Source network load storage AI intelligent scheduling method and system

Abstract

The invention relates to the technical field of source network load storage intelligent scheduling, and discloses a source network load storage AI intelligent scheduling method and system, and the system comprises a data collection module, a fluctuation analysis module, a constraint calculation module, a load modeling module, an energy storage analysis module, a decision engine module, and a safety correction module. Output data of a photovoltaic power station, a wind power plant and a traditional power plant are collected in real time through an Internet of Things sensor, and a multi-source heterogeneous energy data pool is constructed; aligning the corrected source load data with the energy storage state and the power grid operation parameters according to a time sequence to generate a four-dimensional optimization combination matrix; matching an optimal scheduling algorithm through a bionic search strategy, and analyzing the data matrix through the deep reinforcement learning model to generate a three-section scheduling instruction; and finally, the power grid control system executes power generation adjustment, load regulation and control and energy storage charging and discharging instructions. According to the system, the edge computing gateway is adopted to realize data acquisition, the new energy consumption capability and the power grid stability are improved, and the risk of source-grid load-storage collaborative failure is reduced.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology for power generation, grid, load, and storage, specifically to an AI-based intelligent scheduling method and system for power generation, grid, load, and storage. Background Technology

[0002] Intelligent dispatching technology for power generation, grid, load, and energy storage is the core control hub of the new power system. It integrates the four major elements of power source, grid, load, and energy storage through artificial intelligence and big data technology to achieve multi-dimensional dynamic balance and collaborative optimization. The AI-powered intelligent dispatching method and system for power generation, grid, load, and energy storage is an intelligent collaborative system that deeply integrates technologies from multiple fields. It aims to improve the dispatching efficiency and stability of energy production, transmission, consumption, and energy storage. This is the core support for achieving efficient operation of the energy internet. The system uses artificial intelligence algorithms to perform real-time analysis of multi-dimensional data on power generation, grid, load, and energy storage to ensure the rapid generation of optimal dispatching strategies under complex operating conditions.

[0003] Currently, due to the multi-source heterogeneous nature of the energy system, traditional scheduling models rely on static parameters and human experience when conducting coordinated scheduling of energy sources, grids, loads, and storage. They cannot perceive fluctuations in energy supply output and sudden changes in load demand in real time. If there are short-term and drastic fluctuations in new energy power generation, it may cause a mismatch between scheduling instructions and the actual supply and demand status, leading to the failure of coordinated scheduling of energy sources, grids, loads, and storage.

[0004] Therefore, a source-grid-load-storage AI-based intelligent scheduling method and system are proposed to solve the above problems. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an AI-based intelligent scheduling method and system for source-grid-load-storage systems, solving the problems mentioned in the background section.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a source-grid-load-storage AI intelligent scheduling method and system, wherein the method includes the following steps: S1. Collect source, grid, load and storage parameter data, including energy source output data, grid operation status data, load demand data and energy storage system status data; S2. Based on the energy source output data and the load demand data, perform supply and demand forecasting processing to generate source output forecasting data and load demand forecasting data; S3. Based on the source output forecast data and the load demand forecast data, perform supply and demand balance consistency judgment processing to generate a supply and demand balance consistency judgment flag; when supply and demand are balanced, directly execute step S5. S4. When supply and demand are unbalanced, data correction and adjustment are performed based on the source output forecast data and the load demand forecast data to generate corrected source output forecast data and corrected load demand forecast data. S5. Based on the source-end power output prediction data and the corrected source-end power output prediction data, the load demand prediction data and the corrected load demand prediction data, the power grid operation status data and the energy storage system status data, construct source-grid-load-storage optimized combination data; S6. Based on the source-grid-load-storage optimization combination data and the standard source-grid-load-storage optimization combination data corresponding to different AI scheduling algorithms, perform AI scheduling algorithm type matching processing to generate target AI scheduling algorithm type feature data; S7. Construct a summary data of dispatch instructions, perform source-grid-load-storage dispatch execution processing, generate real-time dispatch instruction data, and output it to the power grid control system.

