CN-122026516-A - New energy power distribution strategy decision optimization management method based on deep learning

CN122026516ACN 122026516 ACN122026516 ACN 122026516ACN-122026516-A

Abstract

The invention relates to the technical field of operation optimization of power systems, in particular to a new energy power distribution strategy decision optimization management method based on deep learning. The data is analyzed to determine the optimal input window and prediction step size for the supply and consumption prediction model, and the predictions are performed in parallel to obtain a future supply and consumption prediction set. And carrying out multi-round strategy simulation and evaluation on the prediction sets by using a strategy decision optimization network, and screening out the strategy with the highest internal evaluation score as a target power distribution strategy. According to the invention, the prediction precision is improved by enhancing the time sequence relevance of the data, and the dynamic adaptability and decision robustness of the allocation strategy are enhanced by a closed-loop simulation optimization mechanism.

Inventors

WANG DAN
CHEN CHANG
SHEN PING
CHEN MO

Assignees

沈阳大学

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. The new energy power distribution strategy decision optimization management method based on deep learning is characterized by comprising the following steps of: acquiring a historical power supply data set of a plurality of energy nodes and a historical power consumption data set of a plurality of power consumption units in a preset historical period through a power data acquisition network; performing multidimensional data alignment and time sequence coupling analysis on the historical power supply data set and the historical power consumption data set to generate joint supply consumption data; In a depth prediction model construction space, carrying out fluctuation mode clustering and feature association analysis on the combined supply consumption data, and determining optimal input variable windows and prediction step sizes of an energy supply prediction model and a consumption prediction model; Based on the optimal input variable window and the prediction step length, the energy supply prediction model and the consumption prediction model are operated in parallel to respectively obtain a future supply prediction data set of a plurality of energy nodes and a future consumption prediction data set of a plurality of power consumption units; And carrying out multi-round strategy simulation and strategy evaluation on the future supply prediction data set and the future consumption prediction data set by using a strategy decision optimization network, and screening out a candidate distribution strategy with the highest evaluation score in the strategy decision optimization network as a target power distribution strategy.
2. The deep learning-based new energy power distribution strategy decision optimization management method according to claim 1, wherein the multidimensional data alignment and time sequence coupling analysis of the historical power supply data set and the historical power consumption data set comprises: performing data cleaning on the historical power supply data of each energy node in the historical power supply data set to generate a cleaned supply data sequence; performing data cleaning on the historical power consumption data of each power consumption unit in the historical power consumption data set to generate a cleaned consumption data sequence; establishing a unified time stamp index, and carrying out data slicing and alignment on the cleaned supply data sequence and the cleaned consumption data sequence based on the unified time stamp index to generate supply consumption data with time alignment; and performing time sequence correlation calculation and mutual information analysis on the time-aligned supply consumption data to generate the joint supply consumption data.
3. The deep learning-based new energy power distribution strategy decision optimization management method according to claim 2, wherein the performing, in a deep prediction model construction space, fluctuation mode clustering and feature association analysis on the joint supply consumption data includes: extracting a fluctuation feature vector of a supply data part and a fluctuation feature vector of a consumption data part in the joint supply consumption data; inputting the fluctuation feature vector of the supply data part and the fluctuation feature vector of the consumption data part to independent cluster analysis engines respectively; performing cluster analysis on the fluctuation feature vector of the supply data part and the fluctuation feature vector of the consumption data part through the cluster analysis engine to divide a plurality of fluctuation modes; and calculating the transition probability among different fluctuation modes, and constructing a supply prediction model configuration table and a consumption prediction model configuration table by combining the prediction error historical data.
4. The deep learning-based new energy power distribution strategy decision optimization management method as claimed in claim 3, wherein the determining the optimal input variable window and prediction step size of the energy supply prediction model and the consumption prediction model comprises: traversing a plurality of fluctuation modes in the supply prediction model configuration table, and searching an optimal input variable window and a prediction step length of the energy supply prediction model with the aim of minimizing a prediction error; Traversing a plurality of fluctuation modes in the consumption prediction model configuration table, and searching an optimal input variable window and a prediction step length of the consumption prediction model by taking a minimum prediction error as a target; And respectively storing the searched optimal input variable window and prediction step length of the energy supply prediction model, the optimal input variable window and prediction step length of the consumption prediction model into the supply prediction model configuration table and the consumption prediction model configuration table.
