Multi-power cooperative frequency control method

Abstract

The application provides a multi-power coordinated frequency modulation control method, relates to the power grid dispatching field, and comprises the following steps: acquiring real-time operation data of a power grid, wherein the real-time operation data of the power grid comprises system frequency deviation, load prediction information, and real-time output and regulation capacity data of multiple power sources such as wind power, photovoltaic power, hydropower and thermal power; according to the real-time operation data of the power grid, optimal output distribution schemes of the multiple power sources are calculated by using a mixed integer linear programming algorithm, taking minimization of power generation cost as an objective function, and taking upper and lower limits of the multiple power source output, ramp rate and system frequency deviation limit value as constraint conditions; and the optimal output distribution schemes are converted into power source output adjustment instructions. The application realizes the coordinated frequency modulation of multiple power sources, improves the system frequency stability and economy, enhances the renewable energy grid connection and consumption capacity, and provides an effective solution for large-scale new energy access to the power grid.

Description

Technical Field

[0001] This invention relates to the field of power grid dispatching, and in particular to a multi-source coordinated frequency regulation control method. Background Technology

[0002] In the coordinated frequency regulation control of multiple power sources (wind, solar, hydro, and thermal), the global coordination layer needs to formulate output allocation schemes for multiple power sources based on information such as system frequency deviation and load forecasting to achieve optimal global frequency regulation. However, the response speed and frequency regulation capabilities of different types of power sources vary significantly. How to meet the grid frequency stability requirements while simultaneously considering the economy and reliability of each power source is a complex technical problem. Wind and solar power have large capacities and fast response speeds, but are highly volatile due to weather conditions and difficult to predict accurately. Hydropower has strong regulation capabilities, but is limited by reservoir capacity and downstream ecological flow. Thermal power has a wide regulation range, but its response speed is relatively slow. How to rationally determine the base point power and regulation margin of each power source based on power source characteristics and grid demand, coordinate the output response of each power source at different time scales, avoid conflicting or canceling control commands from multiple power sources, and simultaneously consider unit start-up and shutdown status and reserve capacity to achieve optimal frequency regulation control under multiple time scales and constraints is a key technical problem that urgently needs to be solved. Furthermore, in practical engineering applications, it is also necessary to consider the impact of factors such as communication delay and control time lag on the performance of multi-power supply collaborative control, comprehensively balance the frequency modulation performance and the ease of engineering implementation, and ensure the reliability and effectiveness of the multi-power supply collaborative frequency modulation control strategy. Summary of the Invention

[0003] This invention provides a multi-power source coordinated frequency modulation control method, mainly comprising:

[0004] The system acquires real-time power grid operation data, including system frequency deviation, load forecast information, and real-time output and regulation capacity data of multiple power sources such as wind power, photovoltaic power, hydropower, and thermal power. Based on this real-time power grid operation data, a mixed-integer linear programming algorithm is used to calculate the optimal output allocation scheme for multiple power sources, with the goal of minimizing generation cost and constraints such as upper and lower limits of multi-power source output, ramp rate, and system frequency deviation limit. The optimal output allocation scheme is then converted into power source output adjustment commands and distributed to each power source control system through the power grid dispatch automation system. The system continuously monitors the operation status and performance indicators of the multi-power source coordinated frequency regulation control system, and continuously optimizes the multi-power source coordinated frequency regulation control strategy through online learning and strategy iteration to improve the effectiveness of power grid frequency control.

[0005] Furthermore, after acquiring the real-time operation data of the power grid, the process also includes: preprocessing and feature extraction of the real-time operation data of the power grid, calculating key indicators such as the average system frequency deviation and load forecasting error, and converting the multi-source data into a standardized input format.

[0006] Furthermore, before calculating the optimal power output allocation scheme of multiple power sources based on real-time power grid operation data, the method further includes: classifying the power grid operation status using a decision tree algorithm, determining whether the current status meets the requirements for frequency stability and economic operation, and if not, triggering the multi-power source output optimization and adjustment process.

[0007] Furthermore, after converting the optimal power output allocation scheme into a power output adjustment command, the method further includes: during the power output adjustment process, using power simulation software to monitor the grid operation status and system frequency changes in real time; if the frequency deviation exceeds a preset threshold, triggering an emergency frequency regulation strategy based on fuzzy PID control, and quickly restoring system frequency stability by adjusting power output and load limiting measures.

[0008] Furthermore, after continuously monitoring the operating status and performance indicators of the multi-power source coordinated frequency modulation control system, the method further includes: using a support vector machine algorithm to train historical communication status data, control effect data, and optimization parameters to establish a prediction model, and using the prediction model to achieve autonomous optimization of the star topology and PID control parameters without manual intervention.

[0009] Furthermore, real-time power grid operation data is acquired, including system frequency deviation and real-time power output data of each power source. Simultaneously, power source regulation capacity data and load forecasting results are acquired. The real-time power grid operation data, power source regulation capacity data, and load forecasting results are cleaned and feature extracted. Through data fusion technology, comprehensive evaluation data is generated. This comprehensive evaluation data is input into a Q-learning reinforcement learning model. The power grid operation optimization model is periodically updated by iteratively updating the Q-value table. Based on the power source operation status evaluation results, the optimization objective function and constraints for multi-power source coordinated frequency regulation control are determined. By solving the optimization problem, the optimal frequency regulation control strategy for each power source is obtained. The optimal frequency regulation control strategy includes AGC commands and primary frequency regulation power allocation.

[0010] Furthermore, when continuously optimizing the multi-power source coordinated frequency modulation control strategy, the strategy gradient REINFORCE algorithm is used for online learning. Through continuous trial and error and strategy iteration, a continuously optimized multi-power source coordinated frequency modulation control strategy is obtained, the frequency modulation power allocation and control command issuance scheme is determined, and the grid connection point frequency is quickly and accurately adjusted.

[0011] Furthermore, real-time operating status data of the multi-power supply cooperative control system is acquired, including power supply status and load demand, and used as input to the particle swarm optimization algorithm. A standard particle swarm optimization algorithm is employed, using star topology parameters and PID controller parameters as optimization variables to establish an optimization model. Based on the optimization results, fuzzy control is used to dynamically adjust the control cycle and control gain. If the system performance indicators fail to meet requirements for several consecutive control cycles, the current status data is re-input into the particle swarm optimization algorithm for further optimization and adjustment.

[0012] Furthermore, the power grid dispatch automation system and each power control system adopt a communication network based on the IEC61850 protocol to realize the real-time transmission and interaction of power grid operation information and control commands.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0014] This invention discloses a multi-source coordinated frequency regulation control method. Considering the characteristics of multiple power sources in the power grid, such as wind power, photovoltaic power, hydropower, and thermal power, a mixed-integer linear programming algorithm is used to calculate the optimal power output allocation scheme to ensure system frequency stability and economy. For the volatility of wind and solar power generation, support vector machine prediction and moving average smoothing control are employed. For hydropower constrained by reservoir capacity, dynamic programming is used for optimized scheduling. For thermal power, deep peak shaving technology is used to expand the regulation range. In multi-source coordination, a hierarchical control architecture is adopted, with the global layer defining the base power and regulation margin, and a genetic algorithm used for multi-objective optimization. To reduce the impact of communication delay, Kalman filter predictive control and adaptive algorithms are used. Taking into account performance and implementation difficulty, the control strategy is optimized through simulation and practical verification. This invention achieves coordinated frequency regulation of multiple power sources, improves system frequency stability and economy, enhances the grid connection and absorption capacity of renewable energy, and provides an effective solution for large-scale renewable energy grid integration. Attached Figure Description

[0015] Figure 1 This is a flowchart of a multi-power source coordinated frequency modulation control method according to the present invention. Detailed Implementation

[0016] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0017] like Figure 1 This embodiment of a multi-power source coordinated frequency modulation control method may specifically include:

[0018] S101. Based on real-time grid operation data, obtain system frequency deviation and load forecast information. Combine the real-time output and regulation capabilities of multiple power sources such as wind power, photovoltaic power, hydropower and thermal power, and calculate the optimal output allocation scheme of multiple power sources through mixed integer linear programming algorithm to ensure system frequency stability and power economy.

[0019] Real-time power grid operation data is acquired, including system frequency deviation, load forecasting information, and real-time output and regulation capacity data of multiple power sources such as wind power, photovoltaic power, hydropower, and thermal power. The acquired data undergoes preprocessing and feature extraction to calculate key indicators such as the mean system frequency deviation and load forecasting error, and the multi-power source data is converted into a standardized input format. A decision tree algorithm is used to classify the power grid operating status and determine whether the current status meets the requirements for frequency stability and economic operation. If not, a multi-power source output optimization adjustment process is triggered. A mixed-integer linear programming model is constructed, with the minimization of generation cost as the objective function and constraints such as upper and lower limits of multi-power source output, ramp rate, and system frequency deviation limits, to solve for the optimal output allocation scheme for multiple power sources. The optimal output allocation scheme is converted into power source output adjustment commands and issued to each power source control system through the power grid dispatch automation system. During the power source output adjustment process, power simulation software is used to monitor the power grid operating status and system frequency changes in real time. If the frequency deviation exceeds a preset threshold, an emergency frequency regulation strategy based on fuzzy PID control is triggered, and measures such as adjusting power source output and load curtailment are implemented to quickly restore system frequency stability. After the adjustment, the grid operation status and economic indicators such as generation costs before and after the multi-source output adjustment are compared and analyzed to evaluate the optimization effect. The evaluation results are used to correct the objective function coefficients and constraint parameters in the mixed-integer linear programming model to achieve adaptive optimization of the model. Based on historical grid operation data and optimization results, a predictive model for multi-source output allocation is established using a support vector machine regression algorithm. This model is used to predict the grid operation status and power output allocation in the future, providing auxiliary decision support for optimized scheduling and improving system frequency stability and economic operation.

