Power dispatching method and system based on power load prediction

Through deep learning and reinforcement learning and other technologies, the power load prediction model and scheduling strategy are constructed, and the existing power scheduling methods are solved. The problem of insufficient response capabilities for rapidly changing market demand and emergencies is achieved, and efficient and flexible power resource scheduling is improved. The operation efficiency and safety of the power grid.

CN119029836AInactive Publication Date: 2024-11-26SHANDONG ENERGY SHENGLUNENG CHEM ORDOS NEW ENERGY CO LTD +1

Patent Information

Application Number
CN202410864164.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-30
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing power scheduling methods cannot accurately respond to rapidly changing market demands and emergencies, lack real-time data processing and learning mechanisms, and fail to make full use of advanced data analysis and machine learning technologies, resulting in untimely and inaccurate scheduling decision responses. Inadequate adaptability and flexibility.

Method used

The power load prediction model is constructed using deep learning algorithms, combined with real-time data to perform short-term load prediction, and generate an optimized power scheduling strategy. The genetic algorithm and reinforcement learning optimization scheduling strategy are used to dynamically adjust the start-stop arrangement and power adjustment of the generator set, and real-time monitoring is carried out. And update the model to improve prediction accuracy and the effectiveness of scheduling strategies.

Benefits of technology

It significantly improves the prediction accuracy and response speed of the power scheduling system, optimizes the operating efficiency and stability of the power grid, generates the most cost-effective power scheduling strategy, enhances the ability to respond to complex and variable power demand environments, and improves the Reliability and safety of the power grid.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119029836A_ABST
    Figure CN119029836A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of power system management, in particular to a power dispatching method and system based on power load prediction, and the method comprises the following steps: S1, collecting multi-source data from a power system, and carrying out the preprocessing of the collected data; s2, constructing a power load prediction model by adopting a deep learning algorithm; s3, performing real-time prediction on future power load to generate a short-term load prediction result; s4, generating an optimized power dispatching strategy; s5, the generated power dispatching strategy is issued to each generator set and a load management unit; and S6, carrying out self-learning and updating on the power load prediction model and the scheduling strategy generation model. According to the invention, by integrating the advanced data processing technology, the deep learning algorithm and the real-time scheduling strategy, the accuracy and the response speed of power scheduling are remarkably improved, and meanwhile, the scheduling strategy is continuously optimized through a self-learning mechanism, so that the efficiency, the economical efficiency and the reliability of power grid operation are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power system management, and in particular to a power dispatching method and system based on power load forecasting. Background Art

[0002] With the rapid development of industrialization and urbanization, the growth of electricity demand is accelerating, and the load management of power grids has become an important and complex issue. The dispatch of power systems is the key to ensuring the safety, stability and economic operation of power grids. Traditional power dispatch relies on historical data and experience judgment. This method often seems inadequate when facing emergencies or atypical load changes. In addition, with the large-scale access of renewable energy, the uncertainty of power grid operation has further increased, which has put higher requirements on power dispatch strategies and requires more accurate load forecasting and real-time dispatching capabilities.

[0003] Existing technologies face several major challenges when dealing with power system dispatching issues: First, existing load forecasting methods are often unable to accurately respond to rapidly changing market demands and emergencies, resulting in dispatching decisions that are not timely or accurate enough; second, many dispatching systems lack effective real-time data processing and learning mechanisms and cannot dynamically adjust dispatching strategies based on real-time feedback, which limits the adaptability and flexibility of the system; finally, existing systems often fail to fully utilize advanced data analysis and machine learning technologies and fail to achieve data-driven intelligent decision support.

[0004] Therefore, there is an urgent need for a new power dispatching method and system that can integrate advanced data processing, load forecasting, and dispatching strategy generation technologies to improve the operating efficiency and response capabilities of the power system. Summary of the invention

[0005] Based on the above objectives, the present invention provides a power dispatching method and system based on power load forecasting.