[0007] Preferably, step S1 includes the following steps: S11. Collect power output data from energy sources through IoT sensors. The energy sources include photovoltaic power stations, wind power stations, and traditional power plants. The power output data from the energy sources includes real-time power generation and power generation prediction error rate. Power grid operation status data is collected through power grid monitoring equipment. The power grid operation status data includes voltage fluctuation parameters, frequency stability parameters, and line load rate. Load demand data is collected through smart meters, including real-time power consumption and load forecast curves. The battery management system collects energy storage system status data, which includes battery remaining capacity, charge / discharge efficiency, and health status index.

[0008] Preferably, step S2 includes the following steps: S21. Import the energy source output data and the load demand data into the AI ​​scheduling platform; S22. Use a time series prediction model to perform future time period prediction processing on the energy source output data to generate source output prediction data. The prediction model includes an LSTM neural network. S23. Use clustering algorithms to perform demand pattern recognition and prediction processing on the load demand data to generate load demand prediction data. The clustering algorithms include K-means.

[0009] Preferably, step S3 includes the following steps: S31. Obtain the source-end output prediction data and the load demand prediction data; S32. Compare the source-end output forecast data with the load demand forecast data, and generate a supply-demand balance consistency judgment flag P based on the comparison result: When the predicted output value is greater than or equal to the predicted load demand value, the output P indicates a supply-demand balance. When the predicted output value at the source is less than the predicted load demand value, the output P indicates a supply-demand imbalance.

[0010] Preferably, step S4 includes the following steps: S41. When P is a supply-demand imbalance, the Kalman filter algorithm is used to correct and adjust the source-end power output prediction data to generate corrected source-end power output prediction data. S42. The load demand forecast data is smoothed and corrected using a sliding window averaging algorithm to generate corrected load demand forecast data.

[0011] Preferably, step S5 includes the following steps: S51. The source-end power output prediction data and the corrected source-end power output prediction data, the load demand prediction data and the corrected load demand prediction data, the grid operation status data, and the energy storage system status data are aligned and combined according to time series parameters to construct a source-grid-load-storage optimized combination data matrix. : .

[0012] Preferably, step S6 includes the following steps: S61. Establish a data matrix for the optimal combination of standard source-grid-load-storage systems corresponding to different AI scheduling algorithms.

[0013] in, The h-th AI scheduling algorithm type corresponds to the standard optimized combination data, where h = 1, 2, 3, 4, 5, and the algorithm types include genetic algorithm, particle swarm optimization, deep reinforcement learning, simulated annealing, and linear programming. S62, the above and In Perform similarity matching and search for matching results. Generate feature data for the target AI scheduling algorithm type, corresponding to the algorithm type. This includes the following sub-steps: S621. Initialize the maximum number of iterations T, and randomly initialize the position of the population search algorithm in the optimization space. The position formula is:

[0014] in, For the position, Let be the lower boundary. For the upper boundary, The result is a random number in the range [0,1]. S622, Exploration Phase: Position is updated based on particle swarm behavior simulation. Position update formula:

[0015] in, For the target location, The value is a constant; if the new position is better, the original position is replaced. S623, Development Phase: Calculate new random positions, with the numerator containing positional dimensions and the denominator being dimensionless.

[0016] Normalized boundary parameters, dimensionless:

[0017] Dimensional normalization constraints must satisfy:

[0018] in, Let represent the position of the i-th search entity in the ϑ-dimensional space, with the dimension of the original data. This is the time decay coefficient, a dimensionless threshold, ranging from 0.1 to... ≤1.0, , These are the normalized values ​​of the boundary parameters. A random number in the interval [0,1] with a dimension of 1. The current iteration number is dimensionless. Original position parameters ,η, Let η be the lower boundary of the search space and η be the upper boundary of the search space. For data matrix The largest eigenvalue; S624, Output the matching result after iterating to the maximum number of times. ; S625, Matching-based Generate target AI scheduling algorithm type feature data .

[0019] Preferably, step S7 includes the following steps: S71, the above and stated Combine and construct scheduling instruction summary data ; S72, the AI ​​scheduling platform calls the aforementioned Corresponding algorithm program processing ,extract The data from the power grid and energy storage systems are used to generate real-time dispatch command data. These include power generation adjustment commands, load reduction commands, and energy storage charging and discharging commands.