5. The deep learning-based new energy power distribution strategy decision optimization management method according to claim 4, wherein the parallel running of the energy supply prediction model and the consumption prediction model includes: According to the supply prediction model configuration table, configuring a current input variable window and a prediction step length of the energy supply prediction model, and inputting supply data of the energy nodes acquired in real time into the energy supply prediction model; According to the consumption prediction model configuration table, configuring a current input variable window and a prediction step length of the consumption prediction model, and inputting consumption data of a power consumption unit acquired in real time into the consumption prediction model; The energy supply prediction model outputs the future supply prediction data set based on the supply data input by the energy supply prediction model; the consumption prediction model outputs the future consumption prediction data set based on the consumption data input thereto.
6. The deep learning-based new energy power distribution policy decision optimization management method according to claim 1, wherein the performing, by using a policy decision optimization network, multiple rounds of policy simulation and policy evaluation on the future supply prediction data set and the future consumption prediction data set includes: Initializing the strategy decision optimizing network, wherein the strategy decision optimizing network comprises a strategy generating module, a power network simulation module and a strategy evaluation module; The strategy generation module generates a plurality of initial power distribution strategies according to the future supply prediction data set and the future consumption prediction data set; for each initial power distribution strategy, the power network simulation module simulates the execution process of the power network simulation module in a preset future time period and outputs a simulated supply consumption track; And the strategy evaluation module analyzes the simulated supply consumption track according to a preset evaluation index and calculates an evaluation score of each initial power distribution strategy.
7. The deep learning-based new energy power distribution strategy decision optimization management method according to claim 6, wherein the screening out the candidate distribution strategy with the highest evaluation score in the strategy decision optimization network comprises: The strategy decision optimization network sorts the evaluation scores of all the initial power distribution strategies, and selects one or more initial power distribution strategies with the highest evaluation scores as elite strategy sets; the strategy generation module generates a new power distribution strategy based on the elite strategy set through strategy crossing and strategy mutation operation to form a new generation strategy set; The strategy decision optimizing network repeatedly executes the power network simulation and the strategy evaluation process for each new power distribution strategy in the new generation strategy set to calculate a new evaluation score; And iteratively executing the processes of strategy evaluation, elite strategy selection, strategy generation and simulation until a preset iteration termination condition is reached, and determining the strategy with the highest evaluation score in the elite strategy set after the last iteration as the candidate allocation strategy.
8. The deep learning-based new energy power distribution strategy decision optimization management method according to claim 7, further comprising model refinement and parameter calibration of the strategy decision optimization network: Executing the target power distribution strategy in an actual power network, and collecting actual supply data and actual consumption data generated in an actual execution process; Calculating a supply forecast deviation between the actual supply data and the future supply forecast data set, and a consumption forecast deviation between the actual consumption data and the future consumption forecast data set; reversely adjusting internal parameters of the energy supply prediction model and the consumption prediction model according to the supply prediction deviation and the consumption prediction deviation; and comparing the actual execution effect of the target power distribution strategy with an evaluation score given by the strategy decision optimization network in simulation, and adjusting the evaluation index weight of the strategy evaluation module according to a comparison result.
9. The deep learning-based new energy power distribution strategy decision optimization management method as claimed in claim 2, wherein the data cleaning comprises: Identifying missing data points and outlier data points in the historical power supply data set and the historical power consumption data set; For missing data points, filling the missing data points by adopting data of time points adjacent to the missing data points through a time sequence interpolation method; And for abnormal data points, adopting a statistical distribution-based method, and replacing data points exceeding a preset reasonable range with statistical data of the historical power supply data set or the historical power consumption data set in a corresponding historical period.
10. The deep learning-based new energy power distribution strategy decision optimization management method according to claim 3, wherein the constructing a supply prediction model configuration table and a consumption prediction model configuration table further comprises: The supply prediction model configuration table and the consumption prediction model configuration table are associated, and a supply consumption mode mapping relation is established; The supply consumption pattern mapping relation is used for recording joint probability of the specific fluctuation pattern corresponding to the consumption data part when the supply data part is in the specific fluctuation pattern; and the supply consumption mode mapping relation is used as priori knowledge and is input into the strategy decision optimization network for guiding the strategy generation module to generate an initial power distribution strategy which is more in line with the historical joint fluctuation rule.

