Optimization Method for Decision-Making Rules of New Energy AGC Intelligent Agent

Abstract

This invention discloses a method for optimizing decision-making rules for a new energy AGC (Automatic Generation Control) agent, relating to the field of power system testing technology. The method includes: collecting historical power grid fault data, new energy AGC control data, and simulation test data; preprocessing the collected data to obtain preprocessed data; and optimizing and training the agent's decision-making rule model using a supervised learning method based on the preprocessed data. Specifically, a deep neural network is used to predict the optimal AGC unit power adjustment, and the mean square error is used as the loss function for model optimization to obtain the optimized agent decision-making rule model. This method improves the model's prediction accuracy and the power grid's operating efficiency by optimizing the agent's decision-making rules.

Description

Technical Field

[0001] This invention relates to the field of power system testing technology, and in particular to a method for optimizing decision-making rules of a new energy AGC intelligent agent. Background Technology

[0002] With the increasing global demand for clean energy, the proportion of new energy sources in the power system is rising. Automatic generation control (AGC) for new energy sources, as a key technology to ensure reliable grid connection of new energy sources and maintain stable power system operation, is playing an increasingly important role. In practical applications, new energy AGC has been widely used in various new energy power plants, such as wind farms and photovoltaic power plants. Through the AGC system, new energy power plants can adjust their power generation in real time according to the grid's demand to maintain the power balance and frequency stability of the power system.

[0003] However, current methods for simulating the response of renewable energy AGC systems to complex grid faults have limitations, making it difficult to accurately and comprehensively simulate various potential grid faults and operating conditions. Therefore, a new approach is needed to improve the accuracy and efficiency of simulation testing for renewable energy AGC systems, while optimizing the decision-making rules of the agent to better adapt to the uncertainties of grid operation and the intermittency of renewable energy generation. This will help improve the grid's adaptability to fluctuations in renewable energy generation, enhancing grid stability and reliability. Summary of the Invention

[0004] The main objective of this invention is to provide a method for optimizing decision-making rules for new energy AGC agents, which can improve the accuracy of evaluating the control effect of new energy AGCs under complex fault scenarios.

[0005] To achieve the above objectives, the first aspect of this application provides a method for optimizing decision-making rules for new energy AGC intelligent agents, the method comprising:

[0006] Collect historical power grid fault data, new energy AGC control data, and simulation test data;

[0007] The collected data is preprocessed to obtain preprocessed data;

[0008] Based on the preprocessed data, a supervised learning method is used to optimize and train the decision rule model of the agent. In this method, a deep neural network is used to predict the optimal AGC unit power adjustment amount, and the mean square error is used as the loss function to optimize the model, thereby obtaining the optimized agent decision rule model.

[0009] Optionally, the method further includes:

[0010] Reinforcement learning is employed to further optimize the agent's adaptive decision-making ability. Specifically, by setting the state space, action space, and reward function, the deep Q-learning algorithm is used to enable the agent to iteratively optimize the decision-making strategy through Q-value optimization.

[0011] Optionally, the state space is set as follows:

[0012]

[0013] in, For power grid frequency deviation, For excess power, For changes in renewable energy power, V bus For critical node voltages;

[0014] The action space is set as follows:

[0015]

[0016] in, This refers to the active power regulation of the AGC unit. This is the reactive power regulation quantity;

[0017] The reward function is set as follows:

[0018]

[0019] in: α, β, γ These are the weighting coefficients; This is the rated voltage.

[0020] Optionally, the decision-making strategy optimized through Q-value iteration adopts the following formula:

[0021] in: m The learning rate; q Discount factor; Q ( s,a ) represents the Q-value of the current state-action pair; Q t+1 ( s , a The new Q value is the result of the iteration. r For instant rewards.

[0022] Optionally, the historical power grid fault data includes power flow exceeding limits at transmission sections, power grid frequency deviation, and voltage fluctuation.

[0023] The new energy AGC control data includes wind power and photovoltaic output data, AGC unit response data, and load changes;

[0024] The simulation data includes data related to the interactive behavior of new energy intelligent agents, AGC unit intelligent agents, and power transmission section intelligent agents under different fault scenarios.

[0025] Optionally, the preprocessing includes:

[0026] Normalization, outlier removal, and time series resampling.