[0020] For example, acquiring real-time power grid operation data is like a doctor examining a patient and obtaining various physiological indicators. For instance, system frequency deviation is like the stability of a heartbeat, load forecasting information is like predicting a patient's future activity level, and real-time output and regulation capacity data from multiple power sources such as wind, solar, hydro, and thermal power are like the operational status and potential of various organs in the human body. This data can be acquired through sensors and smart meters deployed at various nodes of the power grid. For example, smart meters can monitor users' electricity consumption in real time, and the collected data constitutes the system load data. Wind turbines in wind farms are equipped with sensors that measure wind speed, direction, etc., and then calculate the output of the wind turbines. Data preprocessing and feature extraction are like a doctor analyzing examination results. The acquired raw data undergoes cleaning and noise reduction, such as removing abnormal load spikes. Calculating key indicators such as the average system frequency deviation and load forecasting error is like calculating a patient's average heart rate and blood pressure. Converting multi-power source data into a standardized input format is like unifying the patient's various indicators into a standard unit of measurement. For example, converting the output of different types of power sources into MW. Classifying the power grid's operating status using decision tree algorithms is like a doctor diagnosing a patient's condition based on various indicators. Decision trees categorize the grid's operating status into frequency stability, slight frequency fluctuation, and severe frequency fluctuation based on indicators such as system frequency deviation and load forecasting error. For example, if the frequency deviation exceeds a certain threshold, it is classified as a frequency fluctuation state. This helps to quickly determine whether the grid's operating status meets the requirements for frequency stability and economical operation. Triggering the multi-source output optimization adjustment process is like a doctor developing a treatment plan based on a patient's condition. If the grid's operating status does not meet the requirements, the output of multiple sources needs to be adjusted. For example, when the system frequency is low, the generator output needs to be increased; conversely, the output needs to be reduced. Constructing a mixed-integer linear programming model is like a doctor developing a medication plan, considering the dosage, interactions, and patient's tolerance of various drugs. Minimizing generation cost as the objective function is like minimizing treatment costs. Constraints such as upper and lower limits for multi-source output, ramp rates, and system frequency deviation limits are like medication dosages, contraindications, and patient's physical limitations. For example, the output of a thermal power unit cannot exceed its maximum output limit, nor can it fall below its minimum technical output limit; its output change rate also cannot exceed its ramp rate limit. By solving this model, the optimal output allocation scheme for each power source can be obtained. Transforming the optimal output allocation scheme into power source output adjustment commands is like a doctor converting a medication plan into a specific prescription. These commands are then distributed to the control systems of each power source through the power grid dispatch automation system, much like a pharmacist dispensing medication according to a prescription. For instance, sending the output adjustment command of the thermal power unit to the control system of the thermal power plant will automatically adjust the fuel supply and turbine speed, thereby changing the generator output.Real-time monitoring of the power grid's operating status and system frequency changes using power simulation software is like a doctor continuously monitoring a patient's vital signs during treatment. If the frequency deviation exceeds a preset threshold, it's like certain indicators of the patient becoming abnormal. Triggering an emergency frequency regulation strategy based on fuzzy PID control is like a doctor taking emergency measures to stabilize the patient's condition. Quickly restoring system frequency stability by adjusting power output and load limiting is like controlling the patient's condition through medication or surgery. For example, when the system frequency suddenly drops significantly, fast-response pumped-storage units or gas turbine units can be activated to increase output, while simultaneously limiting some non-critical loads, thereby rapidly increasing the system frequency. After adjustment, comparing and analyzing the power grid's operating status and economic indicators such as generation costs before and after the multi-power output adjustment to evaluate the optimization effect is like a doctor assessing the treatment effect. For example, comparing the degree of system frequency fluctuation and the level of generation costs before and after adjustment determines whether the adjustment is effective. The evaluation results are used to correct the objective function coefficients and constraint parameters in the mixed-integer linear programming model, achieving adaptive optimization of the model, just like a doctor adjusting medication based on treatment results. For example, if the actual output adjustment effect of a certain power source is found to be less than expected, the output cost coefficient or ramp rate constraint of that power source in the model can be adjusted. Based on historical power grid operation data and optimization results, a predictive model for multi-power source output allocation is established using a support vector machine regression algorithm, much like a doctor predicting the progression of a patient's condition based on their medical history and treatment records. This model is used to predict the power grid's operating status and power source output allocation over a future period, providing auxiliary decision support for optimized scheduling, much like a doctor making preparations in advance based on a patient's condition to prevent it from worsening. This can improve the operational efficiency and security of the power grid.

[0021] S102. To address the volatility and forecasting challenges of wind and solar power generation, a short-term power forecasting algorithm based on support vector machines and a real-time power smoothing control strategy based on moving averages are adopted to reduce the impact of wind and solar power fluctuations on grid frequency. Based on the forecast results and real-time power data, the base point power and regulation margin of wind and solar power sources are dynamically adjusted to improve their ability and reliability to participate in system frequency regulation.

[0022] Historical power and meteorological data from wind farms and photovoltaic power plants, including parameters such as wind speed, wind direction, irradiance, and temperature, are acquired. Data cleaning and anomaly detection algorithms are used to remove abnormal and invalid data, and the data is normalized. Power and meteorological data are time-aligned and synchronized based on their timestamps. Meteorological features with high power correlation, such as wind speed and irradiance, are selected to construct the input feature vector for the prediction model. Support Vector Regression (SVR) algorithm is used to establish a nonlinear mapping relationship between the input feature vector and the output power. Hyperparameters of the SVR model, such as kernel type, penalty coefficient, and kernel parameters, are optimized through grid search and cross-validation. The optimized SVR model is used to predict power from real-time meteorological data of wind farms and photovoltaic power plants. The prediction time scale is set according to grid dispatch requirements, such as 15 minutes or 1 hour. The input feature data is updated every time step, and the predicted power value for the future period is output. The predicted power sequence is smoothed using a moving average filtering algorithm. The initial smoothing window size is set to 10 time steps, and the weight coefficient is 6. Based on the peak-shaving demand of the power grid and the fluctuating characteristics of wind and solar resources, the smoothing window size and weighting coefficients are adaptively adjusted to balance the smoothness and response speed of power output. The smoothed power output is used as the base power for wind farms and photovoltaic power plants. The regulation margin of wind and solar power sources is optimized in real time according to the frequency regulation requirements of the power grid. When the grid frequency is too high, the base power is reduced to increase the regulation margin; when the grid frequency is too low, the base power is increased to decrease the regulation margin. When the grid frequency deviates from the rated value, wind farms and photovoltaic power plants initiate primary and secondary frequency regulation, respectively. Wind farms use a combination of pitch control and variable speed control to quickly adjust active power output based on frequency deviation and regulation margin. Photovoltaic power plants use a combination of power reserve and energy storage to smoothly adjust active power output based on frequency deviation and regulation margin. The effectiveness of wind farms and photovoltaic power plants in frequency regulation is continuously monitored and evaluated using indicators such as frequency regulation contribution, frequency regulation accuracy, and frequency regulation response time. Based on the evaluation results, we can optimize the feature selection and hyperparameter settings of the SVR prediction model, optimize the smoothing window and weight parameters of the moving average filtering algorithm, optimize the regulation margin and control strategy of wind and solar power, and improve the reliability and economy of wind and solar power participating in grid frequency regulation.