[0020] Preferably, after step S7, the method further includes: S8. Based on the real-time scheduling instruction data, perform execution effect feedback processing, use reinforcement learning algorithm to update the prediction model parameters, and generate an optimized prediction model.

[0021] Preferably, the following steps are also included: S11. Perform data desensitization processing on the collected source-grid-load-storage parameter data to generate desensitized source-grid-load-storage parameter data; S12. Perform privacy compliance verification processing based on the de-identified source-network-load-storage parameter data to generate a privacy compliance verification flag; when the verification passes, execute step S2. S13. When the verification fails, perform data anonymization reprocessing, generate anonymized source-network-load-storage parameter data, and re-execute step S12.

[0022] (III) Beneficial Effects Compared with existing technologies, this invention provides an AI-powered intelligent scheduling method and system for source-grid-load-storage systems, which has the following advantages: 1. In this invention, by setting up a data correction unit, when performing source-grid-load-storage coordinated scheduling, the source output and load demand forecast data are corrected in real time to solve the supply-demand mismatch problem caused by the fluctuation of new energy output, ensure the consistency between scheduling instructions and actual energy status, further reduce the risk of source-grid-load-storage coordination failure, and improve system reliability.

[0023] 2. In this invention, by setting up an algorithm matching unit, the optimal scheduling strategy is dynamically matched through artificial intelligence algorithm during the load response scheduling process, which solves the problem of missing coupling assessment between load-side regulation capacity and grid transmission loss. This enables the system to reduce the energy efficiency attenuation of scheduling commands in the transmission link and optimize scheduling parameters in real time when the grid is congested, thus ensuring the overall operating efficiency of source, grid, load and storage.

[0024] 3. In this invention, by setting up an execution feedback unit, the prediction model is updated in real time through a multi-time-scale rolling correction mechanism during energy storage charging and discharging scheduling. This solves the problem of overcharging and discharging and idle energy storage caused by the spatiotemporal propagation of prediction errors, enabling the system to achieve adaptive optimization, reducing the possibility of economic and safety degradation, and further ensuring the long-term stability of source-grid-load-storage scheduling. Attached Figure Description