[0027] Optionally, the method further includes:

[0028] The optimized agent decision rule model is used as the decision rule for the agent in the simulation model. Based on the simulation model, a joint simulation test of active power control of new energy AGC is carried out to obtain simulation results.

[0029] The simulation results are compared and verified with actual power grid monitoring data or theoretical analysis results; if there are deviations, the simulation model and algorithm are calibrated.

[0030] The second aspect of this application provides a new energy AGC intelligent agent decision-making rule optimization device, comprising:

[0031] The data acquisition module is used to collect historical power grid fault data, new energy AGC control data, and simulation test data;

[0032] The preprocessing module is used to preprocess the collected data to obtain preprocessed data;

[0033] The training optimization module is used to optimize and train the decision rule model of the agent based on the preprocessed data using a supervised learning method. Specifically, a deep neural network is used to predict the optimal AGC unit power adjustment amount, and the mean square error is used as the loss function to optimize the model and obtain the optimized agent decision rule model.

[0034] A third aspect of this application provides an electronic device including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform steps as described in the first aspect and any possible implementation thereof.

[0035] This application provides a method for optimizing the decision-making rules of a new energy AGC (Automatic Generation Control) agent. The method involves collecting historical power grid fault data, new energy AGC control data, and simulation test data; preprocessing the collected data to obtain preprocessed data; and then using supervised learning to optimize and train the agent's decision-making rule model based on the preprocessed data. Specifically, a deep neural network is used to predict the optimal AGC unit power adjustment, and the mean square error is used as the loss function for model optimization, resulting in an optimized agent decision-making rule model. This method optimizes the agent's decision-making rules, making the simulation results closer to reality, thereby improving the accuracy of evaluating the control effect of new energy AGC in complex fault scenarios. It not only improves the model's prediction accuracy and the power grid's operating efficiency but also enhances the power grid's adaptability to fluctuations in new energy power generation, thus providing strong technical support for the stable operation of the power grid. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] in:

[0038] Figure 1 A flowchart illustrating a new energy AGC intelligent agent decision rule optimization method provided in an embodiment of this application;

[0039] Figure 2 A schematic diagram of the structure of a new energy AGC intelligent agent decision rule optimization device provided in the embodiments of this application;

[0040] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0041] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0042] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0043] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0044] The following explains some of the terms, concepts, or related background information used in the embodiments of this application:

[0045] Automatic Generation Control (AGC): Within a defined area, when the power system frequency or tie-line power changes, the active power of generator sets is automatically adjusted through a control program to maintain the system frequency or ensure predetermined power exchange between areas. Its technical equipment system mainly includes grid operation control systems of dispatching agencies at all levels, remote transmission channels, remote terminal equipment or computer monitoring systems at power plants, generator set coordination control systems, generator sets and their active power regulation devices, and application software to implement AGC functions. For regulation and control involving renewable energy AGC, under normal circumstances, renewable energy power plants generate electricity at maximum capacity. When the grid experiences issues such as exceeding capacity limits, difficulties in peak shaving, insufficient reserves, or frequency fluctuations, the renewable energy AGC automatically adjusts the output power of renewable energy power generation equipment (such as photovoltaic inverters) to meet the needs of grid safety and stability.

[0046] Power Regulator (PLC): Following instructions from the AGC system, the PLC precisely controls the output power of the renewable energy power generation equipment. It calculates the difference between the total generated power and the sum of the current generated power of each device as the power to be regulated. Then, it prioritizes each device based on its adjustable power ratio and allocates the regulated power to each device. This process ensures that the renewable energy power generation equipment operates efficiently and stably while meeting grid demands. In the renewable energy AGC system, the PLC can also integrate optimization algorithms and strategies, such as proportional allocation and priority ranking methods, to achieve more precise and efficient power regulation. These algorithms and strategies fully consider the characteristics of the renewable energy power generation equipment and grid demands, ensuring stable system operation even under complex and changing conditions.

[0047] The principle of PLC equivalence for regional dispatch centers: Provincial-level coordinated control should support equivalence of a single regional dispatch center into multiple virtual machines based on group objects, grid topology nodes, and cross-sectional partitions, so as to achieve refined control of photovoltaic / wind power plants and related cross-sections at the central dispatch center. The principles for PLC equivalence of regional dispatch centers' new energy power plants at the central dispatch center's main station are as follows:

[0048] 1) Geological survey stations whose output affects the same provincial-level coordination section can be considered as one PLC; other new energy stations not affected by the provincial-level coordination section can be considered as one PLC.