[0023] For example, wind farms and photovoltaic power plants experience significant power output fluctuations, making accurate prediction of their output power crucial for stable grid operation. Acquiring historical power and meteorological data is fundamental to prediction. For instance, a wind farm might record wind speed, wind direction, temperature, and corresponding power generation every 15 minutes, while a photovoltaic power plant might record irradiance, temperature, and corresponding power generation every 15 minutes. This data forms the training set for the prediction model. During data preprocessing, cleaning and anomaly detection are necessary. For example, if wind speed data for a particular day shows abnormal spikes due to sensor malfunction, these need to be removed or interpolated using surrounding data. Power data may also be missing due to equipment maintenance and requires supplementation. Normalization unifies data with different dimensions to the same scale; for example, wind speed, temperature, and irradiance data are normalized to between 0 and 1, preventing certain features from having an excessive impact on the model. Time alignment and synchronization are also critical, ensuring a one-to-one correspondence between meteorological and power data. For example, if wind speed data is timestamped at the hour, while power data is timestamped 10 minutes after the hour, they need to be aligned to the same timestamp. Choosing appropriate input features is crucial for model accuracy. Generally, wind speed and irradiance have the highest correlation with wind and solar power output. In addition, factors such as wind direction and temperature also have some impact on output. Correlation analysis and other methods can be used to select features and construct input feature vectors. For example, analyzing historical data reveals that wind speed and direction have a greater impact on wind farm output than temperature, so wind speed and direction can be used as input features. Support Vector Regression (SVR) algorithms can establish a nonlinear mapping relationship between input feature vectors and output power, making them suitable for handling nonlinear relationships and small sample data. For example, an SVR model can be trained using historical data to learn the complex relationship between wind speed, wind direction, and wind farm output. Grid search and cross-validation can optimize the hyperparameters of the SVR model, improving its generalization ability. For example, grid search can be used to find the optimal penalty coefficient and kernel function parameters, avoiding overfitting or underfitting. Using a trained SVR model, power output can be predicted for a future period. For example, based on 24-hour weather forecast data, the power output of wind farms and photovoltaic power plants can be predicted for the next 24 hours. The prediction timescale can be adjusted according to scheduling needs. For example, a 15-minute prediction timescale can be set for intraday scheduling; for long-term planning, a 1-hour or even longer prediction timescale can be set. Sliding updates of input feature data ensure the real-time performance of the prediction model. For example, meteorological data can be updated every 15 minutes, and the power output for the next 15 minutes can be predicted. Moving average filtering can smooth the predicted power sequence and reduce fluctuations caused by prediction errors. For example, using a moving average filter with a window size of 10 time steps can make the predicted power curve smoother. Adaptively adjusting the smoothing window size and weighting coefficients can balance smoothness and response speed.For example, when wind and solar power output fluctuates significantly, the window size can be reduced to improve response speed; when output fluctuations are small, the window size can be increased to improve smoothness. The smoothed power output serves as the base power for wind and solar power, used for subsequent frequency regulation. The participation of wind and solar power in frequency regulation can improve grid stability. When the grid frequency is too high, the base power is reduced to increase the regulation margin, and vice versa. For example, if the grid frequency increases by 0.1 Hz, the wind farm reduces its base power by 5 MW, and the photovoltaic power station reduces its base power by 3 MW, reserving more output for frequency regulation. Primary frequency regulation is the primary means of regulating grid frequency and requires a rapid response. Wind farms can quickly adjust output through pitch control and speed control. For example, when the grid frequency decreases, the wind turbine increases its pitch angle or speed to increase output. Photovoltaic power stations can participate in primary frequency regulation through power reservation and energy storage. For example, when the grid frequency decreases, reserved power is released or power is output from the energy storage system. Secondary frequency regulation has a relatively slower response speed and is used to restore the frequency to its rated value. Wind and solar power can smoothly adjust their output and participate in secondary frequency regulation based on frequency deviation and regulation margin. Continuous monitoring and evaluation of the effectiveness of wind and solar power in frequency regulation allows for continuous optimization of control strategies. For example, statistical analysis can be conducted on indicators such as the frequency regulation contribution, frequency regulation accuracy, and frequency regulation response time of wind and solar power, analyzing their response characteristics under different frequency deviations. Based on the evaluation results, the SVR prediction model, moving average filtering algorithm, and regulation margin and control strategy of wind and solar power can be adjusted to improve the reliability and economy of wind and solar power participation in grid frequency regulation. For instance, if the frequency regulation response time of a wind farm is found to be too long, its control strategy can be optimized to improve the response speed.

[0024] S103. To address the limitation of hydropower regulation capacity by reservoir capacity and downstream ecological flow, a dynamic programming reservoir optimization scheduling model is adopted to meet the frequency regulation needs of the power grid while taking into account the long-term operational benefits of the reservoir and ecological environmental protection. Based on the reservoir optimization scheduling results, the output distribution and start-up / shutdown status of hydropower units are optimized to improve the flexibility and economy of hydropower frequency regulation.

[0025] First, based on the real-time water level and inflow of the reservoir, combined with the downstream ecological flow demand, a dynamic programming algorithm is used to determine the optimal scheduling strategy of the reservoir over a future period, resulting in the outflow process. The state variable in the dynamic programming is defined as the reservoir water level, and the decision variable is the outflow. The state transition equation is: Water level (t+1) = Water level (t) + Inflow (t) - Outflow (t) - Evaporation (t) - Leakage (t). The objective function is to maximize hydropower generation over a future period while satisfying the downstream ecological flow demand. Then, based on the frequency regulation requirements of the power grid and combined with the outflow obtained from the optimized reservoir scheduling, a mixed-integer programming algorithm is used to optimize the output allocation and start-up / shutdown status of the hydropower units. Decision variables include the output and start-up / shutdown status of each unit. Constraints include reservoir outflow constraints, unit output limits, and unit start-up / shutdown time constraints. The objective function is to maximize the long-term operational benefits of the hydropower station, such as maximizing power generation revenue or minimizing fuel costs. The unit combination scheme is determined by dividing the units into several levels based on their capacity. For example, small, medium, and large generating units are selected and combined within each category to form multiple optional unit combination schemes, improving the flexibility of hydropower frequency regulation. In mixed-integer programming, the output allocation and start-up / shutdown states of different unit combination schemes are considered to meet frequency regulation requirements and maximize long-term benefits. By setting minimum duration constraints for unit start-up and shutdown, frequent start-up and shutdown are avoided, reducing unit wear. Finally, based on long-term reservoir operation data, the Q-learning algorithm is used to adaptively adjust various parameters in the dynamic programming model, such as the weight coefficients in the objective function and the evaporation / leakage coefficients in the state transition equation, enabling the optimized scheduling strategy to adapt to changes in hydrological conditions and ensuring the maximization of long-term reservoir benefits. The state space is defined as reservoir water level and inflow, the action space as outflow, and the reward function as power generation. Learning rates and discount factors are set based on hydrological conditions and operational experience.

[0026] For example, the optimal scheduling strategy of a reservoir is crucial to both the efficiency of hydropower generation and the guarantee of downstream ecological water demand. Dynamic programming algorithms can effectively solve this type of multi-stage decision-making problem. Taking a reservoir as an example, suppose we need to determine the outflow strategy for the next week. First, the reservoir water level is divided into multiple discrete states, for example, several water level states from dead water level to flood control limit water level at 0.5-meter intervals. Then, the outflow is also divided into multiple discrete values, for example, several outflow values ​​from minimum ecological flow to maximum flood discharge flow at 10 cubic meters per second intervals. The state transition equation describes the change law of the reservoir water level over time, that is, the water level at the next moment is jointly determined by the current water level, inflow, outflow, evaporation, and leakage. For example, if the current water level is 100 meters, the inflow is 50 cubic meters per second, the outflow is 30 cubic meters per second, and the evaporation and leakage are 2 cubic meters per second, then the water level at the next moment is 100 + 50 - 30 - 2 = 118 meters. The objective function aims to maximize hydropower generation over the next week while satisfying downstream ecological flow constraints. For example, if the minimum downstream ecological flow is set at 20 cubic meters per second, the outflow from the reservoir at any given time must not fall below this value during the optimization process. Dynamic programming can be used to obtain the optimal outflow for each moment within the next week, thus guiding the reservoir's operation. The power output allocation and start-up / shutdown status of the hydropower units directly affect the operational efficiency of the hydropower station. Mixed-integer programming can optimize the operating status of the units while satisfying various constraints. Assume the hydropower station has four units, each with different output limits and start-up / shutdown time constraints. For example, unit 1 has an output range of 50 to 100 megawatts and a minimum start-up / shutdown time of 2 hours; unit 2 has an output range of 30 to 80 megawatts and a minimum start-up / shutdown time of 1 hour. The reservoir's outflow determines the total power output of the hydropower station. For instance, based on the reservoir's optimized scheduling results, the total power output of the hydropower station corresponding to the outflow at a certain future moment is 200 megawatts. The objective function of mixed-integer programming can be set to maximize power generation revenue or minimize fuel costs. Constraints include reservoir outflow constraints (the total output of the four generating units must equal 200 MW), unit output limits (each unit's output must be within its specified range), and unit start-up and shutdown time constraints (each unit's start-up and shutdown time must meet its minimum start-up and shutdown time limit). By solving the mixed-integer programming problem, the optimal output and start-up / shutdown state of each generating unit can be obtained, thereby maximizing the operational benefits of the hydropower station. To meet the frequency regulation requirements of the power grid, it is necessary to consider the output allocation and start-up / shutdown states of different generating unit combinations. Assume that the four generating units are divided into two groups: units 1 and 2 in one group, and units 3 and 4 in another. This results in two generating unit combination schemes: Scheme 1, operating units 1 and 2; Scheme 2, operating units 3 and 4.In mixed-integer programming, the output allocation and start-up / shutdown states of two schemes are considered separately, and the optimal scheme is selected based on frequency regulation requirements and long-term benefits. For example, when the grid frequency decreases, hydropower stations need to rapidly increase output; in this case, a unit combination scheme with faster start-up speed and stronger regulation capability can be selected. By setting minimum duration constraints for unit start-up and shutdown, such as stipulating that each unit runs for at least 2 hours after startup, frequent start-up and shutdown of units can be avoided, reducing unit wear and extending service life. The long-term operation of a reservoir is affected by various factors, such as inflow conditions and climate change. The Q-learning algorithm can adaptively adjust the parameters in the dynamic programming model based on the long-term operation data of the reservoir, thereby improving the adaptability of the scheduling strategy. The reservoir water level and inflow are used as state variables, the outflow as action variables, and power generation as the reward function. For example, when the reservoir water level is high and the inflow is large, the outflow can be appropriately increased to increase power generation; when the reservoir water level is low and the inflow is small, the outflow needs to be reduced to ensure water supply security. Through continuous learning and adjustment, the Q-learning algorithm can find the optimal outflow strategy, thereby maximizing the long-term benefits of the reservoir. The learning rate and discount factor are crucial parameters in the Q-learning algorithm and need to be set based on hydrological conditions and operational experience. For example, the learning rate determines the speed at which the algorithm learns new knowledge, while the discount factor determines the degree to which future rewards influence current decisions.

