a load mutation event occurs. If it occurs, the power grid is load-shifted. Determine whether a line overload event occurs based on the real-time load rate of the line. If it occurs, the power grid is dynamically load-shedded. It should be noted that abnormal event determination can also be performed based on other short-term fluctuation characteristics and real-time operation data. The specific abnormal event type and the corresponding determination method can be flexibly set according to the actual situation. This embodiment is only used as a preferred method and not a specific limitation.

[0046] In a preferred embodiment, the step of hourly optimization scheduling includes: According to the long-term trend characteristics in the fusion feature matrix, a multi-objective optimization scheduling model is established with minimization of comprehensive operating costs and maximization of new energy consumption as objective functions; The multi-objective optimization scheduling model is solved to obtain an hourly optimization scheduling strategy.

[0047] In this embodiment, a multi-objective optimization scheduling model is established according to the long-term trend characteristics to achieve hourly optimization scheduling, wherein the multi-objective optimization scheduling model takes minimization of comprehensive operating costs and maximization of new energy consumption as objective functions, and its formula is expressed as follows: In the formula, represents the long-term trend characteristics of time step t, Represents the weather forecast data at time step t. The weather forecast value can be obtained from the weather forecast department. represents the historical load data at time step t, Represents the dispatch instruction vector, which is the power allocation plan for each power source (such as coal-fired units, new energy power stations, energy storage systems, etc.) that needs to be determined in the optimization model within a specified time period. θ represents the weight of the meteorological matching item, which is used to emphasize the dominant influence of meteorological forecasts on dispatching and prevent insufficient output due to sudden meteorological changes. γ represents the weight of the historical fitting item, which is used to balance the innovation of the model with historical experience and avoid overfitting new data. η represents the sparsity constraint coefficient, which is used to control the sparsity of the dispatch instruction. The larger the value, the fewer units are started and stopped. T represents the time window length, which is used to cover the typical cycle of power grid dispatching, and θ+γ=1. Preferably, θ is set to 0.8 and γ is set to 0.2. It should be noted that in multiple formulas of the present invention, there are time window length T and time step t. The specific values ​​are different in different formulas, but their meanings are the same. They will no longer be represented by different parameters.

[0048] The above objective function includes three parts, namely, meteorological matching term, historical matching term and sparsity constraint. The meteorological matching term is the square of the L2 norm between the long-term trend characteristics and the meteorological forecast data: The objective function minimizes the difference between long-term trend characteristics (such as load daily cycle, new energy output deviation) and meteorological forecast data (such as temperature trend, wind speed change) to ensure that the dispatch plan conforms to the changes in meteorological conditions, prevent insufficient output due to sudden meteorological changes, and reduce unplanned costs caused by sudden weather changes. At the same time, using the square of the L2 norm in the meteorological matching term can ensure that the objective function is a convex function, so that there is a global optimal solution, and give higher weight to large deviations (because the error will grow squarely), prompting the model to pay more attention to the overall matching of meteorological trends.

[0049] The historical fitting term is the L1 norm between the long-term trend characteristics and the historical load data: The historical fitting item is to constrain the deviation of the dispatch plan from the historical experience pattern (such as holiday load peak) and avoid excessive investment costs caused by over-reliance on forecasts.

[0050] The sparsity constraint is the L1 norm of the scheduling instruction vector: Sparse constraints can reduce the frequent start and stop of units, thereby reducing the loss cost of equipment.

[0051] Furthermore, since the long-term trend characteristics include the long-term forecast deviation of renewable energy output (such as the daily cumulative deviation of wind power), during the optimization process, the model can dynamically adjust the ratio of thermal power to renewable energy output by minimizing the deviation. For example, if the predicted wind power output is higher than the actual output (the deviation is negative), the thermal power reserve is increased; if the prediction is low (the deviation is positive), the thermal power is reduced and renewable energy is consumed first. Although the wind abandonment term is not explicitly included in the objective function, the wind abandonment rate can be indirectly reduced through the L2 norm constraint of the meteorological matching term. In addition, the sparsity constraint prompts energy storage to charge at the peak of renewable energy output (low electricity price) and discharge at the valley (reducing the demand for thermal power), which indirectly improves the utilization rate of renewable energy. Therefore, by iteratively solving the objective function, an hourly optimization scheduling strategy that minimizes the comprehensive operating cost and maximizes the consumption of renewable energy can be obtained.

[0052] The present embodiment provides a method for dynamic load control of distributed renewable energy access to a power grid. The present invention solves the problem of inconsistent time granularity and reduces feature loss and noise interference through multi-source data alignment optimization. It effectively separates short-term fluctuations from long-term trends through adaptive spatiotemporal attention templates and spatiotemporal coding models. It enhances the time series modeling capability by combining gated recurrent units with long short-term memory neural networks, realizes the fusion of short-term and long-term features, improves the accuracy of load forecasting results, and improves the efficiency of scheduling response through short-term response scheduling and long-term optimization scheduling, providing high-timeliness support for the intelligent scheduling of power grids with high penetration of renewable energy.