[0049] 2) New energy power stations connected to the same 220kV node can be equivalent to one PLC;

[0050] 3) Each prefecture-level new energy power station is connected to only one PLC.

[0051] 4. Provincial-Local Coordination: An effective coordination mechanism is established between provincial and prefecture-level power grid dispatching. Generally, the provincial dispatching center formulates overall control strategies and objectives based on the overall power supply and demand situation and the forecast results of renewable energy generation, and then issues them to the prefecture-level dispatching centers. The prefecture-level dispatching centers, in turn, formulate specific control plans based on the provincial dispatching center's control strategies and objectives, combined with the local renewable energy generation situation and power grid structure. This provincial-local coordinated control strategy optimizes the dispatching and control of renewable energy generation. This not only improves the safety and stability of the power grid but also enhances the utilization rate and power generation efficiency of renewable energy.

[0052] The embodiments of this application mentioned below are described in conjunction with the accompanying drawings.

[0053] Figure 1 This is a flowchart illustrating a new energy AGC intelligent agent decision rule optimization method provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes:

[0054] 101. Collect historical power grid fault data, new energy AGC control data, and simulation test data.

[0055] The method in this application embodiment can be implemented by a new energy AGC intelligent agent decision rule optimization device, which can be implemented on an electronic device in practical applications.

[0056] In multi-agent modeling methods, each agent's decision-making rules need to be adaptive to cope with the uncertainties of complex grid faults and fluctuations in renewable energy output. Therefore, in this embodiment, machine learning algorithms can be used to optimize and train the agents, enabling them to learn the optimal decision-making strategy based on historical grid fault data and renewable energy AGC control data. The training objective is to enable the agents to dynamically adjust their strategies based on existing experience and real-time data when facing different fault conditions, thereby improving the response speed, regulation accuracy, and grid stability of the renewable energy AGC system.

[0057] Specifically, the optimization training of intelligent agents includes data acquisition and model training.

[0058] The data obtained in this data acquisition step can be used as training sample data, mainly including historical power grid fault data, new energy AGC control data, and simulation test data.

[0059] Optionally, the aforementioned historical power grid fault data includes power flow exceeding limits at transmission sections, power grid frequency deviation, and voltage fluctuations.

[0060] The aforementioned new energy AGC control data includes wind power and photovoltaic output data, AGC unit response data, and load changes;

[0061] The simulation data mentioned above includes data related to the interactive behavior of new energy intelligent agents, AGC unit intelligent agents, and power transmission section intelligent agents under different fault scenarios.

[0062] Specifically, historical power grid fault data can mainly include power flow exceeding limits at transmission sections (fault time, section number, active power flow, reactive power flow, and exceeding limit magnitude), power grid frequency deviation (frequency changes before and after the fault), and voltage fluctuations (voltage changes at key nodes).

[0063] New energy AGC control data can mainly include wind power and photovoltaic output data (real-time power, wind speed, and solar intensity), AGC unit response data (adjustment rate, adjustment range, and control delay), and load change data (system load forecast and actual load data).

[0064] Simulation data can mainly include data related to the interactive behavior of new energy intelligent agents, AGC unit intelligent agents, and power transmission section intelligent agents under different fault scenarios.

[0065] 102. Preprocess the collected data to obtain preprocessed data.

[0066] Optionally, the above preprocessing includes:

[0067] Normalization, outlier removal, and time series resampling.

[0068] The preprocessing mentioned in this embodiment refers to a series of processing steps performed on the collected data before training the decision rule model of the agent using machine learning algorithms. The purpose of preprocessing is to improve data quality, making it more suitable for model training and improving the performance and accuracy of the model. Appropriate data preprocessing methods can be selected and adjusted as needed.

[0069] The main normalization method can be the Min-Max normalization method:

[0070]

[0071] in: X This is the original data; , These are the minimum and maximum values ​​of the data, respectively. X′ This is the normalized data.