[0027] S104. To address the issue of thermal power plants having a wide regulation range but relatively slow response speed, deep peak-shaving technology for thermal power units is adopted to expand their regulation range and improve their response speed. Based on the regulation characteristics of thermal power units, load allocation and reserve capacity are optimized to improve their ability and reliability in participating in system frequency regulation.

[0028] Real-time operating parameters of the thermal power units are acquired, including current load, upper and lower limits of the regulation range, and response speed, forming a parameter set. Based on the parameter set, a BP neural network algorithm is used to establish a regulation characteristic model of the thermal power units, obtaining optimized regulation range and response speed thresholds. If the regulation range and response speed of the thermal power units do not reach the optimized thresholds, deep peak shaving control is triggered, expanding the regulation range and improving the response speed by optimizing the combustion and steam-water systems. Based on system load forecasts and reserve capacity requirements, a particle swarm optimization algorithm is used to calculate the optimal load allocation scheme for the thermal power units, balancing the unit load rate and reserve capacity. The load allocation scheme is then sent to the thermal power unit control system, and the unit output is adjusted through the automatic generation control (AGC) system, improving the unit's ability to participate in system frequency regulation. During frequency regulation, a PID control algorithm is used to dynamically optimize frequency regulation parameters, improving frequency regulation accuracy and reliability. Based on the unit's frequency regulation performance evaluation results, the PID control parameters are adjusted and optimized in real time during the frequency regulation process, continuously improving the regulation performance of the thermal power units and the system frequency regulation level. By continuously optimizing deep peak shaving strategies and load allocation schemes, a closed-loop optimization is formed, which continuously improves the regulation performance of thermal power units and the system frequency regulation level.

[0029] For example, real-time operating parameters of thermal power units are the foundation for deep peak-shaving control and load allocation. Through sensors and data acquisition systems, key parameters such as the current load, upper and lower limits of the regulation range, and response speed of each thermal power unit can be acquired in real time and aggregated into a parameter set. For instance, a 600MW thermal power unit might have a current load of 500MW, an upper limit of the regulation range of 550MW, a lower limit of 400MW, and a response speed of 2MW / min. Collecting these parameters helps understand the current state and regulation capacity of the unit. Using the collected parameter set, a BP neural network algorithm can be employed to establish a model of the thermal power unit's regulation characteristics. A BP neural network is a multi-layer feedforward neural network that, by learning from a large amount of historical data, can establish a mapping relationship between unit parameters and regulation characteristics. For example, historical load, regulation range, and response speed data can be used as input, and actual peak-shaving performance indicators can be used as output to train the BP neural network model. Through training, the model can predict the optimal regulation range and response speed thresholds of the unit under different load and parameter conditions. Assuming that model calculations show the optimal adjustment range for this 600MW unit at a 500MW load is 450MW to 580MW, with a response speed threshold of 4MW / min, if the actual adjustment range and response speed of the thermal power unit do not reach the optimal threshold, deep peak shaving control needs to be triggered. Deep peak shaving control refers to further expanding the unit's adjustment range and improving its response speed by optimizing the combustion system (e.g., adjusting burner tilt angle, changing fuel ratio) and the steam-water system (e.g., optimizing steam temperature and pressure control). For example, for the aforementioned 600MW unit, optimizing the combustion system can increase its response speed from 2MW / min to 3MW / min, closer to the optimal threshold of 4MW / min. Simultaneously, by improving the steam-water system control strategy, its lower limit of adjustment range can be extended from 400MW to 380MW. To balance the unit's load factor and reserve capacity, the optimal load allocation scheme for the thermal power unit needs to be calculated using the particle swarm optimization algorithm based on system load forecasts and reserve capacity requirements. The particle swarm optimization algorithm is an intelligent optimization algorithm that simulates the foraging behavior of bird flocks and can effectively search for the optimal solution. For example, assuming the system load forecast for the next hour is 1000MW, the reserve capacity requirement is 200MW, and two 600MW thermal power units are available, the optimal load allocation for each unit can be calculated using particle swarm optimization. For instance, one unit could handle 550MW of load, and the other 450MW, satisfying the system load requirement while leaving sufficient reserve capacity. The calculated optimal load allocation scheme needs to be sent to the thermal power unit control system, and the unit output is adjusted through the Automatic Generation Control (AGC) system. The AGC system can automatically adjust the output of each unit according to the load allocation scheme to achieve the target value.For example, if the load allocated to a certain generating unit is 550MW, the AGC system will automatically control the unit's output to stabilize it around 550MW, thereby improving the unit's ability to participate in system frequency regulation. During frequency regulation, to improve regulation accuracy and reliability, PID control algorithms are needed to dynamically optimize regulation parameters. PID control is a classic control algorithm that achieves precise control of the controlled object through proportional, integral, and derivative control actions. For example, when the system frequency drops, the PID controller will adjust the output of the thermal power unit in real time according to the frequency deviation, rapidly increasing output and restoring system frequency stability. To continuously improve the regulation performance of thermal power units and the system frequency regulation level, the PID control parameters during the frequency regulation process need to be adjusted and optimized in real time based on the unit's frequency regulation performance evaluation results. For example, if the unit's frequency regulation response speed is found to be slow, the proportional coefficient of the PID controller can be appropriately increased to improve the response speed. By continuously optimizing deep peak shaving strategies and load allocation schemes, combined with dynamic adjustment of PID parameters, a closed-loop optimization can be formed, continuously improving the regulation performance of thermal power units and the system frequency regulation level, better meeting the needs of the power system.

[0030] S105. In multi-power source coordinated frequency regulation control, a hierarchical coordinated control architecture is adopted. The global coordination layer determines the base power and regulation margin of multiple power sources based on system frequency deviation, load forecast information, and the real-time output and regulation capabilities of each power source, coordinating the output response of each power source at different time scales. A multi-objective optimization algorithm based on genetic algorithm is adopted to satisfy the grid frequency stability constraints while taking into account the economy and reliability of each power source, avoiding contradictions or cancellations between multi-power source control commands.

[0031] Based on real-time grid operation data, information such as system frequency deviation, real-time output and regulation capacity of each power source is acquired and combined with load forecasting results, then input to the global coordination control layer. The global coordination control layer employs the NSGA-II multi-objective optimization algorithm, with grid frequency stability, power source economy, and reliability as optimization objectives, to generate optimized results for multi-power source base point power and regulation margin. The multi-power source base point power and regulation margin generated by the global coordination control layer are then distributed to each power source control unit as operational constraints. Each power source control unit, based on the received base point power and regulation margin and its own operating status, determines its own output regulation strategy using a sequential quadratic programming method. During real-time operation, each power source control unit dynamically adjusts its output using a PID control algorithm based on the instructions issued by the global coordination control layer and its own optimization results, achieving multi-power source coordinated frequency regulation. The effectiveness of multi-power source coordinated frequency regulation is assessed through real-time monitoring of the grid frequency. If the frequency deviation exceeds a preset threshold, an early warning system triggers the global coordination control layer to re-optimize the multi-power source and generate new control instructions. Based on the load change trend obtained from the power grid load forecasting model and the evaluation results of the power source operation status assessment system, the objective function and constraints of the optimization model in the global coordination control layer are updated periodically using an adaptive dynamic programming algorithm to adapt to the dynamic changes in power grid operation and ensure the continuous optimization of multi-power source coordinated frequency regulation control.