[0053] See also Figure 2 Based on the same inventive concept, a load dynamic control system for accessing a distributed renewable energy source to a power grid is proposed in a second embodiment of the present invention, comprising: The data preprocessing module 10 is used to obtain multi-source data of distributed renewable energy access to the power grid, and dynamically align and extract features from the multi-source data to obtain a multi-source feature tensor, wherein the multi-source data includes renewable energy power generation data, meteorological data, and load data; A spatiotemporal coding module 20 is used to input the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding model is constructed based on a deep learning model of a multi-head attention mechanism; A load forecasting module 30, configured to input the spatiotemporal coding matrix into a preset load forecasting model to obtain a load forecasting result, wherein the load forecasting model is obtained based on cross-scale joint modeling of a gated recurrent unit and a long short-term memory neural network; The dynamic control module is used to obtain real-time operation data of distributed renewable energy access to the power grid, and dynamically control the load of distributed renewable energy access to the power grid based on the real-time operation data and the load forecast results.

[0054] The technical features and technical effects of the load dynamic control system for accessing a distributed renewable energy source to a power grid proposed in an embodiment of the present invention are the same as those of the method proposed in an embodiment of the present invention, and will not be described in detail here. Each module in the above-mentioned load dynamic control system for accessing a distributed renewable energy source to a power grid can be implemented in whole or in part through software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above modules.

[0055] In summary, an embodiment of the present invention proposes a method and system for dynamic load regulation of distributed renewable energy access to a power grid. The method obtains multi-source data of distributed renewable energy access to a power grid, and dynamically aligns and extracts features from the multi-source data to obtain a multi-source feature tensor, wherein the multi-source data includes renewable energy power generation data, meteorological data, and load data; the multi-source feature tensor is input into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, and the spatiotemporal coding model is constructed based on a deep learning model of a multi-head attention mechanism; the spatiotemporal coding matrix is ​​input into a preset load forecasting model to obtain a load forecasting result, and the load forecasting model is obtained based on cross-scale joint modeling of a gated recurrent unit and a long short-term memory neural network; real-time operation data of distributed renewable energy access to a power grid is obtained, and according to the real-time operation data and the load forecasting result, the load of distributed renewable energy access to a power grid is dynamically regulated. The present invention solves the problem of inconsistent time granularity and reduces feature loss and noise interference through multi-source data alignment optimization. It effectively separates short-term fluctuations from long-term trends through adaptive spatiotemporal attention templates and spatiotemporal coding models. It enhances the timing modeling capability by combining gated recurrent units and long short-term memory neural networks, realizes the fusion of short-term and long-term features, improves the accuracy of load forecasting results, and improves the efficiency of dispatch response through short-term response dispatch and long-term optimization dispatch, providing high-timeliness support for the intelligent dispatch of new energy high-penetration power grids.

[0056] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0057] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A method for dynamically controlling the load of distributed renewable energy connected to a power grid, characterized in that: include: Acquire multi-source data of distributed renewable energy access to the power grid, and dynamically align and extract features from the multi-source data to obtain a multi-source feature tensor, wherein the multi-source data includes renewable energy power generation data, meteorological data, and load data; Inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding model is constructed based on a deep learning model of a multi-head attention mechanism; Inputting the spatiotemporal coding matrix into a preset load forecasting model to obtain a load forecasting result, wherein the load forecasting model is obtained based on cross-scale joint modeling of a gated recurrent unit and a long short-term memory neural network; Real-time operation data of distributed renewable energy access to the power grid is obtained, and the load of distributed renewable energy access to the power grid is dynamically regulated based on the real-time operation data and the load forecast result.

2. The method for dynamic load control of distributed renewable energy access to a power grid according to claim 1, characterized in that: Before the step of inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, the method further includes: Performing speed-layered feature engineering on the multi-source feature tensor to obtain a speed-layered feature matrix; The speed hierarchical feature matrix is ​​input into a preset weight mapping model to obtain a spatiotemporal attention template, and the spatiotemporal attention template is embedded into a preset spatiotemporal coding model. The weight mapping model is constructed based on a multi-layer perceptron.

3. The method for dynamic load control of distributed renewable energy access to a power grid according to claim 2, characterized in that: The step of performing speed hierarchical feature engineering on the multi-source feature tensor to obtain a speed hierarchical feature matrix comprises: respectively calculating the fluctuation variance of the renewable energy power generation data, the trend sensitivity of the meteorological data, and the stability index of the load data; The fluctuation variance, the trend sensitivity and the stability index are combined to obtain a speed stratified feature matrix.