[0072] 103. Based on the preprocessed data described above, a supervised learning method is used to optimize and train the decision rule model of the agent. In this process, a deep neural network is used to predict the optimal AGC unit power adjustment amount, and the mean square error is used as the loss function to optimize the model, thereby obtaining the optimized agent decision rule model.

[0073] Supervised learning, as mentioned in the embodiments of this application, is a method in machine learning that involves learning a model from labeled training data to predict the output of unseen data. In supervised learning, each training sample contains input data (features) and a corresponding output label (target value). The goal of the model is to learn the relationship between the input features and the output label so that it can make accurate predictions on new, unseen data.

[0074] For some tasks of AGC control in new energy sources (such as power regulation decisions of AGC units), supervised learning methods are used for training so that the AGC unit can learn to make appropriate power adjustment strategies under different fault conditions.

[0075] The deep neural network (DNN) mentioned in this application embodiment is a type of artificial neural network with multiple hidden layers, capable of learning complex patterns and relationships in data. By increasing the depth of the network (i.e., the number of hidden layers), deep neural networks can capture more abstract data features, thereby achieving superior performance in many tasks.

[0076] Specifically, in the embodiments of this application, a deep neural network (DNN) can be used, the goal of which is to predict the optimal power adjustment amount of the AGC unit:

[0077]

[0078] Where: Input X includes the power grid frequency deviation Changes in new energy output Trend exceeding limits etc.; output This refers to the power adjustment amount of the AGC unit.

[0079] The loss function for neural networks can be the mean squared error (MSE):

[0080]

[0081] in: The target output is the result of the training data; Adjusted power for model prediction; N The total number of samples.

[0082] In an optional implementation, the method further includes:

[0083] Reinforcement learning is employed to further optimize the agent's adaptive decision-making ability. Specifically, by setting the state space, action space, and reward function, the deep Q-learning algorithm is used to enable the agent to iteratively optimize the decision-making strategy through Q-value optimization.

[0084] Reinforcement learning (RL) mentioned in the embodiments of this application is a method in machine learning that learns how to make decisions through interaction with the environment. In reinforcement learning, an agent takes actions in a given environment and receives feedback (rewards) based on these actions, with the goal of maximizing cumulative rewards. This method is particularly useful in tasks that require decision sequences, such as games, robot control, and resource management.

[0085] When dealing with complex environments such as fluctuations in new energy sources and sudden failures, intelligent agents need to possess adaptive decision-making capabilities. Reinforcement learning can be used for optimization, enabling the agent to learn the optimal decision-making strategy through continuous trial and error. The core of reinforcement learning is that the agent performs actions in the environment and adjusts its strategy based on reward feedback.

[0086] In one implementation, the state, action, and reward can be set as follows:

[0087] State space S:

[0088]

[0089] in, For power grid frequency deviation, For excess power, Changes in renewable energy power refer to changes in the power generation capacity of renewable energy sources such as wind and solar energy. V bus This refers to the voltage at critical nodes.

[0090] Action Space A:

[0091]

[0092] in, This refers to the active power regulation of the AGC unit. This is the reactive power regulation quantity.

[0093] Reward function R:

[0094]

[0095] in: α, β, γ These are the weighting coefficients; This is the rated voltage.

[0096] Deep Q-Learning (DQN) mentioned in this embodiment is a technique combining deep learning and reinforcement learning. It selects the optimal action by estimating the value of each action. During training, the agent interacts with the environment, receives reward signals, and updates its estimates of action values ​​based on these signals. By repeatedly performing this process, the agent gradually learns how to make optimal decisions in different situations. The core idea of ​​Deep Q-Learning is to construct a deep neural network as the agent's decision-making model. The input of this neural network is the agent's observation of the environment, and the output is the value estimate of each action. During training, reinforcement learning algorithms are used to update the parameters of the neural network so that the agent can better estimate the value of actions.

[0097] Specifically, deep Q-learning can be used to train the agent, enabling it to iteratively optimize its decision-making strategy through Q-value optimization.

[0098]

[0099] in: m The learning rate; q Discount factor; Q ( s,a ) represents the Q-value of the current state-action pair; Q t+1 ( s , a The new Q value is the result of the iteration. r For instant rewards.