[0032] For example, the global coordination control layer is responsible for coordinating the output of all power sources to maintain grid frequency stability. It receives real-time operational data from the grid, such as system frequency deviation, real-time output and regulation capabilities of each power source (e.g., thermal, hydro, photovoltaic), and load forecast results. For instance, at a certain moment, the system frequency deviation is -0.1Hz, indicating the frequency is slightly below the standard value; a thermal power unit currently outputs 500MW, with an adjustable range of 400MW to 550MW; a hydropower unit currently outputs 200MW, with an adjustable range of 100MW to 300MW; and load forecast results show that the load will increase by 100MW in the next hour. All this information is input into the global coordination control layer. The global coordination control layer uses the NSGA-II multi-objective optimization algorithm. The NSGA-II algorithm can simultaneously consider multiple optimization objectives, such as grid frequency stability, power source economy, and reliability. Based on the input real-time data and load forecast results, it calculates a set of optimal base point power and regulation margins and allocates them to each power source. For example, to cope with a future load increase of 100MW per hour, the global coordination control layer might increase the base power of thermal power units to 520MW, maintaining a regulation margin of 50MW; and increase the base power of hydropower units to 230MW, maintaining a regulation margin of 70MW. This approach satisfies the demand for increased load while maintaining a certain frequency stability margin, and also considers the economy and reliability of different power sources. The base power and regulation margin of the multiple power sources calculated by the global coordination control layer will be distributed to each power source control unit. For example, the thermal power unit control unit will receive an instruction for a base power of 520MW and a regulation margin of 50MW; the hydropower unit control unit will receive an instruction for a base power of 230MW and a regulation margin of 70MW. These instructions serve as constraints for the operation of each power source, guiding its output adjustment. After receiving the instructions from the global coordination control layer, each power source control unit, based on its own operating status, such as current output, adjustable range, and response speed, will determine its own output adjustment strategy using a sequential quadratic programming method. For example, a thermal power unit control unit, based on a base power of 520MW and a regulation margin of 50MW, as well as its own operating status, might formulate a strategy to gradually increase output, such as increasing by 2MW per minute until reaching 520MW. During real-time operation, each power control unit dynamically adjusts its output using a PID control algorithm based on instructions from the global coordination control layer and its own optimization results. The PID control algorithm is a classic control algorithm that dynamically adjusts the power output based on real-time frequency deviations, achieving precise control. For example, if the system frequency is lower than the standard value, the PID controller will automatically increase the power output according to the magnitude of the frequency deviation until the frequency returns to the standard value. Real-time monitoring of the grid frequency can determine the effectiveness of multi-power source coordinated frequency regulation.If the frequency deviation exceeds a preset threshold, such as ±0.2Hz, the early warning system will trigger the global coordinated control layer to re-optimize the multi-source power supply and generate new control commands. For example, if the system frequency is detected to be continuously decreasing and exceeding the -0.2Hz threshold, the early warning system will immediately notify the global coordinated control layer. The global coordinated control layer will reassess the current grid operating status and load forecast results, recalculate the base point power and regulation margin of each power source, and generate new control commands to be issued to each power source control unit. To adapt to the dynamic changes in grid operation, the optimization model in the global coordinated control layer needs to be updated periodically. Using an adaptive dynamic programming algorithm, the objective function and constraints in the global coordinated control layer can be updated periodically based on the load change trend obtained from the grid load forecast model and the evaluation results of the operating status of each power source. For example, if it is predicted that the load will continue to grow in the future, the objective function of the optimization model can be adjusted to place greater emphasis on frequency stability; if the operating status of a power source changes, such as a reduction in its adjustable range, the corresponding constraints need to be updated. Through continuous optimization, it can be ensured that the multi-source coordinated frequency regulation control is always in the optimal state, better meeting the needs of the power system.

[0033] S106. Based on real-time power grid operation data, obtain system frequency deviation and real-time output of each power source. Combine power source regulation capability and load forecast results, and periodically update the global coordination control layer optimization model through adaptive dynamic programming algorithm. Judge the power source operation status evaluation results, determine the optimization objective function and constraints, realize continuous optimization of multi-power source coordinated frequency regulation control, and adapt to the dynamic changes in power grid operation.

[0034] Real-time power grid operation data is acquired, including system frequency deviation and real-time output data of each power source, along with power source regulation capacity data and load forecasting results. The real-time power grid operation data, power source regulation capacity data, and load forecasting results are cleaned and feature extracted. Data fusion techniques, such as Kalman filtering, are used to generate comprehensive evaluation data. The Q-learning reinforcement learning algorithm within adaptive dynamic programming is employed. The comprehensive evaluation data is input into the Q-learning model, and the Q-value table is iteratively updated, periodically updating the power grid operation optimization model of the global coordination control layer. Using the power grid operation optimization model, a multi-objective optimization function is constructed, comprehensively considering multiple objectives such as power grid security, economy, and low carbon emissions, to determine the current power source operating status and obtain the power source operating status evaluation results. Based on the power source operating status evaluation results, the optimization objective function for multi-power source coordinated frequency regulation control is determined, such as minimizing frequency deviation and minimizing frequency regulation costs, along with corresponding power source output constraints and ramp-up rate constraints, serving as the basis for coordinated frequency regulation control. The objective function and constraints are input into the multi-source coordinated frequency regulation control model. This model, based on rolling optimization and model predictive control methods, solves the optimization problem to obtain the optimal frequency regulation control strategy for each power source, including AGC commands and primary frequency regulation power allocation. The optimal frequency regulation control strategy is then distributed to each power source control unit, including thermal power, hydropower, wind power, and photovoltaic power, controlling each power source to adjust its output according to the frequency regulation commands, achieving rapid recovery and stabilization of the grid frequency. Through rolling optimization and feedforward correction, the multi-source coordinated frequency regulation control model is continuously optimized using real-time grid operation data, adaptively adjusting the frequency regulation control strategy to adapt to dynamic changes in grid operating conditions such as load fluctuations and changes in renewable energy output, ensuring long-term stability of the grid frequency near its rated value.

[0035] For example, real-time power grid operation data, such as system frequency deviation and real-time output data of various power sources (thermal power, hydropower, photovoltaic, wind power, etc.), forms the basis for multi-power source coordinated frequency regulation control. System frequency deviation reflects the real-time supply and demand balance of the power grid; for example, a frequency deviation of -0.1Hz indicates insufficient power supply. Real-time output data of each power source reflects its current operating status; for example, a thermal power unit outputs 500MW, and a hydropower unit outputs 200MW. In addition to real-time power grid operation data, it is also necessary to obtain power source regulation capacity data, such as the adjustable range of a thermal power unit being 400MW to 550MW, and load forecast data, such as a predicted load increase of 100MW in the next hour. These data constitute the input information for multi-power source coordinated frequency regulation control. After obtaining the raw data, data cleaning and feature extraction are required. For example, outlier data can be removed, the data can be smoothed, and features such as frequency change trends and load change trends can be extracted. Then, data fusion techniques, such as Kalman filtering, are used to fuse data from multiple sources into comprehensive evaluation data. For example, system frequency deviation, load forecast results, and power output data from each power source can be fused into a comprehensive index reflecting the real-time operating status of the power grid. The Q-learning reinforcement learning algorithm in adaptive dynamic programming can be used to periodically update the power grid operation optimization model of the global coordinated control layer. Taking Q-learning as an example, it learns an optimal action strategy by continuously interacting with the environment. The comprehensive evaluation data is input into the Q-learning model, and the model selects an action based on the current state, such as adjusting the output of a power source. Then, the model updates the Q-value table based on environmental feedback (such as new frequency deviations), continuously optimizing the control strategy. The power grid operation optimization model needs to comprehensively consider multiple objectives such as power grid security, economy, and low carbon emissions. For example, a multi-objective optimization function can be constructed, simultaneously considering minimizing frequency deviation, minimizing frequency regulation costs, and minimizing carbon emissions. By solving this optimization function, the power source operating status evaluation results can be obtained, such as evaluating the economy, reliability, and low carbon emissions of each power source. Based on the power source operating status evaluation results, the optimization objective function and constraints for multi-power source coordinated frequency regulation control can be determined. For example, if the current frequency deviation is large, minimizing the frequency deviation should be prioritized; if the system operating cost is high, minimizing the frequency regulation cost should be prioritized. Simultaneously, power output constraints and ramp-up rate constraints need to be considered. For instance, the output of thermal power units cannot exceed their maximum output limit, and their output change rate cannot exceed their ramp-up rate limit. The multi-power coordinated frequency regulation control model is based on rolling optimization and model predictive control methods. It obtains the optimal frequency regulation control strategy for each power source by solving the optimization problem based on the objective function and constraints, including AGC (Automatic Generation Control) commands and primary frequency regulation power allocation. For example, the model might issue a command requiring thermal power units to increase output by 20MW and hydropower units to decrease output by 10MW.After receiving the optimal frequency regulation control strategy, each power source control unit (thermal power, hydropower, wind power, photovoltaic, etc.) adjusts its output according to the instructions. For example, the thermal power unit control unit increases fuel supply and output according to the AGC instruction; the hydropower unit control unit adjusts turbine opening and output according to the instruction. Rolling optimization and feedforward correction mechanisms enable the multi-power source coordinated frequency regulation control model to adaptively adjust the frequency regulation control strategy. For example, when a sudden increase in load is detected, the model re-optimizes based on the latest grid operation data, generating new control instructions to quickly restore grid frequency stability. Through continuous optimization, the grid frequency can be kept stable near its rated value for a long period.

[0036] S107. To address the impact of communication delay and control time lag on the performance of multi-power supply cooperative control, a predictive control strategy based on Kalman filtering and an adaptive control algorithm are adopted to reduce the impact of communication delay and control time lag on system performance. Based on real-time communication status and control effect, the communication network topology and control parameters are optimized to improve the real-time performance and reliability of multi-power supply cooperative control.