[0006] The power dispatching method according to the power load forecast comprises the following steps:

[0007] S1: Collect multi-source data from the power system, including historical load data, meteorological data, and user power consumption behavior data, and pre-process the collected data, including data cleaning, data normalization, and data noise reduction;

[0008] S2: Based on the preprocessed data, a deep learning algorithm is used to build a power load forecasting model to predict the power load in different time periods in the future;

[0009] S3: Using the constructed power load forecasting model and the real-time collected power system data, the future power load is predicted in real time to generate short-term load forecast results, which include load curves in different time periods in the future;

[0010] S4: Generate an optimized power dispatching strategy based on the real-time load forecast results, combined with the operation constraints of the power grid and the power market price information, the dispatching strategy includes the start and stop arrangement of the generator set, power adjustment instructions, and load transfer plan;

[0011] S5: The generated power dispatching strategy is issued to each generator set and load management unit, and the dispatching execution is monitored in real time. The power dispatching strategy is dynamically adjusted according to the actual execution status and feedback information;

[0012] S6: Conduct self-learning and updating of the power load forecasting model and the dispatching strategy generation model, and use the feedback data during the dispatching execution process to continuously optimize the model parameters to improve the prediction accuracy of the model and the effectiveness of the dispatching strategy.

[0013] Furthermore, S1 also includes clustering analysis of user electricity consumption behavior data to identify electricity consumption characteristics of different types of users. Specifically, the user electricity consumption behavior data is first preliminarily cleaned to remove noise and abnormal data; then a clustering algorithm is used to perform cluster analysis on the cleaned data to divide the users into several categories; finally, the electricity consumption characteristics of each type of user are analyzed, and the typical electricity consumption pattern of each type of user is extracted to provide input features for the load forecasting model.

[0014] Furthermore, the load forecasting model in S2 is constructed by using a hybrid model combining a convolutional neural network and a long short-term memory network. Specifically, the convolutional neural network is used to extract local features in the power load time series; the features extracted by the convolutional neural network are then input into the long short-term memory network to capture the long-term dependencies of the time series; finally, the output of the long short-term memory network is converted into a load forecasting result through a fully connected layer, and the output layer of the model uses a linear activation function to ensure the continuity of the predicted value.

[0015] Furthermore, the real-time load forecast in S3 uses sliding window technology to dynamically update the input data of the forecast model to improve the real-time performance of the forecast. Specifically, the power load time series data is divided into windows of fixed length, and the data in each window is used as an input sample; each time the sliding window moves forward one time step, the latest data of the input sample is updated to ensure that the input data of the forecast model always reflects the latest load change trend; finally, load forecasting is performed on the data in each window to generate a rolling forecast result.

[0016] Furthermore, the dispatching strategy generation in S4 uses a genetic algorithm to optimize the power dispatching scheme to obtain the optimal dispatching strategy. Specifically, the power dispatching problem is first expressed as an optimization problem, and a fitness function is defined; then a group of candidate dispatching schemes are initialized as the initial population of the genetic algorithm; then, crossover and mutation operations are performed on the candidate schemes in the population to generate new candidate schemes, and selection is performed according to the fitness function to retain the schemes with higher fitness; finally, the crossover, mutation and selection operations are iteratively performed until the stopping condition is met, and the optimal dispatching strategy is output.

[0017] Furthermore, the scheduling execution and feedback step in S5 includes dynamically adjusting the start and stop schedule of the generator set to cope with the fluctuation of the power load, specifically including:

[0018] Real-time monitoring of load changes: Through sensors and monitoring systems, the actual change data of power load is obtained in real time, compared with the predicted results, and the load deviation is calculated. The load deviation is set as ΔP(t). The specific calculation formula is: ΔP(t) = P 实际 (t)-P 预测 (t), where P 实际 (t) is the actual load, P 预测 (t) is the predicted load;

[0019] Start-stop adjustment strategy: adjust the start-stop arrangement of the generator set according to the load deviation. If ΔP(t)>0, increase the output power of the generator set or enable the standby generator set; if ΔP(t)<0, reduce the output power of the generator set or disable some generator sets;

[0020] Power change during the adjustment process: the power adjustment of the generator set ΔQ(t), the specific calculation formula is:

[0021] ΔQ(t)=K·ΔP(t), where K is the adjustment coefficient, which is determined according to the adjustment capability of the generator set and the system requirements;

[0022] Dynamic adjustment strategy: During the adjustment process, the operating status and load changes of the generator sets are monitored in real time, and the start-stop schedule and power output are continuously adjusted to ensure the stable operation of the power system;

[0023] Feedback data collection: Collect the adjusted generator set operation data and power load data as input data for model self-learning and updating, and optimize the prediction model and scheduling strategy.