[0100] In the above formula s The "state space" representing the state of the intelligent agent, in the power grid simulation scenario of this application, refers to this space. s May include grid frequency deviation f Power flow exceeding limits at transmission sections P Changes in new energy output P N Critical node voltage V bus These are key pieces of information that comprehensively reflect the current operating status of the power grid; a The "action space" represents the actions that an intelligent agent can perform. In the context of a power grid scenario, this corresponds to... a This may include the active power regulation of the AGC unit. P adjust Reactive power regulation Q adjust This refers to the specific actions that an intelligent agent can take in response to the current state.

[0101] Q ( s , a () is the "state-action value function", which represents the agent's state in the current state. s Next, execute the action. a Then, the expected long-term cumulative rewards that can be obtained. Q ( s , a The higher the value, the more likely the action will be taken under the current power grid conditions. a This is more beneficial for maintaining grid stability and optimizing AGC control performance.

[0102] Q ( s' , a' This indicates when the agent enters a new state. s' Then, for all possible actions a'Value prediction. For example, when the power grid state changes due to frequency deviation or cross-sectional power flow. s' Then, the agent determines and performs an action. a' This refers to the benefits that can be obtained after AGC adjusts its power.

[0103] Q ( s' , a' States and actions are "potential". s' It is to perform an action a The state that will appear later a' Is s 'All possible actions that can be selected in this state.' max Q ( s' , a' ) is in state s' The maximum value of Q among all possible actions.

[0104] During the training process of deep Q-learning, the agent continuously interacts with the simulation environment and updates its algorithm using the formula described above. Q ( s , a The value of ), where r It is to perform an action a Immediate rewards after the action, such as a decrease in frequency deviation, will result in a positive reward; if the cross-sectional exceedance worsens, a negative reward will be given. s' It is to perform an action a The new state that is entered later, max Q ( s' , a' ) is in the new state s' The maximum value of Q among all possible actions. m It is the learning rate (which controls the magnitude of each update). q It's a discount factor (balancing the importance of immediate rewards and future rewards). Through such iterative updates, Q ( s , a It can gradually approach the optimal value, allowing the intelligent agent to continuously learn the optimal decision-making strategy in complex power grid fault scenarios. For example, when multiple faults overlap, it can accurately select the adjustment amount of the AGC unit to achieve stable control of the power grid.

[0105] The reference values ​​for each weight in the reward function R above can be determined by following a process of theoretical derivation, testing and verification, and continuous iteration. Specifically:

[0106] 1. Theoretical Derivation

[0107] (1) Determine the "weight of influence" of each indicator on the power grid

[0108] Based on the power grid's goals of "stabilizing frequency, controlling power flow, and ensuring voltage," key indicators can be initially prioritized. For example, frequency collapse directly leads to widespread power outages and requires priority control, thus warranting a higher weight. Power flow exceeding limits may cause equipment overload and line tripping, therefore it should be set as a secondary priority. Based on the deviation limits of each indicator, the theoretical weight can be initially set as follows: α : β : γ =0.5:0.2:0.1=5:2:1.

[0109] (2) Correction based on "control sensitivity"

[0110] Using small-disturbance stability analysis, calculate the adjustment sensitivity of AGC control to various indicators, such as the adjustment sensitivity of AGC to frequency. S f =0.8Hz / MW means that for every 1MW increase in output, the frequency rises by 0.8Hz. The adjustment sensitivity of each indicator is calculated and tested. Higher sensitivity indicates that the AGC can more easily control that indicator, allowing for a more appropriate reduction in its weight. The weights are then adjusted accordingly.

[0111] 2. Testing and Verification

[0112] After obtaining the theoretically calculated weights, different weight combinations can be set in simulation tests to observe the control effect of AGC, thereby judging the error in the weight setting and adjusting it in a timely manner. If the weights... α If the value is too small, the frequency exceeding the limit cannot be suppressed in time during the simulation, resulting in significant fault losses. Therefore, the corresponding weight should be increased appropriately.

[0113] Through multi-scenario testing and algorithmic optimization of weights, the expected failure loss is minimized, ultimately determining the project's usability. α , β , γ .

[0114] 3. Continuous iteration

[0115] The system should be recalibrated periodically as the grid topology and the proportion of new energy sources connected change. This could include annual updates or re-optimization when large-scale new energy power plants are added.

[0116] Further optionally, after step 103 above, the method further includes:

[0117] Using the optimized agent decision rule model as the decision rule for the agent in the simulation model, a joint simulation test of active power control of new energy AGC was conducted based on the simulation model to obtain simulation results.