[0037] The system acquires real-time communication status and control performance data of the multi-power supply system. Based on preset delay and time-delay thresholds, it determines whether the data exceeds allowable limits; if so, it triggers an optimization process. A Kalman filter algorithm is used to estimate and predict the state of the multi-power supply system, obtaining predicted state values ​​for a future period. These predicted state values ​​are used as input for coordinated control, enabling advance control decisions and execution to reduce the impact of communication delays and control time-delays. Based on the real-time and reliability requirements of multi-power supply coordinated control, the optimization objective function for the star topology and PID control parameters is determined. A particle swarm optimization algorithm is used to adjust and optimize the number of nodes, connection method, and PID control parameters of the star topology online, minimizing the objective function. The optimized star topology and PID control parameters are acquired, and the configuration of the multi-power supply coordinated control system is updated to improve the system's real-time performance and reliability. The communication status and control performance of the multi-power supply system are continuously monitored. If communication delays or control time-delays exceed thresholds again, the optimization process is re-triggered, forming a closed-loop control. A support vector machine algorithm is used to train historical communication status data, control performance data, and optimization parameters to establish a predictive model. A predictive model is used to autonomously optimize the star topology and PID control parameters without manual intervention. Based on the optimized star topology and PID control parameters, the control cycle and control gain of the multi-power source cooperative control system are dynamically adjusted to ensure efficient and reliable cooperative control even with communication delays and control time lags. When communication delays and control time lags decrease, the control cycle is shortened and the control gain is reduced accordingly; when communication delays and control time lags increase, the control cycle is lengthened and the control gain is increased accordingly. By adaptively adjusting the control cycle and control gain, the stability and dynamic performance of the multi-power source cooperative control system are maintained.

[0038] For example, in multi-power source coordinated frequency regulation control, communication delay and control lag can affect the control effect. To address this issue, a series of optimization measures can be taken. First, it is necessary to acquire real-time communication status data and control effect data of the multi-power source system, such as the output data of each power source, frequency deviation data, communication delay time, and control command execution time. Thresholds for delay and lag should be set, for example, communication delay not exceeding 10ms and control lag not exceeding 50ms. If the monitored data exceeds the threshold, the optimization process is triggered. The Kalman filter algorithm can be used to estimate and predict the state of the multi-power source system. For example, based on historical frequency deviation data, power source output data, and load change data, the system frequency deviation over a future period can be predicted. Assuming the current system frequency deviation is -0.1Hz, it is predicted that the frequency deviation will decrease to -0.15Hz within the next minute. This predicted value will serve as the input for coordinated control, allowing for advance adjustment of the output of each power source, thereby reducing the impact of communication delay and control lag. For example, the output of hydropower units can be increased in advance to cope with the predicted frequency decrease. To improve the real-time performance and reliability of multi-power source coordinated control, the star topology and PID control parameters need to be optimized. A star topology refers to a network topology centered on a central node and connecting multiple power supply nodes. PID control parameters include proportional, integral, and derivative coefficients. The optimization objective is to enable the system to respond to frequency deviations as quickly as possible while meeting stability requirements and keeping the frequency deviations within acceptable limits. For example, the objective function can be set as a weighted sum of the squares of the frequency deviations and the squares of the changes in control commands. Particle swarm optimization (PSO) can be used to find the optimal number of nodes, connection method, and PID control parameters for the star topology. Assume that PSO determines the optimal number of nodes to be 5, the connection method to be fully connected, and the PID control parameters to be a proportional coefficient of 0.8, an integral coefficient of 0.1, and a derivative coefficient of 0.05. After obtaining the optimized parameters, the configuration of the multi-power supply cooperative control system needs to be updated. For example, the communication interfaces of the central node are set to 5, and the PID controller parameters are set to the optimized values. The communication status and control effect of the system are continuously monitored. If the communication delay or control lag exceeds the threshold again, the optimization process is retried to form closed-loop control. To achieve autonomous optimization, a support vector machine (SVM) algorithm can be used to train a predictive model on historical data. For example, historical communication status data, control effect data, and optimization parameters can be used as input, with the optimized star topology and PID control parameters as output, to train an SVM model. Using this model, the optimal star topology and PID control parameters can be predicted based on the current system state without manual intervention. Assuming the predictive model predicts optimal PID parameters of a proportional gain of 0.9, an integral gain of 0.12, and a derivative gain of 0.06, the system will automatically update the PID controller parameters.Based on the optimized star topology and PID control parameters, the control cycle and control gain are dynamically adjusted. The control cycle refers to the time interval during which the controller performs one control operation. The control gain refers to the controller's sensitivity to frequency deviations. When communication delay and control lag are large, the control cycle and control gain need to be increased to ensure control effectiveness. Conversely, the control cycle and control gain need to be decreased to improve control efficiency. For example, when the communication delay increases to 15ms, the control cycle is adjusted from 10ms to 20ms, and the control gain is adjusted from 1.0 to 1.2. By adaptively adjusting the control cycle and control gain, the system can maintain stability and dynamic performance under different communication delays and control lags.

[0039] S108. Based on the real-time and reliability requirements of multi-power source cooperative control, the particle swarm optimization algorithm is used to adjust and optimize the number of nodes, connection method, and PID control parameters of the star topology online, obtain the optimized star topology and PID control parameters, dynamically adjust the control cycle and control gain of the multi-power source cooperative control system, and maintain the stability and dynamic performance of the multi-power source cooperative control system by adaptively adjusting the control cycle and control gain.

[0040] To meet the real-time and reliability requirements of the multi-power supply cooperative control system, current system operating status data, including power supply status and load demand, is acquired and used as input to the particle swarm optimization (PSO) algorithm. The standard PSO algorithm is employed, using the number of nodes in the star topology, connection method, and the proportional, integral, and derivative parameters of the PID controller as optimization variables to establish an optimization model. A fitness function is set, considering the system's stability margin and dynamic response indices, such as overshoot and settling time. Through iterative optimization, the optimal combination of node number, connection method, and PID control parameters is searched to obtain the optimized star topology and control parameters. The optimization results are applied to the multi-power supply cooperative control system. A star communication network is constructed based on the optimized number of nodes and connection method, and the PID controller parameters are set to optimized values. Based on the optimization results, a fuzzy control method is used to dynamically adjust the control cycle and control gain. A fuzzy rule base is established between the control cycle and gain and the system's stability margin and dynamic performance indices. Based on the system's real-time operating status, the control cycle and gain values ​​are dynamically calculated through fuzzy inference to adaptively maintain the system's stability and dynamic performance. During the adjustment process, the system's operating status is continuously monitored, acquiring data such as power status and load demand to determine whether the system's stability margin and dynamic performance indicators meet preset threshold requirements. If the system performance indicators fail to meet requirements for multiple consecutive control cycles, the current status data is re-input into the particle swarm optimization algorithm for further optimization and adjustment. A maximum limit is set for the number of iterative optimizations and dynamic adjustments. If the system performance indicators still fail to meet requirements when the maximum number of adjustments is reached, an alarm is issued, requiring manual intervention. If the system performance indicators continuously meet requirements, the optimization and adjustment are considered to have reached stability, the iterative optimization and dynamic adjustment process is stopped, and the system transitions to regular monitoring mode.

[0041] For example, real-time performance and reliability are crucial for multi-power supply cooperative control systems. To achieve this, a control strategy capable of dynamically adjusting itself based on the system's operating status is needed. Here, a particle swarm optimization algorithm combined with fuzzy control is employed to achieve this goal. First, real-time operating status data of the system needs to be acquired. This data includes the output power, voltage, and current of each power supply, as well as the load demand of the entire system, such as total active and reactive power. For instance, suppose a system consists of three power supplies: power supply A outputs 10MW, power supply B outputs 15MW, and power supply C outputs 5MW, with a current total system load demand of 28MW. This data will serve as input to the particle swarm optimization algorithm. The goal of the particle swarm optimization algorithm is to find the optimal combination of a star topology and PID control parameters. A star topology refers to a network topology centered on a central control node, connecting all power supply nodes. The connection method can be fully connected or partially connected. The parameters of the PID controller include proportional, integral, and derivative coefficients. These parameters determine the controller's response speed and stability to system deviations. For example, a possible optimization objective is to minimize frequency deviation and control energy consumption while meeting system stability requirements. The design of the fitness function is crucial, as it reflects the optimization objective. Typically, the fitness function considers the system's stability margin and dynamic response metrics, such as overshoot and settling time. Assuming the system requires a stability margin greater than 0.1, overshoot less than 5%, and settling time less than 1 second, a fitness function can be designed that multiplies these metrics by their weights and then sums the results. Through iterative search using the particle swarm optimization algorithm, the optimal number of nodes, connection method, and PID control parameter combination for the star topology can be found. For example, the final optimization result might be a fully connected star topology with 5 nodes, and PID parameters of proportional gain 0.8, integral gain 0.1, and derivative gain 0.05. After obtaining the optimization result, it needs to be applied to the actual system. First, the star communication network is reconstructed based on the optimized number of nodes and connection method. Then, the PID controller parameters for each power supply are set to the optimized values. For example, the proportional gain of all power supply PID controllers is set to 0.8, the integral gain to 0.1, and the derivative gain to 0.05. To further improve the system's adaptability, a fuzzy control method is used to dynamically adjust the control cycle and control gain. The core of fuzzy control is the fuzzy rule base, which describes the relationship between the control period and gain and the system's stability margin and dynamic performance indicators. For example, a fuzzy rule might be: if the system has a low stability margin and a slow dynamic response, increase the control gain and shorten the control period. Assume the system's stability margin is 0.08, less than the preset threshold of 0.1, and the dynamic response time is 1.2 seconds, greater than the preset threshold of 1 second. Based on the fuzzy rule base, the fuzzy inference engine will calculate the appropriate control period and gain adjustment values.For example, the control cycle can be shortened from 10ms to 8ms, and the control gain increased from 1.0 to 1.2. During the adjustment process, the system's operating status needs continuous monitoring. If the system performance indicators fail to meet requirements for several consecutive control cycles, optimization needs to be repeated. For example, if the system stability margin is below 0.1 for three consecutive cycles, the current state data is re-input into the particle swarm optimization algorithm to search for optimal parameters again. To avoid infinite loops, a maximum limit needs to be set for the number of iterative optimizations and dynamic adjustments. For example, the maximum number of iterations can be set to 10. If the system performance still fails to meet the requirements after reaching the maximum number of iterations, an alarm is issued, requiring manual intervention. If the system performance indicators continuously meet the requirements, the system is considered to have reached a stable state, and iterative optimization and dynamic adjustments cease.