Description

New energy power distribution strategy decision optimization management method based on deep learning Technical Field The invention relates to the technical field of operation optimization of power systems, in particular to a new energy power distribution strategy decision optimization management method based on deep learning. Background In the field of power system scheduling containing high-proportion new energy, the prior art mainly carries out power distribution decision according to independent time sequence prediction results. Conventional methods model and predict historical data of an energy supply end and an electric power consumption end respectively, and the prediction process usually adopts a fixed time window or an empirical step size. The technology regards supply and consumption as separable subsystems, the data preprocessing stage lacks systematic processing of space-time consistency among multi-source data, the selection of the input variables and the step length of the prediction model also often depends on priori knowledge, and the optimal time sequence dependency relation cannot be dynamically mined from the data. The generation of the existing allocation strategy is mostly based on the prediction result, and the calculation is performed by using a static optimization model such as a linear programming model, a heuristic rule and the like. These models rely on deterministic input scenarios, the decision process of which is essentially a single, open-loop optimization. They cannot simulate and evaluate the linkage effects and long-term performance of allocation strategies in various future uncertain scenarios at the decision stage, and lack a strategy learning mechanism that can dynamically interact with the environment and self-iterate through feedback. This results in the existing schemes being insufficiently robust and adaptive to decisions in the face of strong fluctuations and load randomness of the new energy. There is a need for a method that can improve the quality and relevance of data for prediction from the source and achieve closed-loop learning and optimization of allocation strategies in a simulated environment to cope with complex uncertainties in system operation. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a new energy power distribution strategy decision optimization management method based on deep learning. In order to achieve the purpose, the invention adopts the following technical scheme that the new energy power distribution strategy decision optimization management method based on deep learning comprises the following steps: acquiring a historical power supply data set of a plurality of energy nodes and a historical power consumption data set of a plurality of power consumption units in a preset historical period through a power data acquisition network; performing multidimensional data alignment and time sequence coupling analysis on the historical power supply data set and the historical power consumption data set to generate joint supply consumption data; In a depth prediction model construction space, carrying out fluctuation mode clustering and feature association analysis on the combined supply consumption data, and determining optimal input variable windows and prediction step sizes of an energy supply prediction model and a consumption prediction model; Based on the optimal input variable window and the prediction step length, the energy supply prediction model and the consumption prediction model are operated in parallel to respectively obtain a future supply prediction data set of a plurality of energy nodes and a future consumption prediction data set of a plurality of power consumption units; And carrying out multi-round strategy simulation and strategy evaluation on the future supply prediction data set and the future consumption prediction data set by using a strategy decision optimization network, and screening out a candidate distribution strategy with the highest evaluation score in the strategy decision optimization network as a target power distribution strategy. As a further aspect of the present invention, the performing multidimensional data alignment and time-series coupling analysis on the historical power supply data set and the historical power consumption data set includes: performing data cleaning on the historical power supply data of each energy node in the historical power supply data set to generate a cleaned supply data sequence; performing data cleaning on the historical power consumption data of each power consumption unit in the historical power consumption data set to generate a cleaned consumption data sequence; establishing a unified time stamp index, and carrying out data slicing and alignment on the cleaned supply data sequence and the cleaned consumption data sequence based on the unified time stamp index to generate supply consumption data with time alignment; and performi