[0118] The simulation results are compared and verified with actual power grid monitoring data or theoretical analysis results; if there are deviations, the simulation model and algorithm are calibrated.

[0119] Specifically, after obtaining the optimized agent decision-making rule model, it can be used in a simulation model to conduct joint simulation tests of active power control for new energy AGC. This simulation model can be obtained based on a multi-agent modeling method, treating transmission line overruns, the impact of natural conditions on new energy power generation equipment, and the new energy AGC system as independent agents. Each agent is assigned a specific function, initial state, and decision-making rules, enabling it to make real-time decisions and adjust its behavior based on its own state and information from other agents. This accurately simulates the dynamic interactions of various factors in complex scenarios, thereby improving test accuracy. The agent decision-making rules used in this simulation can be implemented using the optimized model in this application embodiment. By optimizing the agent decision-making rules, it can make more realistic decisions in different situations, further improving the accuracy and efficiency of the test.

[0120] In setting and simulating complex fault scenarios, it is necessary to consider various power grid devices and their interactions, including different factors such as transmission lines, wind farms, and photovoltaic power stations, while also introducing extreme natural condition changes and drastic fluctuations in grid frequency and voltage. Simulation enables dynamic simulation of these fault scenarios and allows for fault injection and scheduling decision optimization. When setting specific fault scenarios, the situation of transmission line overruns must first be considered. To simulate this fault, load fluctuations in the system are simulated by increasing or decreasing the load or adjusting the power output at one end of the transmission line, thereby triggering the phenomenon of transmission line overruns. In this case, the specific methods of increasing or decreasing the load can be defined using the following algorithm:

[0121]

[0122] in, This represents the load change. It is the load adjustment factor, which represents the proportion by which the load increases or decreases. This represents the load during normal operation. This adjustment simulates different load fluctuations and, by adjusting power output, simulates power overload or shortage fault scenarios. This process requires real-time monitoring of the status of each transmission section of the power grid and uses an intelligent agent model to decide on the range of load changes to ensure the system's dynamic response and stability.

[0123] In this embodiment, after simulating complex fault scenarios, the simulation results can be compared and verified with monitoring data of similar faults in the actual power grid or authoritative theoretical analysis results. If there is a large deviation, the interaction coefficients of various factors in the model and the parameters of the simulation algorithm can be adjusted to make the simulation results closer to the actual situation. At the same time, sensitivity analysis methods can be used to identify the factors and parameters that have a significant impact on the simulation results, and these key factors and parameters can be adjusted and calibrated first to improve calibration efficiency and thus improve the accuracy of the evaluation of the control effect of new energy AGC under complex fault scenarios.

[0124] Optionally, after adjusting and calibrating the parameters, the simulation test can be performed again, and the above steps can be repeated until the deviation between the simulation results and the actual data is within an acceptable range or the predetermined accuracy requirements are met.

[0125] In this embodiment, a data acquisition step is used to compare the simulation results with actual power grid monitoring data or theoretical analysis results to evaluate the control effect of the new energy AGC system under complex fault scenarios. Comparative analysis reveals the differences between the simulation model and the actual system, providing a basis for subsequent model adjustments and algorithm optimization. The main comparisons include the active power control response characteristics of the new energy AGC system, the regulation effect of power grid frequency and voltage, and the control of power flow at transmission sections. Verification of these factors ensures the accuracy of the simulation model and allows for adjustments when necessary to improve its consistency with actual power grid operation. Specifically, the comparative analysis may include:

[0126] When comparing simulation results with actual or theoretical analysis data, model calibration is necessary. The goal of model calibration is to make the simulation results more closely reflect reality, thereby improving the accuracy of evaluating the control effect of new energy AGC. During calibration, the interaction coefficients of various factors in the model can be adjusted, such as changing the influence weights between different faulty agents and the new energy AGC agent. By adjusting these weights, the interaction between agents can be altered, thus optimizing the accuracy of the simulation results. For example, if extreme wind speed changes at wind farms have a significant impact on grid frequency, the adjustment of the weights on the response of the new energy AGC system to wind speed changes can optimize the regulation effect.