[0042] S109. Based on the communication optimization results, in the design of the multi-power source coordinated frequency modulation control strategy, considering both frequency modulation performance and the ease of engineering implementation, a combination of simulation testing and actual engineering verification is adopted to optimize the control strategy parameters and implementation scheme. By monitoring the operating status and performance indicators of the multi-power source coordinated frequency modulation control system in real time, the control strategy is continuously optimized to ensure its reliability and effectiveness.

[0043] Based on the communication optimization results, the network topology and communication link quality parameters of the multi-power source coordinated frequency modulation control system are obtained. The optimization objectives and constraints of the control strategy are determined, such as frequency modulation response time, frequency modulation accuracy, and communication delay. Multiple alternative control strategy parameter combinations are designed, including parameters such as frequency modulation power allocation ratio, frequency modulation dead zone, and frequency modulation rate. For each control strategy parameter combination, a simulation model of the multi-power source coordinated frequency modulation control system is built using a simulation test platform, such as MATLAB / Simulink, to simulate the system's operation under different operating conditions and obtain simulation results for various frequency modulation performance indicators. The simulation test results are analyzed to evaluate the frequency modulation performance of each control strategy parameter combination and determine whether it meets the optimization objectives and constraints. Parameter combinations that meet the requirements are added to the candidate set, while combinations that do not meet the requirements are eliminated. Among the candidate control strategy parameter combinations, considering both the implementation difficulty and frequency regulation performance, and combining engineering verification results, the multi-objective optimization algorithm NSGA-II is adopted. With frequency regulation response time and frequency regulation accuracy as optimization objectives, a Pareto optimal solution set is generated. The optimal parameter combination balancing difficulty and performance is selected as the final parameter configuration and implementation scheme for the multi-power source coordinated frequency regulation control strategy. The optimized multi-power source coordinated frequency regulation control strategy is applied to coordinate the frequency regulation power allocation and control command issuance among multiple power sources, achieving rapid and accurate adjustment of the grid connection point frequency. A communication network based on the IEC61850 protocol is deployed to monitor key indicators such as the grid connection point frequency and active power of each power source in real time. When the frequency deviation exceeds ±2Hz or the active power fluctuation rate exceeds 5%, the re-optimization process of the control strategy is triggered. In the re-optimization process, based on the real-time operating conditions of the power grid and the latest communication optimization results, the strategy gradient REINFORCE algorithm is used for online learning. Through continuous trial and error and strategy iteration, the multi-power source coordinated frequency regulation control strategy is continuously optimized, improving the robustness and dynamic performance of the system and achieving adaptive updates of the frequency regulation strategy.

[0044] For example, the results of communication optimization provide the network topology and communication link quality parameters of the multi-power source coordinated frequency modulation control system. For instance, the topology might be star, ring, or mesh, and the communication link quality parameters include bandwidth, latency, and packet loss rate. These parameters influence the formulation of the control strategy. The goal of control strategy optimization is to improve frequency modulation performance, such as reducing the frequency modulation response time to within 200ms, controlling the frequency modulation accuracy to within 0.1Hz, and ensuring communication latency is less than 50ms. Constraints include ensuring system stability and considering communication resource limitations. Multiple alternative combinations of control strategy parameters can be designed. For example, combination 1: the frequency modulation power allocation ratio is hydropower:thermal power:wind power = 4:3:3, the frequency modulation dead zone is 0.02Hz, and the frequency modulation rate is 1MW / s; combination 2: the frequency modulation power allocation ratio is 3:4:3, the frequency modulation dead zone is 0.01Hz, and the frequency modulation rate is 2MW / s, etc. These parameter combinations will be evaluated in subsequent simulations. A simulation model of the multi-power source coordinated frequency modulation control system is built using MATLAB / Simulink. The model includes dynamic models of each power source, a communication network model, and a control algorithm module. It simulates the system's operation under different conditions, such as load step changes and power output fluctuations. Simulation results for various frequency regulation performance indicators are recorded and analyzed, such as frequency regulation response time, frequency regulation accuracy, and communication delay. It is assumed that combination 1 has a frequency regulation response time of 180ms, a frequency regulation accuracy of 0.08Hz, and a communication delay of 40ms, satisfying the optimization objective and constraints; while combination 2 has a frequency regulation response time of 150ms, but a frequency regulation accuracy of 0.12Hz, exceeding the set target, so combination 2 is eliminated. Among the candidate control strategy parameter combinations, a comprehensive evaluation is performed based on engineering verification results, such as test data from actual power grids. It is assumed that combination 1 performs stably in actual tests and is relatively easy to implement, while other candidate combinations offer limited performance improvement but are more difficult to implement. The multi-objective optimization algorithm NSGA-II is used, with frequency regulation response time and frequency regulation accuracy as optimization objectives, to generate a Pareto optimal solution set. For example, one solution in the solution set has a frequency modulation response time of 170ms and a frequency modulation accuracy of 0.09Hz; another solution has a frequency modulation response time of 190ms and a frequency modulation accuracy of 0.07Hz. Ultimately, combination 1 with a frequency modulation response time of 180ms and a frequency modulation accuracy of 0.08Hz is chosen as the final parameter configuration scheme because it balances implementation difficulty and frequency modulation performance. An optimized multi-power source coordinated frequency modulation control strategy is applied. When the system detects a frequency deviation exceeding ±2Hz or an active power fluctuation exceeding 5%, a re-optimization process is triggered. For example, a sudden increase in load at a certain moment causes the frequency to drop to 49.8Hz, triggering the re-optimization process. In the re-optimization process, the policy gradient REINFORCE algorithm is used for online learning. For example, the algorithm tries different frequency modulation power allocation ratios and rewards or penalizes based on changes in frequency deviation.Through continuous trial and error and strategy iteration, the algorithm gradually learns the optimal control strategy, such as increasing the frequency regulation power allocation ratio of hydropower to 50% to restore the frequency more quickly. Continuous optimization of the multi-power source coordinated frequency regulation control strategy improves the system's robustness and dynamic performance, enabling adaptive updates of the frequency regulation strategy.

[0045] S1010. Based on the real-time operating conditions of the power grid and the latest communication optimization results, the strategy gradient REINFORCE algorithm is used for online learning. Through continuous trial and error and strategy iteration, a continuously optimized multi-power source coordinated frequency regulation control strategy is obtained, the frequency regulation power allocation and control command issuance scheme is determined, and the grid connection point frequency is quickly and accurately adjusted.

[0046] Real-time power grid operation status data and communication optimization results are acquired and used as input to the reinforcement learning algorithm to initialize the policy network parameters. The policy network adopts a multilayer perceptron structure, with power grid state features as input and probability distribution of multi-source coordinated frequency regulation control strategies as output. Based on the current policy network parameters, a batch of multi-source coordinated frequency regulation control strategies are sampled and generated. Each strategy includes information such as frequency regulation power allocation and control command issuance time for each power source. These strategies are tested in a power grid simulation environment to obtain evaluation indicators such as grid connection point frequency deviation and frequency regulation response time, and the reward value of each strategy is calculated. The reward value considers factors such as the sum of absolute values ​​of frequency deviations and the weighted average of frequency regulation response times. The policy network parameters are updated using the policy gradient method. Based on the reward value and gradient estimate of each strategy, the policy gradient is calculated, and the weights of the policy network are updated using the gradient ascent method to obtain the optimized policy network. Steps 2-4 are repeated for multiple rounds of policy iterative optimization until the policy network converges or reaches the preset number of iterations. The convergence criterion is that the variance of the policy reward value is less than a preset threshold. The optimal multi-source coordinated frequency regulation control strategy is generated by sampling from the converged strategy network and applied to the actual operation environment of the power grid. The actual control effect of the strategy is evaluated by monitoring indicators such as the frequency deviation at the grid connection point. If the actual effect does not meet expectations, i.e., the frequency deviation exceeds the allowable range or the frequency regulation time is too long, the real-time operating status data of the power grid and the evaluation results are fed back to the reinforcement learning model, triggering a new round of online learning and strategy optimization. Steps 1-6 are repeated to continuously adapt the control strategy to changes in the power grid operating conditions.