4. The method for dynamic load control of distributed renewable energy access to a power grid according to claim 2, characterized in that: The step of inputting the speed hierarchical feature matrix into a preset weight mapping model to obtain a spatiotemporal attention template comprises: Mapping the speed hierarchical feature matrix into attention weights through a weight mapping model, wherein the attention weights include short-term attention weights and long-term attention weights; The attention weights are smoothly optimized through a dynamic programming algorithm to generate a spatiotemporal attention template.

5. The method for dynamic load control of distributed renewable energy access to a power grid according to claim 1, characterized in that: The step of inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix comprises: According to the multi-head attention separation mechanism of the spatiotemporal coding model, feature decomposition is performed on the multi-source feature tensor; Performing spatiotemporal position coding enhancement according to the dynamic load information to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding matrix includes short-term sensitive features and long-term trend features, and the dynamic load information includes historical load information and real-time load information; Among them, the enhanced spatiotemporal position coding is expressed by the following formula: In the formula, t represents the time step, pos represents the position dimension, and PE (t,pos) represents the spatiotemporal position encoding, d represents the model dimension, α represents the load dynamic enhancement coefficient, l t Indicates real-time load information, represents the historical load mean, Represents the historical load standard deviation.

6. The method for dynamic load control of distributed renewable energy access to a power grid according to claim 1, characterized in that: The step of inputting the space-time coding matrix into a preset load forecasting model to obtain a load forecasting result comprises: Inputting the spatiotemporal coding matrix into a preset load forecasting model, wherein the load forecasting model includes a gated recurrent unit, a long short-term memory neural network, and a time series load forecasting module, wherein the gated recurrent unit and the long short-term memory neural network are connected in parallel and then connected in series with the time series load forecasting module after being connected in parallel; Performing short-term feature extraction on the spatiotemporal coding matrix through a gated recurrent unit to obtain short-term fluctuation features; Extracting long-term features from the spatiotemporal coding matrix through a long short-term memory neural network to obtain long-term trend features; Performing weighted fusion on the short-term fluctuation feature and the long-term trend feature to obtain a fusion feature matrix; The fused feature matrix is ​​input into a time series load forecasting module to obtain a load forecasting result, which includes a short-term forecasting result and a long-term forecasting result.

7. The method for dynamic load control of distributed renewable energy access to a power grid according to claim 6, characterized in that: The step of dynamically regulating the load of the distributed renewable energy source connected to the power grid according to the real-time operation data and the load forecast result comprises: According to the real-time operation data and the fusion feature matrix, minute-level response scheduling is performed on the load of the distributed renewable energy access to the power grid; According to the fusion characteristic matrix, the load of the distributed renewable energy connected to the power grid is optimally scheduled at the hourly level.

8. The method for dynamic load control of distributed renewable energy access to a power grid according to claim 7, characterized in that: The step of performing minute-level response scheduling on the load of the distributed renewable energy access to the power grid according to the real-time operation data and the fusion feature matrix comprises: Judging whether an abnormal event occurs when the distributed renewable energy is connected to the power grid based on the real-time operation data and the short-term fluctuation characteristics in the fusion feature matrix; According to the event type of the abnormal event, execute the corresponding minute-level response scheduling strategy; Among them, the event types include sudden drop in new energy output, sudden load change and line overload, and the minute-level response scheduling strategies corresponding to the event types are starting backup power supply, load transfer and dynamic load shedding.

9. The method for dynamic load control of distributed renewable energy access to a power grid according to claim 7, characterized in that: The step of performing hourly optimization scheduling on the load of the distributed renewable energy connected to the power grid according to the fusion characteristic matrix comprises: According to the long-term trend characteristics in the fusion feature matrix, a multi-objective optimization scheduling model is established with minimization of comprehensive operating costs and maximization of new energy consumption as objective functions; The multi-objective optimization scheduling model is solved to obtain an hourly optimization scheduling strategy.

10. A load dynamic control system for accessing a distributed renewable energy source to a power grid, characterized in that: include: A data preprocessing module is used to obtain multi-source data of distributed renewable energy access to the power grid, and dynamically align and extract features from the multi-source data to obtain a multi-source feature tensor, wherein the multi-source data includes renewable energy power generation data, meteorological data, and load data; A spatiotemporal coding module, used for inputting the multi-source feature tensor into a preset spatiotemporal coding model to obtain a spatiotemporal coding matrix, wherein the spatiotemporal coding model is constructed based on a deep learning model of a multi-head attention mechanism; A load forecasting module, used for inputting the spatiotemporal coding matrix into a preset load forecasting model to obtain a load forecasting result, wherein the load forecasting model is obtained based on cross-scale joint modeling of a gated recurrent unit and a long short-term memory neural network; The dynamic control module is used to obtain real-time operation data of distributed renewable energy access to the power grid, and dynamically control the load of distributed renewable energy access to the power grid based on the real-time operation data and the load forecast results.

Images