[0025] Figure 1This is a schematic diagram of the architecture of the AI-powered intelligent scheduling method and system for source-grid-load-storage of the present invention; Figure 2 This is a flowchart illustrating the steps of the AI-powered intelligent scheduling method and system for source-grid-load-storage of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Please see Figure 1-2 The source-grid-load-storage AI intelligent scheduling method and system includes the following steps: S1. Collect source, grid, load and storage parameter data, including energy source output data, grid operation status data, load demand data and energy storage system status data; S2. Based on the power output data of energy source end and the load demand data of intelligent dispatching of energy source, grid, load and storage, perform supply and demand forecasting to generate power output forecasting data and load demand forecasting data. S3. Based on the source-end output forecast data and the load demand forecast data of the intelligent dispatching of power generation, grid, load and storage, perform supply and demand balance consistency judgment processing to generate a supply and demand balance consistency judgment flag; when supply and demand are balanced, directly execute step S5. S4. When supply and demand are unbalanced, data correction and adjustment are performed based on the source output forecast data and load demand forecast data of the intelligent scheduling of source, grid, load and storage, to generate corrected source output forecast data and corrected load demand forecast data. S5. Based on the source-end output prediction data and the corrected source-end output prediction data of the intelligent dispatching of power generation, grid, load and storage, the load demand prediction data and the corrected load demand prediction data of the intelligent dispatching of power generation, grid, load and storage, the grid operation status data of the intelligent dispatching of power generation, grid, load and storage, and the energy storage system status data of the intelligent dispatching of power generation, grid, load and storage, construct the optimized combination data of power generation, grid, load and storage. S6. Based on the source-grid-load-storage intelligent scheduling source-grid-load-storage optimized combination data and the standard source-grid-load-storage optimized combination data corresponding to different AI scheduling algorithms, perform AI scheduling algorithm type matching processing to generate target AI scheduling algorithm type feature data; S7. Construct a summary data of dispatch instructions, process the source-grid-load-storage dispatch execution, generate real-time dispatch instruction data, and output it to the power grid control system. S11. Collect power output data from energy sources through IoT sensors. The energy sources for intelligent dispatching of power generation, grid, load and storage include photovoltaic power stations, wind power stations and traditional power plants. The power output data from energy sources for intelligent dispatching of power generation, grid, load and storage includes real-time power generation and power generation prediction error rate. Power grid operation status data is collected through power grid monitoring equipment. The power grid operation status data for intelligent dispatching of power sources, grids, loads and storage includes voltage fluctuation parameters, frequency stability parameters and line load rate. Load demand data is collected through smart meters, and the load demand data for intelligent dispatching of power sources, grids, loads, and storage includes real-time power consumption and load forecast curves. The battery management system collects energy storage system status data, which includes battery remaining capacity, charge and discharge efficiency, and health status index. S21. Import the energy source output data and load demand data of the intelligent dispatching of energy sources, grids, loads and storage into the AI ​​dispatching platform. S22. Use a time series prediction model to process the power output data of the energy source end of the intelligent dispatching of energy source, grid, load and storage for future periods to generate power output prediction data. The prediction model includes an LSTM neural network. S23. Clustering algorithms are used to identify and predict the demand patterns of the intelligent scheduling load demand data of source-grid-load-storage system, generating load demand prediction data. Clustering algorithms include K-means. S31. Obtain source-end output forecast data and source-grid-load-storage intelligent dispatching load demand forecast data. S32. Compare the source-end output forecast data of the intelligent dispatching system for power generation, grid, load, and storage with the load demand forecast data of the intelligent dispatching system for power generation, grid, load, and storage, and generate a supply-demand balance consistency judgment flag P based on the comparison results. When the predicted output value is greater than or equal to the predicted load demand value, the output P indicates a supply-demand balance. When the predicted output value is less than the predicted load demand value, the output P indicates a supply-demand imbalance. S41. When the intelligent dispatch of source, grid, load and storage is unbalanced in supply and demand, the Kalman filter algorithm is used to correct and adjust the source output prediction data of the intelligent dispatch of source, grid, load and storage to generate corrected source output prediction data. S42. The sliding window averaging algorithm is used to smooth and correct the load demand forecast data of intelligent dispatching of source, grid, load and storage, and generate corrected load demand forecast data. S51. Combine the source-end output forecast data and the corrected source-end output forecast data, the load demand forecast data and the corrected load demand forecast data, the grid operation status data, and the energy storage system status data of the intelligent dispatching system for power generation, grid, load, and storage, according to time series parameters, to construct an optimized combination data matrix for power generation, grid, load, and storage. : ; S61. Establish a data matrix for the optimal combination of standard source-grid-load-storage systems corresponding to different AI scheduling algorithms.

[0028] in, The h-th AI scheduling algorithm type corresponds to the standard optimized combination data, where h = 1, 2, 3, 4, 5, and the algorithm types include genetic algorithm, particle swarm optimization, deep reinforcement learning, simulated annealing, and linear programming. S62, Intelligent Dispatch of Power Grid Load Storage and In Perform similarity matching and search for matching results. Generate feature data for the target AI scheduling algorithm type, corresponding to the algorithm type. This includes the following sub-steps: S621. Initialize the maximum number of iterations T, and randomly initialize the position of the population search algorithm in the optimization space. The position formula is:

[0029] in, For the position, Let be the lower boundary. For the upper boundary, The result is a random number in the range [0,1]. S622, Exploration Phase: Position is updated based on particle swarm behavior simulation. Position update formula:

[0030] in, For the target location, The value is a constant; if the new position is better, the original position is replaced. S623, Development Phase: Calculate new random positions, with the numerator containing positional dimensions and the denominator being dimensionless.