[0024] Furthermore, the S6 self-learning and model updating specifically utilizes a reinforcement learning algorithm to continuously adjust model parameters based on historical dispatch data and feedback data. Specifically, the power dispatch process is first regarded as a reinforcement learning problem, and the state space, action space and reward function are defined; then the reinforcement learning algorithm is trained using historical dispatch data to adjust the parameters of the load forecasting model and the dispatch strategy generation model; in the actual dispatch process, the model parameters are updated in real time according to the dispatch execution results and feedback data to improve the model's adaptability.

[0025] Furthermore, the power dispatch strategy generation model includes a constraint generation module that takes into account power grid safety constraints to ensure the safety of the dispatch strategy. Specifically, the power grid safety constraints are first defined, including thermal stability constraints of power lines, node voltage constraints, and system frequency constraints; then, when generating the dispatch strategy, it is checked whether each dispatch scheme meets the safety constraints, and the schemes that do not meet the requirements are modified or discarded; when optimizing the dispatch strategy, the schemes that meet the safety constraints are given priority.

[0026] Furthermore, the dispatch strategy generation also includes considering the electricity market price information and optimizing the economic efficiency of the dispatch strategy. Specifically, the real-time price data of the electricity market is first collected as one of the inputs for the dispatch strategy generation; then, when generating the dispatch strategy, the economic costs of different dispatch schemes are calculated by combining the load forecast results and market price information; finally, the dispatch scheme with the lowest economic cost is selected to ensure that the power system minimizes operating costs while meeting load demand and safety constraints.

[0027] A power dispatching system based on power load forecasting, used to implement the above-mentioned power dispatching method based on power load forecasting, comprises the following modules:

[0028] Data collection and processing module: used to collect multi-source data from the power system, including historical load data, meteorological data, and user power consumption behavior data; and perform data cleaning, data normalization, and data noise reduction processing steps to ensure the accuracy and completeness of the data;

[0029] Load forecasting model module: receives pre-processed data from the data collection and processing module, and uses deep learning algorithms to build a power load forecasting model;

[0030] Real-time load forecasting module: uses the model built by the load forecasting model module, combined with the power system data collected in real time from the data collection and processing module, to make real-time forecasts of future power loads and generate short-term load forecast results;

[0031] Dispatch strategy generation module: Based on the prediction results provided by the real-time load prediction module, combined with the operation constraints of the power grid and the price information of the power market, it generates an optimized power dispatch strategy, including the start and stop arrangements of the generator sets, power adjustment instructions, and load transfer plans;

[0032] Dispatch execution and feedback module: executes the power dispatch strategy formulated by the dispatch strategy generation module, including issuing the dispatch strategy to each generator set and load management unit, and monitoring the dispatch execution in real time; according to the actual execution status and feedback information, feedbacks the adjustment information to the dispatch strategy generation module to dynamically adjust the power dispatch strategy;

[0033] Self-learning and model updating module: Based on the feedback data provided by the dispatch execution and feedback module, the power load forecasting model and dispatch strategy generation model are self-learned and updated.

[0034] Beneficial effects of the present invention:

[0035] The present invention significantly improves the prediction accuracy and response speed of the power dispatching system by integrating advanced data processing technology, deep learning algorithm and real-time dispatching strategy generation mechanism. By utilizing the real-time data acquisition and processing module and the load forecasting model module, the system can accurately predict the changing trend of power load and adjust the allocation of power resources in time, thereby optimizing the operation efficiency and stability of the power grid. In addition, the system's dispatching strategy generation module combines the price information of the power market and the constraints of the power grid operation to generate a cost-effective power dispatching strategy, effectively reduce operating costs and improve economic benefits.

[0036] The present invention continuously collects feedback information on scheduling execution and updates the prediction and scheduling models. The system can continuously learn and adapt to the complex and changeable power demand environment. This dynamic learning and adjustment mechanism ensures that the power scheduling strategy always remains in the optimal state, while enhancing the system's ability to respond to future load changes and emergencies, significantly improving the reliability and safety of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 A schematic diagram of a power dispatching method according to power load forecasting according to an embodiment of the present invention;

[0039] Figure 2Schematic diagram of a power dispatching system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0041] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0042] like Figure 1 As shown, the power dispatching method according to the power load forecast includes the following steps:

[0043] S1: Collect multi-source data from the power system, including historical load data, meteorological data, and user power consumption behavior data, and pre-process the collected data, including data cleaning, data normalization, and data noise reduction, to ensure the accuracy and completeness of the data;

[0044] S2: Based on the preprocessed data, a deep learning algorithm is used to build a power load forecasting model to predict the power load in different time periods in the future. The forecasting model includes an input layer, multiple hidden layers, and an output layer. The model parameters are optimized through the back propagation algorithm to improve the model's forecasting accuracy.