[0127] Furthermore, parameters in the simulation algorithm can be modified, such as adjusting thresholds and step sizes in the agent's decision-making model. These parameters determine the agent's decision sensitivity and adjustment speed during the simulation process. For example, if the power regulation dead zone set in the agent's regulation rules is too large, the new energy AGC system may not respond sensitively enough to grid frequency fluctuations. In this case, the sensitivity to frequency fluctuations can be enhanced by reducing the dead zone range, thereby improving control accuracy.

[0128] These adjustments effectively reduce the discrepancy between simulation results and actual power grid operation data, improving model accuracy. For example, during simulation, if the power regulation of the new energy AGC system is not timely enough under extreme wind speed conditions, adjusting the step size and regulation rate in the agent's decision-making rules can make the regulation faster, avoiding excessive fluctuations in grid frequency. This approach not only improves control performance but also provides more accurate decision support for fault response in actual system operation. Ultimately, the calibrated simulation model can better simulate the power grid's response under complex fault scenarios, thereby optimizing the performance of the new energy AGC system and ensuring stable power grid operation.

[0129] The method in this embodiment collects historical power grid fault data, new energy AGC control data, and simulation test data; preprocesses the collected data to obtain preprocessed data; based on the preprocessed data, a supervised learning method is used to optimize and train the decision rule model of the agent. Specifically, a deep neural network is used to predict the optimal AGC unit power adjustment, and the mean square error is used as the loss function for model optimization to obtain the optimized agent decision rule model. This method optimizes the agent's decision rules, making the simulation results closer to reality, thereby improving the accuracy of evaluating the control effect of new energy AGC in complex fault scenarios. This not only improves the model's prediction accuracy and the power grid's operating efficiency but also enhances the power grid's adaptability to fluctuations in new energy power generation, thus providing strong technical support for the stable operation of the power grid.

[0130] Based on the description of the foregoing method embodiments, this application also provides a new energy AGC intelligent agent decision-making rule optimization device.

[0131] Figure 2 This is a schematic diagram of a new energy AGC intelligent agent decision rule optimization device provided in an embodiment of this application. Figure 2 As shown, the new energy AGC intelligent agent decision rule optimization device 200 includes:

[0132] Data acquisition module 210 is used to collect historical power grid fault data, new energy AGC control data and simulation test data;

[0133] The preprocessing module 220 is used to preprocess the collected data to obtain preprocessed data;

[0134] The training optimization module 230 is used to optimize and train the decision rule model of the agent based on the preprocessed data using a supervised learning method. Specifically, a deep neural network is used to predict the optimal AGC unit power adjustment amount, and the mean square error is used as the loss function to optimize the model and obtain the optimized agent decision rule model.

[0135] Optionally, the training optimization module 230 described above is further used to: further optimize the agent's adaptive decision-making ability using reinforcement learning, wherein, by setting the state space, action space and reward function, the agent uses a deep Q-learning algorithm to iteratively optimize the decision-making strategy through Q-values.

[0136] Optionally, the above state space is set as follows:

[0137]

[0138] in, For power grid frequency deviation, For excess power, For changes in renewable energy power, V bus For critical node voltages;

[0139] The above motion space is set as follows:

[0140]

[0141] in, This refers to the active power regulation of the AGC unit. This is the reactive power regulation quantity;

[0142] The above reward function is set as follows:

[0143]

[0144] in: α, β, γ These are the weighting coefficients; This is the rated voltage.

[0145] Optionally, the above-mentioned decision-making strategy optimized through Q-value iteration adopts the following formula:

[0146]

[0147] in: m The learning rate; q Discount factor; Q ( s,a ) represents the Q-value of the current state-action pair; Q t+1 ( s , a The new Q value is the result of the iteration. r For instant rewards.

[0148] Optionally, the aforementioned historical power grid fault data includes power flow exceeding limits at transmission sections, power grid frequency deviation, and voltage fluctuations.

[0149] The aforementioned new energy AGC control data includes wind power and photovoltaic output data, AGC unit response data, and load changes;

[0150] The simulation data mentioned above includes data related to the interactive behavior of new energy intelligent agents, AGC unit intelligent agents, and power transmission section intelligent agents under different fault scenarios.

[0151] Optionally, the above preprocessing includes:

[0152] Normalization, outlier removal, and time series resampling.