[0031] Normalized boundary parameters, dimensionless:

[0032] Dimensional normalization constraints must satisfy:

[0033] in, Let represent the position of the i-th search entity in the ϑ-dimensional space, with the dimension of the original data. This is the time decay coefficient, a dimensionless threshold, ranging from 0.1 to... ≤1.0, , These are the normalized values ​​of the boundary parameters. A random number in the interval [0,1] with a dimension of 1. The current iteration number is dimensionless. Original position parameters ,η, Let η be the lower boundary of the search space and η be the upper boundary of the search space. For data matrix The largest eigenvalue; S624, Output the matching result after iterating to the maximum number of times. ; S625, Matching-based Generate target AI scheduling algorithm type feature data ; S71, Intelligent Dispatch of Power Grid Load Storage Heyuan Power Grid Intelligent Dispatch Combine and construct scheduling instruction summary data ; S72, AI scheduling platform calls source-grid-load-storage intelligent scheduling Corresponding algorithm program processing ,extract The data from the power grid and energy storage systems are used to generate real-time dispatch command data. These include power generation adjustment commands, load reduction commands, and energy storage charging and discharging commands; S8. Based on the real-time scheduling instruction data of source-grid-load-storage intelligent scheduling, the execution effect feedback processing is performed, and the prediction model parameters are updated by reinforcement learning algorithm to generate an optimized prediction model. S11. Perform data anonymization processing on the source-grid-load-storage parameter data collected by the intelligent scheduling of source-grid-load-storage to generate anonymized source-grid-load-storage parameter data. S12. Perform privacy compliance verification on the de-identified source-grid-load-storage parameter data based on the intelligent scheduling of source-grid-load-storage, and generate a privacy compliance verification flag; when the verification passes, execute step S2. S13. When the verification fails, perform data anonymization reprocessing, generate anonymized source-network-load-storage parameter data, and re-execute step S12.

[0034] Example 1: Data Acquisition of Source-Grid-Load-Storage Parameters In practice, IoT sensors are first deployed at photovoltaic and wind power plants to monitor power generation and prediction error rate in real time. Voltage fluctuation parameters and frequency stability data are collected through grid monitoring equipment. Smart meters are used to obtain the real-time power consumption curve of the industrial area load. At the same time, the battery management system records the remaining capacity and health status of the energy storage batteries. All data is transmitted to the AI ​​scheduling platform through the 5G network to ensure the synchronization and integrity of data collection.

[0035] Example 2: Supply and Demand Forecasting Processing The collected source-end power output and load data are imported into an LSTM neural network model. The model is trained based on historical 24-hour data to generate a power output prediction curve for the next 6 hours. On the load side, the K-means clustering algorithm is used to identify the differences in electricity consumption patterns between weekdays and holidays, and output time-of-day load demand prediction results. The prediction process fully considers the impact of sudden weather changes on new energy power generation and dynamically adjusts the prediction weight coefficients.

[0036] Example 3: Supply and Demand Balance Judgment and Correction Compare the source power output forecast with the load demand forecast: When the wind power output drops sharply and the forecast is lower than the load demand, the Kalman filter algorithm is triggered to correct the source data, eliminate short-term fluctuation noise, and the load data is smoothed by the sliding window algorithm to eliminate abnormal power consumption peaks. The corrected data is then rebalanced until the grid carrying capacity threshold is met.

[0037] Example 4: Optimized Combined Data Construction The corrected source data, load data, real-time grid status, energy storage status, and charging / discharging efficiency are aligned with timestamps to construct a four-dimensional combined data matrix. Each column in the matrix represents the complete system status at 5-minute intervals, and the row vectors contain normalized values ​​of four types of parameters: source, grid, load, and storage, forming a spatiotemporally correlated data cube.

[0038] Example 5: AI Scheduling Algorithm Matching The algorithm library pre-stores standard optimization combination data of five algorithms, including genetic algorithm and particle swarm optimization. A biomimetic search strategy is adopted: the particle swarm is initialized and randomly distributed in the algorithm space. During the exploration phase, the particle swarm is simulated to migrate towards the optimal solution. During the development phase, the search step size is reduced according to the number of iterations. Finally, the deep reinforcement learning algorithm with the highest matching degree with the current combination data is locked, and an algorithm type identification code is generated.

[0039] Example 6: Execution of Scheduling Instructions The combined data and algorithm identifier are packaged and sent to the scheduling engine. The engine uses a deep reinforcement learning algorithm to analyze grid congestion points and generate a three-stage instruction: a power generation increase instruction to thermal power plants, an air conditioning temperature control request to commercial loads, and an instruction for the energy storage system to release stored electricity within 15 minutes. All instructions are issued and executed via a dedicated power communication protocol.