[0045] S3: Using the constructed power load forecasting model and the real-time collected power system data, the future power load is predicted in real time to generate short-term load forecast results, which include load curves in different time periods in the future;

[0046] S4: Based on the real-time load forecast results, combined with the operation constraints of the power grid and the price information of the power market, an optimized power dispatching strategy is generated. The dispatching strategy includes the start and stop arrangements of the generator sets, power adjustment instructions, and load transfer plans;

[0047] S5: The generated power dispatching strategy is issued to each generator set and load management unit, and the dispatching execution is monitored in real time. According to the actual execution status and feedback information, the power dispatching strategy is dynamically adjusted to ensure the safe and stable operation of the power system;

[0048] S6: Conduct self-learning and updating of the power load forecasting model and the dispatching strategy generation model, and use the feedback data during the dispatching execution process to continuously optimize the model parameters to improve the prediction accuracy of the model and the effectiveness of the dispatching strategy.

[0049] S1 also includes clustering analysis of user electricity consumption behavior data to identify the electricity consumption characteristics of different types of users. Specifically, the user electricity consumption behavior data is first preliminarily cleaned to remove noise and abnormal data; then a clustering algorithm (such as the K-means algorithm or the DBSCAN algorithm) is used to cluster the cleaned data to divide the users into several categories; finally, the electricity consumption characteristics of each type of user are analyzed, and the typical electricity consumption pattern of each type of user is extracted to provide input features for the load forecasting model.

[0050] The load forecasting model in S2 is constructed by a hybrid model combining convolutional neural network (CNN) and long short-term memory network (LSTM). Specifically, convolutional neural network (CNN) is used to extract local features in the power load time series, and the size and number of convolution kernels of the convolution layer are set according to the characteristics of the data; then the features extracted by the convolutional neural network are input into the long short-term memory network (LSTM) to capture the long-term dependencies of the time series, and the number of layers and hidden units of the long short-term memory network are adjusted as needed; finally, the output of the long short-term memory network is converted into a load forecasting result through a fully connected layer, and the output layer of the model uses a linear activation function to ensure the continuity of the predicted value.

[0051] The real-time load forecasting in S3 uses sliding window technology to dynamically update the input data of the forecasting model and improve the real-time performance of the forecast. Specifically, the power load time series data is divided into windows of fixed length, and the data in each window is used as an input sample. Each time the sliding window moves forward one time step, the latest data of the input sample is updated to ensure that the input data of the forecasting model always reflects the latest load change trend. Finally, load forecasting is performed on the data in each window to generate rolling forecast results.

[0052] The dispatch strategy generation in S4 uses a genetic algorithm to optimize the power dispatch plan to obtain the optimal dispatch strategy. Specifically, the power dispatch problem is first expressed as an optimization problem, and a fitness function is defined, which comprehensively considers the operating cost of the power system, the start-up and shutdown cost of the generator set, the load transfer cost, etc.; then a group of candidate dispatch plans are initialized as the initial population of the genetic algorithm; then the candidate plans in the population are crossover and mutation operations are performed to generate new candidate plans, and they are selected according to the fitness function to retain the plans with higher fitness; finally, the crossover, mutation and selection operations are iteratively performed until the stopping condition is met (such as reaching the maximum number of iterations or the fitness function converges), and the optimal dispatch strategy is output.