[0047] For example, reinforcement learning is a machine learning method that allows an agent to learn optimal behavior by interacting with its environment. The agent receives rewards or penalties to guide its learning process. In this example, reinforcement learning is used to optimize a multi-source coordinated frequency regulation control strategy. First, real-time grid operating status data and communication optimization results are acquired. For example, real-time grid operating status data includes the current frequency, load power, and output of each power source. Communication optimization results include network topology, communication bandwidth, latency, and packet loss rate. This data will serve as input to the reinforcement learning algorithm. Simultaneously, the policy network parameters are initialized. The policy network employs a multilayer perceptron structure, with grid state characteristics as input and a probability distribution of the multi-source coordinated frequency regulation control strategy as output. For example, the input to the policy network could be frequency deviation, load change rate, and power source output. The output could be the frequency regulation power allocation ratio of different power sources. Next, based on the current policy network parameters, a batch of multi-source coordinated frequency regulation control strategies is sampled and generated. Each strategy includes information such as the frequency regulation power allocation of each power source and the control command issuance time. For example, one strategy could be a frequency regulation power allocation ratio of hydropower:thermal power:wind power of 4:3:3, with the control command issuance time being the current time plus 5 milliseconds. Another strategy could be a frequency regulation power allocation ratio of hydropower:thermal power:wind power of 3:4:3, with the control command issuance time being the current time plus 10 milliseconds, and so on. Then, these strategies are tested in a power grid simulation environment. The power grid simulation environment is a software platform that simulates the operation of a real power grid. It can simulate various operating conditions, such as load step changes and power output fluctuations. In the simulation environment, the performance of each strategy can be evaluated. Evaluation indicators such as grid connection point frequency deviation and frequency regulation response time are obtained, and the reward value for each strategy is calculated. The reward value considers factors such as the sum of the absolute values ​​of frequency deviations and the weighted average of frequency regulation response times. For example, a strategy that results in a smaller frequency deviation and a shorter frequency regulation response time will have a higher reward value. Conversely, if a strategy results in a larger frequency deviation and a longer frequency regulation response time, its reward value will be lower. Afterwards, the strategy network parameters are updated using the strategy gradient method. The policy gradient method is a gradient-based optimization algorithm. It calculates the policy gradient based on the reward value and gradient estimate of each policy, and updates the weights of the policy network using gradient ascent, resulting in an optimized policy network. This is done to enable the policy network to generate better policies. The policy iterative optimization process is repeated multiple times until the policy network converges or reaches a preset number of iterations. Convergence is determined by the variance of the policy reward value being less than a preset threshold. This means the policy network has learned a stable policy. The optimal multi-source coordinated frequency regulation control policy is sampled from the converged policy network and applied to the actual operation environment of the power grid. The actual control effect of the policy is evaluated by monitoring indicators such as the frequency deviation at the grid connection point.If the actual effect does not meet expectations—that is, the frequency deviation exceeds the allowable range or the frequency regulation time is too long—the real-time operating status data of the power grid and the evaluation results are fed back to the reinforcement learning model, triggering a new round of online learning and strategy optimization. This is a continuous learning and optimization process to ensure that the control strategy can adapt to changes in power grid operating conditions. For example, suppose the load on the power grid suddenly increases at a certain moment, causing the frequency to drop. At this time, the reinforcement learning model will generate a new frequency regulation control strategy based on the current power grid state. This strategy might increase the output of hydropower to quickly restore the frequency. If the actual control effect of this strategy is good, the reinforcement learning model will retain the strategy. If the actual control effect is poor, the reinforcement learning model will relearn and optimize, generating a new strategy. This process is repeated until an optimal control strategy that can adapt to various changes in operating conditions is obtained.

[0048] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A multi-power source coordinated frequency modulation control method, characterized in that, include: Acquire real-time power grid operation data, which includes system frequency deviation, load forecast information, and real-time output and regulation capacity data of multiple power sources such as wind power, photovoltaic power, hydropower and thermal power. Based on the real-time operation data of the power grid, the optimal power allocation scheme of multiple power sources is calculated using a mixed integer linear programming algorithm, with the goal of minimizing power generation cost and the constraints of upper and lower limits of power output of multiple power sources, ramp rate and system frequency deviation limit. The optimal power output allocation scheme for multiple power sources also includes addressing the volatility and prediction difficulty of wind and solar power generation by employing a short-term power prediction algorithm based on support vector machines and a real-time power smoothing control strategy based on moving averages to reduce the impact of wind and solar power fluctuations on grid frequency. To address the limitation of hydropower regulation capacity by reservoir capacity and downstream ecological flow, a dynamic programming reservoir optimization scheduling model is adopted to meet grid frequency regulation requirements while considering the long-term operational benefits of reservoirs and ecological environmental protection. For thermal power, which has a wide regulation range but relatively slow response speed, deep peak shaving technology for thermal power units is adopted to expand the regulation range and improve the response speed. Based on the regulation characteristics of thermal power units, the load allocation and reserve capacity of thermal power units are optimized to improve the ability and reliability of thermal power units to participate in system frequency regulation. The optimal power output allocation scheme is converted into a power output adjustment command and sent to each power control system through the power grid dispatch automation system. Each power control unit dynamically adjusts its output using a PID control algorithm based on the issued instructions and its own optimization results, thereby achieving multi-power coordinated frequency regulation. Real-time monitoring of grid frequency determines the effectiveness of multi-source coordinated frequency regulation. Through online learning and strategy iteration, the multi-source coordinated frequency regulation control strategy is continuously optimized to improve the effectiveness of grid frequency control.

2. The multi-power source coordinated frequency modulation control method as described in claim 1, characterized in that, After acquiring the real-time operation data of the power grid, the method further includes: The real-time operation data of the power grid is preprocessed and features are extracted to calculate the average system frequency deviation and key indicators of load forecasting error, and the multi-source data is converted into a standardized input format.

3. The multi-power source coordinated frequency modulation control method as described in claim 1, characterized in that, Before calculating the optimal power output allocation scheme for multiple power sources based on real-time power grid operation data, the following steps are also included: The decision tree algorithm is used to classify the power grid operation status and determine whether the current status meets the requirements of frequency stability and economic operation. If not, the multi-source output optimization and adjustment process is triggered.

4. The multi-power source coordinated frequency modulation control method as described in claim 1, characterized in that, After converting the optimal power output allocation scheme into a power output adjustment command, the method further includes: During the power output adjustment process, the power grid operation status and system frequency changes are monitored in real time using power simulation software. If the frequency deviation exceeds the preset threshold, an emergency frequency regulation strategy based on fuzzy PID control is triggered. By adjusting the power output and load limiting measures, the system frequency stability is quickly restored.

5. The multi-power source coordinated frequency modulation control method as described in claim 1, characterized in that, The multi-power source coordinated frequency modulation control method further includes: using a support vector machine algorithm to train historical communication status data, control effect data, and optimization parameters to establish a prediction model, and using the prediction model to achieve autonomous optimization of the star topology and PID control parameters without manual intervention; An optimization model was established using the standard particle swarm optimization algorithm, with star topology parameters and PID controller parameters as optimization variables. Based on the optimization results, a fuzzy control method is adopted to dynamically adjust the control cycle and control gain, so as to ensure that efficient and reliable collaborative control can still be achieved even in the presence of communication delay and control time delay. When the communication delay and control time delay decrease, the control cycle is shortened and the control gain is reduced accordingly. When the communication delay and control time delay increase, the control cycle is extended and the control gain is increased accordingly. If the system performance indicators fail to meet the requirements for several consecutive control cycles, the current state data will be re-input into the particle swarm optimization algorithm for further optimization and adjustment.

6. The multi-power source coordinated frequency modulation control method as described in claim 1, characterized in that, The system acquires real-time power grid operation data, including system frequency deviation and real-time output data of each power source, and simultaneously acquires power regulation capacity data and load forecast results data. The system performs data cleaning and feature extraction, and generates comprehensive evaluation data through data fusion technology. The comprehensive evaluation data is input into the Q-learning reinforcement learning model. The Q-value table is updated iteratively, and the power grid operation optimization model of the global coordination control layer is updated periodically. Through the power grid operation optimization model, the multi-objective optimization function is constructed by comprehensively considering the multiple objectives of power grid safety, economy, and low carbon, so as to determine the current power supply operation status and obtain the power supply operation status evaluation result. Based on the power supply operation status assessment results, the optimization objective function and constraints of multi-power supply coordinated frequency regulation control are determined. By solving the optimization problem, the optimal frequency regulation control strategy of each power supply is obtained. The optimal frequency regulation control strategy includes AGC commands and primary frequency regulation power allocation. The optimal frequency regulation control strategy is sent to each power supply control unit to control each power supply to adjust its output according to the frequency regulation command, so as to achieve rapid recovery and stabilization of the grid frequency.

7. The multi-power source coordinated frequency modulation control method as described in claim 1, characterized in that: When continuously optimizing the multi-power source coordinated frequency modulation control strategy, the strategy gradient REINFORCE algorithm is used for online learning. Through continuous trial and error and strategy iteration, the continuously optimized multi-power source coordinated frequency modulation control strategy is obtained, the frequency modulation power allocation and control command issuance scheme is determined, and the grid connection point frequency is quickly and accurately adjusted.

8. The multi-power source coordinated frequency modulation control method according to any one of claims 1-7, characterized in that: The power grid dispatch automation system and each power control system use a communication network based on the IEC61850 protocol to realize the real-time transmission and interaction of power grid operation information and control commands.

Images