[0040] Example 7: Dynamic Feedback Optimization After the dispatch command is executed, the grid frequency recovery rate and energy storage SOC change data are collected and input into the reinforcement learning model. The model automatically adjusts the hidden layer node weights of the LSTM prediction network and optimizes the boundary constraint parameters of the algorithm matching link to reduce the subsequent prediction error rate. The system completes a closed-loop self-update every 6 hours.

[0041] Example 8: System Hardware Deployment The data acquisition and processing module adopts a distributed edge computing gateway and is deployed in new energy power plants and substations. The scheduling optimization and analysis module runs on a cloud platform GPU cluster and is configured with a dedicated algorithm acceleration card. The scheduling execution feedback module is embedded in the power grid control system and interacts with the SCADA system through the OPC protocol. The modules use an industrial-grade time synchronization protocol to ensure timing consistency.

[0042] It should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in a process, method, article, or apparatus that includes the source-grid-load-storage intelligent scheduling element.

[0043] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A source-grid-load-storage AI intelligent dispatching system, characterized by: include: Data acquisition and processing module, including: The source-end parameter acquisition unit collects power output data from the energy source through IoT sensors; The load parameter acquisition unit collects load demand data through smart meters; The power grid parameter acquisition unit collects power grid operating status data through power grid monitoring equipment; The energy storage parameter acquisition unit collects energy storage system status data through the battery management system. The supply and demand forecasting unit generates forecast data based on source and load data. The data correction unit performs a supply and demand balance assessment on the corrected data. The scheduling optimization analysis module includes: Optimize the combined data generation unit to construct optimized combined data of source, grid, load and storage; Standard algorithm data storage unit, storing standard optimized combination data of different AI scheduling algorithms; The algorithm matching unit generates target AI scheduling algorithm type data based on similarity matching; The scheduling execution feedback module includes: The scheduling instruction generation unit generates and outputs real-time scheduling instruction data. The execution feedback unit updates the prediction model based on the scheduling effect feedback.

2. A source-grid-load-storage AI intelligent scheduling method, characterized by: The method includes the following steps: S1. Collect source, grid, load and storage parameter data, including energy source output data, grid operation status data, load demand data and energy storage system status data; S2. Based on the energy source output data and the load demand data, perform supply and demand forecasting processing to generate source output forecasting data and load demand forecasting data; S3. Based on the source output forecast data and the load demand forecast data, perform supply and demand balance consistency judgment processing to generate a supply and demand balance consistency judgment flag; when supply and demand are balanced, directly execute step S5. S4. When supply and demand are unbalanced, data correction and adjustment are performed based on the source output forecast data and the load demand forecast data to generate corrected source output forecast data and corrected load demand forecast data. S5. Based on the source-end power output prediction data and the corrected source-end power output prediction data, the load demand prediction data and the corrected load demand prediction data, the power grid operation status data and the energy storage system status data, construct source-grid-load-storage optimized combination data; S6. Based on the source-grid-load-storage optimization combination data and the standard source-grid-load-storage optimization combination data corresponding to different AI scheduling algorithms, perform AI scheduling algorithm type matching processing to generate target AI scheduling algorithm type feature data; S7. Construct a summary data of dispatch instructions, perform source-grid-load-storage dispatch execution processing, generate real-time dispatch instruction data, and output it to the power grid control system.

3. The AI-powered intelligent scheduling method for source-grid-load-storage as described in claim 2, characterized in that: S1 includes the following steps: S11. Collect power output data from energy sources through IoT sensors. The energy sources include photovoltaic power stations, wind power stations, and traditional power plants. The power output data from the energy sources includes real-time power generation and power generation prediction error rate. Power grid operation status data is collected through power grid monitoring equipment. The power grid operation status data includes voltage fluctuation parameters, frequency stability parameters, and line load rate. Load demand data is collected through smart meters, including real-time power consumption and load forecast curves. The battery management system collects energy storage system status data, which includes battery remaining capacity, charge / discharge efficiency, and health status index.