[0053] The dispatch execution and feedback step in S5 includes dynamic adjustment of the start and stop schedule of the generator set to cope with the fluctuation of power load, including:

[0054] Real-time monitoring of load changes: Through sensors and monitoring systems, the actual change data of power load is obtained in real time, compared with the predicted results, and the load deviation is calculated. The load deviation is set as ΔP(t). The specific calculation formula is: ΔP(t) = P 实际 (t)-P 预测 (t), where P 实际 (t) is the actual load, P 预测 (t) is the predicted load;

[0055] Start-stop adjustment strategy: adjust the start-stop arrangement of the generator set according to the load deviation. If ΔP(t)>0, increase the output power of the generator set or enable the standby generator set; if ΔP(t)<0, reduce the output power of the generator set or disable some generator sets;

[0056] Power change during the adjustment process: the power adjustment of the generator set ΔQ(t), the specific calculation formula is:

[0057] ΔQ(t)=K·ΔP(t), where K is the adjustment coefficient, which is determined according to the adjustment capability of the generator set and the system requirements;

[0058] Dynamic adjustment strategy: During the adjustment process, the operating status and load changes of the generator sets are monitored in real time, and the start-stop schedule and power output are continuously adjusted to ensure the stable operation of the power system;

[0059] Feedback data collection: Collect the adjusted generator set operation data and power load data as input data for model self-learning and updating, and optimize the prediction model and scheduling strategy.

[0060] S6 self-learning and model updating specifically utilizes reinforcement learning algorithms to continuously adjust model parameters based on historical dispatch data and feedback data. Specifically, the power dispatch process is first regarded as a reinforcement learning problem, and the state space, action space and reward function are defined; then the reinforcement learning algorithm (such as deep Q network DQN or policy gradient algorithm) is trained using historical dispatch data to adjust the parameters of the load forecasting model and the dispatch strategy generation model; in the actual dispatch process, the model parameters are updated in real time according to the dispatch execution results and feedback data to improve the model's adaptability.

[0061] The power dispatch strategy generation model includes a constraint generation module that considers the grid security constraints to ensure the safety of the dispatch strategy. Specifically, the grid security constraints are defined first, including the thermal stability constraints of the power lines, the node voltage constraints, and the system frequency constraints. Then, when generating the dispatch strategy, it is checked whether each dispatch scheme meets the security constraints, and the schemes that do not meet the requirements are modified or discarded. When optimizing the dispatch strategy, the schemes that meet the security constraints are given priority to ensure the safe and stable operation of the power system.

[0062] Dispatch strategy generation also includes considering electricity market price information and optimizing the economic efficiency of the dispatch strategy. Specifically, real-time price data from the electricity market is first collected as one of the inputs for dispatch strategy generation. Then, when generating the dispatch strategy, the economic costs of different dispatch schemes are calculated by combining load forecast results and market price information. Finally, the dispatch scheme with the lowest economic cost is selected to ensure that the power system minimizes operating costs while meeting load demand and safety constraints.

[0063] like Figure 2 As shown, a power dispatching system based on power load forecasting is used to implement the above-mentioned power dispatching method based on power load forecasting, and includes the following modules:

[0064] Data collection and processing module: used to collect multi-source data from the power system, including historical load data, meteorological data, and user power consumption behavior data; and perform data cleaning, data normalization, and data noise reduction processing steps to ensure the accuracy and completeness of the data;

[0065] Load forecasting model module: receives pre-processed data from the data collection and processing module, and uses deep learning algorithms to build a power load forecasting model;

[0066] Real-time load forecasting module: uses the model built by the load forecasting model module, combined with the power system data collected in real time from the data collection and processing module, to make real-time forecasts of future power loads and generate short-term load forecast results;

[0067] Dispatch strategy generation module: Based on the prediction results provided by the real-time load prediction module, combined with the operation constraints of the power grid and the price information of the power market, it generates an optimized power dispatch strategy, including the start and stop arrangements of the generator sets, power adjustment instructions, and load transfer plans;

[0068] Dispatch execution and feedback module: executes the power dispatch strategy formulated by the dispatch strategy generation module, including issuing the dispatch strategy to each generator set and load management unit, and monitoring the dispatch execution in real time; according to the actual execution status and feedback information, feedbacks the adjustment information to the dispatch strategy generation module to dynamically adjust the power dispatch strategy;

[0069] Self-learning and model updating module: Based on the feedback data provided by the dispatch execution and feedback module, the power load forecasting model and dispatch strategy generation model are self-learned and updated.