[0153] Optional, also includes:

[0154] The simulation execution module 240 is used to use the optimized agent decision rule model as the decision rule of the agent in the simulation model, and to perform joint simulation test of new energy AGC active power control based on the simulation model to obtain simulation results.

[0155] The verification and calibration module 250 is used to compare and verify the above simulation results with actual power grid monitoring data or theoretical analysis results; if there is a deviation, the above simulation model and algorithm are calibrated.

[0156] Understandably, this involves Figure 2 The relevant content of each module in the above method embodiments has been described in detail, and you can refer to the content of the method embodiments for details; that is... Figure 2 The provided new energy AGC intelligent agent decision-making rule optimization device 200 can perform, for example... Figure 1 Any steps in the illustrated embodiments will not be described in detail here.

[0157] In one embodiment of this application, an electronic device is also provided. See also... Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 300 includes a processor 301 and a memory 302. The memory 302 stores a computer program, which, when executed by the processor 301, will perform actions such as... Figure 1 Any step in the method embodiment shown. The electronic device 300 may also include input / output devices, etc. In a specific embodiment, the electronic device may be a terminal device, etc.

[0158] In one embodiment, a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor 301, causes the processor 301 to perform any of the steps in the above method embodiments.

[0159] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0160] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0161] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for optimizing decision-making rules for a new energy AGC intelligent agent, characterized in that, The method includes: Collect historical power grid fault data, new energy AGC control data, and simulation test data; The collected data is preprocessed to obtain preprocessed data; Based on the preprocessed data, a supervised learning method is used to optimize and train the decision rule model of the agent. In this process, a deep neural network is used to predict the optimal AGC unit power adjustment amount, and the mean square error is used as the loss function to optimize the model and obtain the optimized agent decision rule model. Reinforcement learning is used to further optimize the adaptive decision-making ability of the agent. Specifically, by setting the state space, action space and reward function, the deep Q-learning algorithm is used to enable the agent to iteratively optimize the decision-making strategy through Q-value optimization. The state space is set as follows: in, For power grid frequency deviation, For excess power, For changes in renewable energy power, V bus For critical node voltages; The action space is set as follows: in, This refers to the active power regulation of the AGC unit. This is the reactive power regulation quantity; The reward function is set as follows: in: α, β, γ These are the weighting coefficients; Rated voltage; The decision-making strategy optimized through Q-value iteration uses the following formula: in: m The learning rate; q Discount factor; Q t ( s,a ) represents the Q-value of the current state-action pair; Q t+1 ( s , a ) represents the new Q value after iteration; r represents the immediate reward.

2. The new energy AGC intelligent agent decision-making rule optimization method according to claim 1, characterized in that, The historical power grid fault data includes power flow exceeding limits at transmission sections, power grid frequency deviation, and voltage fluctuations. The new energy AGC control data includes wind power and photovoltaic output data, AGC unit response data, and load changes; The simulation test data includes data related to the interactive behavior of new energy intelligent agents, AGC unit intelligent agents, and power transmission section intelligent agents under different fault scenarios.

3. The new energy AGC intelligent agent decision-making rule optimization method according to claim 1, characterized in that, The preprocessing includes: Normalization, outlier removal, and time series resampling.

4. The new energy AGC intelligent agent decision-making rule optimization method according to claim 1, characterized in that, The method further includes: The optimized agent decision rule model is used as the decision rule for the agent in the simulation model. Based on the simulation model, a joint simulation test of active power control of new energy AGC is carried out to obtain simulation results. The simulation results are compared and verified with actual power grid monitoring data or theoretical analysis results; if there are deviations, the simulation model and algorithm are calibrated.

5. A new energy AGC intelligent agent decision-making rule optimization device, characterized in that, For performing the method as described in any one of claims 1-4; the apparatus comprises: The data acquisition module is used to collect historical power grid fault data, new energy AGC control data, and simulation test data; The preprocessing module is used to preprocess the collected data to obtain preprocessed data; The training optimization module is used to optimize and train the decision rule model of the agent based on the preprocessed data using a supervised learning method. Specifically, a deep neural network is used to predict the optimal AGC unit power adjustment amount, and the mean square error is used as the loss function to optimize the model and obtain the optimized agent decision rule model.

6. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1-4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method as described in any one of claims 1-4.

Images