4. The AI-powered intelligent scheduling method for source-grid-load-storage as described in claim 2, characterized in that: S2 includes the following steps: S21. Import the energy source output data and the load demand data into the AI ​​scheduling platform; S22. Use a time series prediction model to perform future time period prediction processing on the energy source output data to generate source output prediction data. The prediction model includes an LSTM neural network. S23. Use clustering algorithms to perform demand pattern recognition and prediction processing on the load demand data to generate load demand prediction data. The clustering algorithms include K-means.

5. The AI-powered intelligent scheduling method for source-grid-load-storage as described in claim 2, characterized in that: S3 includes the following steps: S31. Obtain the source-end output prediction data and the load demand prediction data; S32. Compare the source-end output forecast data with the load demand forecast data, and generate a supply-demand balance consistency judgment flag P based on the comparison result: When the predicted output value is greater than or equal to the predicted load demand value, the output P indicates a supply-demand balance. When the predicted output value at the source is less than the predicted load demand value, the output P indicates a supply-demand imbalance.

6. The AI-powered intelligent scheduling method for source-grid-load-storage as described in claim 2, characterized in that: S4 includes the following steps: S41. When P is a supply-demand imbalance, the Kalman filter algorithm is used to correct and adjust the source-end power output prediction data to generate corrected source-end power output prediction data. S42. The load demand forecast data is smoothed and corrected using a sliding window averaging algorithm to generate corrected load demand forecast data.

7. The AI-powered intelligent scheduling method for source-grid-load-storage according to claim 2, characterized in that: S5 includes the following steps: S51. The source-end power output prediction data and the corrected source-end power output prediction data, the load demand prediction data and the corrected load demand prediction data, the grid operation status data, and the energy storage system status data are aligned and combined according to time series parameters to construct a source-grid-load-storage optimized combination data matrix. : 。 8. The AI-powered intelligent scheduling method for source-grid-load-storage according to claim 2, characterized in that: S6 includes the following steps: S61. Establish a data matrix for the optimal combination of standard source-grid-load-storage systems corresponding to different AI scheduling algorithms. in, The h-th AI scheduling algorithm type corresponds to the standard optimized combination data, where h = 1, 2, 3, 4, 5, and the algorithm types include genetic algorithm, particle swarm optimization, deep reinforcement learning, simulated annealing, and linear programming. S62, the above and In Perform similarity matching and search for matching results. Generate feature data for the target AI scheduling algorithm type, corresponding to the algorithm type. This includes the following sub-steps: S621. Initialize the maximum number of iterations T, and randomly initialize the position of the population search algorithm in the optimization space. The position formula is: in, For the position, Let be the lower boundary. For the upper boundary, The result is a random number in the range [0,1]. S622, Exploration Phase: Position is updated based on particle swarm behavior simulation. Position update formula: in, For the target location, The value is a constant; if the new position is better, the original position is replaced. S623, Development Phase: Calculate new random positions, with the numerator containing positional dimensions and the denominator being dimensionless. Normalized boundary parameters, dimensionless: Dimensional normalization constraints must satisfy: in, Let represent the position of the i-th search entity in the ϑ-dimensional space, with the dimension of the original data. This is the time decay coefficient, a dimensionless threshold, ranging from 0.1 to... ≤1.0, , These are the normalized values ​​of the boundary parameters. A random number in the interval [0,1] with a dimension of 1. The current iteration number is dimensionless. Original position parameters ,η, Let η be the lower boundary of the search space and η be the upper boundary of the search space. For data matrix The largest eigenvalue; S624, Output the matching result after iterating to the maximum number of times. ; S625, Matching-based Generate target AI scheduling algorithm type feature data .

9. The AI-powered intelligent scheduling method for source-grid-load-storage according to claim 2, characterized in that: S7 includes the following steps: S71, the above and stated Combine and construct scheduling instruction summary data ; S72, the AI ​​scheduling platform calls the aforementioned Corresponding algorithm program processing ,extract The data from the power grid and energy storage systems are used to generate real-time dispatch command data. These include power generation adjustment commands, load reduction commands, and energy storage charging and discharging commands.

10. The source-grid-load-storage AI intelligent scheduling method according to claim 2, characterized in that: Following S7, the following is also included: S8. Based on the real-time scheduling instruction data, perform execution effect feedback processing, use reinforcement learning algorithm to update the prediction model parameters, and generate an optimized prediction model.

Images