[0070] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0071] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Power dispatching method based on power load forecast, It is characterized in that The following steps are involved: S1: Collect multi-source data from the power system, including historical load data, meteorological data, and user power consumption behavior data, and pre-process the collected data, including data cleaning, data normalization, and data noise reduction; S2: Based on the preprocessed data, a deep learning algorithm is used to build a power load forecasting model to predict the power load in different time periods in the future; S3: Using the constructed power load forecasting model and the real-time collected power system data, the future power load is predicted in real time to generate short-term load forecast results, which include load curves in different time periods in the future; S4: Generate an optimized power dispatching strategy based on the real-time load forecast results, combined with the operation constraints of the power grid and the power market price information, the dispatching strategy includes the start and stop arrangement of the generator set, power adjustment instructions, and load transfer plan; S5: The generated power dispatching strategy is issued to each generator set and load management unit, and the dispatching execution is monitored in real time. The power dispatching strategy is dynamically adjusted according to the actual execution status and feedback information; S6: Conduct self-learning and updating of the power load forecasting model and the dispatching strategy generation model, and use the feedback data during the dispatching execution process to continuously optimize the model parameters to improve the prediction accuracy of the model and the effectiveness of the dispatching strategy.

2. The method for dispatching electric power according to electric power load forecasting according to claim 1, It is characterized in that The S1 also includes clustering analysis of user electricity consumption behavior data to identify electricity consumption characteristics of different types of users. Specifically, the user electricity consumption behavior data is first preliminarily cleaned to remove noise and abnormal data; then a clustering algorithm is used to perform cluster analysis on the cleaned data to divide the users into several categories; finally, the electricity consumption characteristics of each type of user are analyzed, and the typical electricity consumption pattern of each type of user is extracted to provide input features for the load forecasting model.

3. The power dispatching method according to power load forecasting according to claim 1, It is characterized in that The load forecasting model in S2 is constructed by using a hybrid model combining a convolutional neural network and a long short-term memory network. Specifically, the convolutional neural network is used to extract local features in the power load time series; the features extracted by the convolutional neural network are then input into the long short-term memory network to capture the long-term dependencies of the time series; finally, the output of the long short-term memory network is converted into a load forecasting result through a fully connected layer, and the output layer of the model uses a linear activation function to ensure the continuity of the predicted value.

4. The method for dispatching electric power according to electric power load forecasting according to claim 1, It is characterized in that The real-time load forecast in S3 uses sliding window technology to dynamically update the input data of the forecast model and improve the real-time performance of the forecast. Specifically, the power load time series data is divided into windows of fixed length, and the data in each window is used as an input sample; each time the sliding window moves forward by one time step, the latest data of the input sample is updated to ensure that the input data of the forecast model always reflects the latest load change trend; finally, load forecasting is performed on the data in each window to generate a rolling forecast result.

5. The power dispatching method according to power load forecasting according to claim 1, It is characterized in that The dispatch strategy generation in S4 uses a genetic algorithm to optimize the power dispatch plan to obtain the optimal dispatch strategy. Specifically, the power dispatch problem is first expressed as an optimization problem, and a fitness function is defined; then a group of candidate dispatch plans are initialized as the initial population of the genetic algorithm; then, crossover and mutation operations are performed on the candidate plans in the population to generate new candidate plans, and selection is performed according to the fitness function to retain plans with higher fitness; finally, crossover, mutation and selection operations are iteratively performed until the stop condition is met, and the optimal dispatch strategy is output.

6. The method for dispatching electric power according to electric power load forecasting according to claim 1, It is characterized in that The scheduling execution and feedback step in S5 includes dynamically adjusting the start and stop schedule of the generator set to cope with the fluctuation of the power load, specifically including: Real-time monitoring of load changes: Through sensors and monitoring systems, the actual change data of power load is obtained in real time, compared with the predicted results, and the load deviation is calculated. The load deviation is set as ΔP(t). The specific calculation formula is: ΔP(t) = P 实际 (t)-P 预测 (t), where P 实际 (t) is the actual load, P 预测 (t) is the predicted load; Start-stop adjustment strategy: adjust the start-stop arrangement of the generator set according to the load deviation. If ΔP(t)>0, increase the output power of the generator set or enable the standby generator set; if ΔP(t)<0, reduce the output power of the generator set or disable some generator sets; Power change during the adjustment process: the power adjustment of the generator set ΔQ(t), the specific calculation formula is: ΔQ(t)=K·ΔP(t), where K is the adjustment coefficient, which is determined according to the adjustment capability of the generator set and the system requirements; Dynamic adjustment strategy: During the adjustment process, the operating status and load changes of the generator sets are monitored in real time, and the start-stop schedule and power output are continuously adjusted to ensure the stable operation of the power system; Feedback data collection: Collect the adjusted generator set operation data and power load data as input data for model self-learning and updating, and optimize the prediction model and scheduling strategy.

7. The method for dispatching electric power according to electric power load forecasting according to claim 1, It is characterized in that The S6 self-learning and model updating specifically utilizes a reinforcement learning algorithm to continuously adjust model parameters based on historical dispatch data and feedback data. Specifically, the power dispatch process is first regarded as a reinforcement learning problem, and the state space, action space, and reward function are defined; then, the reinforcement learning algorithm is trained using historical dispatch data to adjust the parameters of the load forecasting model and the dispatch strategy generation model; in the actual dispatch process, the model parameters are updated in real time according to the dispatch execution results and feedback data to improve the model's adaptability.

8. The method for dispatching electric power according to electric power load forecasting according to claim 1, It is characterized in that The power dispatch strategy generation model includes a constraint generation module that considers power grid safety constraints to ensure the safety of the dispatch strategy. Specifically, the power grid safety constraints are first defined, including thermal stability constraints of power lines, node voltage constraints, and system frequency constraints; then, when generating the dispatch strategy, it is checked whether each dispatch scheme meets the safety constraints, and the schemes that do not meet the requirements are modified or discarded; when optimizing the dispatch strategy, the schemes that meet the safety constraints are given priority.

9. The method for dispatching electric power according to electric power load forecasting according to claim 1, It is characterized in that The dispatch strategy generation also includes considering the price information of the electricity market and optimizing the economic efficiency of the dispatch strategy. Specifically, the real-time price data of the electricity market is first collected as one of the inputs for the generation of the dispatch strategy; then, when generating the dispatch strategy, the economic costs of different dispatch schemes are calculated by combining the load forecast results and the market price information; finally, the dispatch scheme with the lowest economic cost is selected to ensure that the power system minimizes the operating cost while meeting the load demand and safety constraints.

10. A power dispatching system based on power load forecasting, used to implement the power dispatching method based on power load forecasting as claimed in any one of claims 1 to 9, It is characterized in that Includes the following modules: Data collection and processing module: used to collect multi-source data from the power system, including historical load data, meteorological data, and user power consumption behavior data; And perform data cleaning, data normalization and data noise reduction processing steps to ensure the accuracy and completeness of the data; Load forecasting model module: receives pre-processed data from the data collection and processing module, and uses deep learning algorithms to build a power load forecasting model; Real-time load forecasting module: uses the model built by the load forecasting model module, combined with the power system data collected in real time from the data collection and processing module, to make real-time forecasts of future power loads and generate short-term load forecast results; Dispatch strategy generation module: Based on the prediction results provided by the real-time load prediction module, combined with the operation constraints of the power grid and the price information of the power market, it generates an optimized power dispatch strategy, including the start and stop arrangements of the generator sets, power adjustment instructions, and load transfer plans; Dispatch execution and feedback module: executes the power dispatch strategy formulated by the dispatch strategy generation module, including issuing the dispatch strategy to each generator set and load management unit, and monitoring the dispatch execution in real time; according to the actual execution status and feedback information, feedbacks the adjustment information to the dispatch strategy generation module to dynamically adjust the power dispatch strategy; Self-learning and model updating module: Based on the feedback data provided by the dispatch execution and feedback module, the power load forecasting model and dispatch strategy generation model are self-learned and updated.

Citation Information

Patent Citations

  • Power system load prediction method and system

    CN115293400A

  • Load ultra-short-term prediction method and system based on two-stage intelligent feature engineering

    CN116011655A

  • Power load scheduling method and system based on big data

    CN116646933A

  • Power grid planning dynamic method and system based on reinforcement learning and predictive analysis

    CN116805198A

  • Market clearing optimization pricing method and system and storage medium

    CN117455532A

Cited By

  • Power grid dispatching method and system

    CN119231519A

  • Medium-voltage distribution network capacity-load ratio calculation method based on deep neural network

    CN119419790A

  • A method for calculating the capacity-to-load ratio of medium-voltage distribution networks based on deep neural networks

    CN119419790B

  • Power system monitoring data analysis method and system based on big data

    CN119726691A

  • A power system monitoring data analysis method and system based on big data

    CN119726691B