Load optimization and emergency support control method, system and device for energy storage control system

Abstract

The invention discloses an energy storage control system load optimization and emergency support control method, system and device. The method comprises the following steps: firstly, constructing a multi-level water head sensor network to monitor water head data in real time, and dynamically determining power generation duration based on a hydraulic model; thirdly, combining the power generation duration capacity and the unit efficiency characteristic curve, and using an optimization algorithm to obtain an optimal load distribution scheme; then, a start control strategy is constructed, unit start-stop and load regulation unified management is achieved through an automatic control algorithm of a computer monitoring platform, and the system operation state is obtained; according to the state, a support scheme is determined through a capacity calculation model, and different requirements of a power grid can be flexibly met; when the power grid is abnormal, a corresponding scheme is called and executed in time, normal operation of the power grid can be recovered quickly, safety and stability are guaranteed, and influences on users are reduced. The method establishes a perfect emergency response mechanism, improves the emergency processing capability of the power system, and effectively improves the water energy utilization efficiency and the power grid stability.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent control technology for hydropower stations, and in particular to a load optimization and emergency support control method, system, and device for an energy storage control system. Background Technology

[0002] With hydropower serving as a crucial clean energy source in power systems, its efficient utilization has always been a goal pursued by the industry. Traditional power generation control methods often struggle to adequately adapt to complex and ever-changing actual operating conditions. For instance, during hydropower generation, changes in water head significantly impact power generation capacity, but traditional control methods cannot incorporate real-time water head monitoring to dynamically calculate the duration of power generation or pumping, resulting in inaccurate and inefficient utilization of hydropower resources. Furthermore, the lack of efficiency-prioritized algorithms in unit load allocation prevents units from operating at their optimal state, further reducing hydropower utilization.

[0003] While current automatic generation systems have achieved load distribution and start-up / shutdown control for generating units to a certain extent, they still have many shortcomings. Existing automatic generation systems mostly operate in relatively fixed modes and cannot be well integrated into computer monitoring systems to achieve flexible regulation. When the power plant is used as the control unit, the response to total active power load commands is not intelligent enough, and it cannot accurately calculate and distribute the unit load according to the actual situation, making it difficult to fully meet the diverse needs of the power grid for power plant load. When facing special situations in the power grid, such as large power loss or severe system frequency anomalies, the existing systems have limited capabilities in rapid response and providing emergency power support, and cannot restore the power grid frequency in a timely and effective manner. Summary of the Invention

[0004] In view of this, the present disclosure provides a method, system, and apparatus for load optimization and emergency support control of an energy storage control system, which can solve the problems of traditional power generation control being difficult to adapt to operating conditions, inflexible regulation of automatic power generation systems, and weak emergency support capabilities in the prior art.

[0005] In a first aspect, embodiments of this disclosure provide a method for load optimization and emergency support control of an energy storage control system, including: The head data is monitored in real time by constructing a multi-level head sensor network, and the power generation duration and capacity are dynamically determined based on the hydraulic model. Based on the power generation duration and capacity and the efficiency characteristic curves of each unit, the optimal load allocation scheme is obtained through an optimization algorithm. Based on the optimal load allocation scheme, a startup control strategy is constructed; According to the aforementioned startup control strategy, the automatic control algorithm of the computer monitoring platform is used to achieve unified management of unit startup and shutdown and load regulation, thereby obtaining the operating status of the unified management system. Based on the operational status of the unified management system, a support plan for the hydropower station is determined using a capacity calculation model; When the power grid experiences an abnormal state, the corresponding support scheme is invoked and executed.

[0006] Secondly, this disclosure also provides a load optimization and emergency support control system for an energy storage control system, including: The power generation duration and capacity acquisition module is used to monitor head data in real time based on the constructed multi-level head sensor network and dynamically determine the power generation duration and capacity based on the hydraulic model. The optimal load allocation scheme determination module is used to obtain the optimal load allocation scheme based on the power generation duration capacity and the efficiency characteristic curves of each unit through an optimization algorithm. A startup control strategy construction module is used to construct a startup control strategy based on the optimal load allocation scheme. The unified management system operation status acquisition module is used to obtain the unified management system operation status by realizing the unified management of unit start-up and shutdown and load regulation through the automatic control algorithm of the computer monitoring platform according to the startup control strategy. The support scheme acquisition module is used to determine the support scheme for the hydropower station based on the operating status of the unified management system and through the capacity calculation model. The trigger module is used to invoke and execute the corresponding support scheme when an abnormal state occurs in the power grid.

[0007] Thirdly, this disclosure also provides a computer device, which adopts the following technical solution: The computer device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform any of the above-described energy storage control system load optimization and emergency support control methods.

[0008] Fourthly, embodiments of this disclosure also provide a computer-readable storage medium storing computer instructions for causing a computer to execute any of the above-described energy storage control system load optimization and emergency support control methods.

[0009] Fifthly, embodiments of this disclosure also provide a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of any of the methods described above.

[0010] The energy storage control system load optimization and emergency support control method disclosed in this application firstly monitors head data in real time based on a constructed multi-level head sensor network. This multi-level head sensor network can acquire head data accurately and in real time, providing a reliable basis for calculating power generation duration and capacity. Then, it dynamically determines the power generation duration and capacity based on a hydraulic model. Based on the power generation duration and capacity and the efficiency characteristic curves of each unit, an optimization algorithm obtains the optimal load allocation scheme. Next, based on the optimal load allocation scheme, a start-up control strategy is constructed. According to the start-up control strategy, the automatic control algorithm of the computer monitoring platform achieves unified management of unit start-up and shutdown and load regulation, thus obtaining unified management system operation. Based on the operational status of the unified management system, the support plan for the hydropower station is determined through a capacity calculation model. This dynamic determination of the support plan allows the hydropower station to flexibly adjust its support strategy according to actual conditions, better meeting the diverse needs of the power grid. When an abnormal state occurs in the power grid, the corresponding support plan is invoked and executed. Timely invocation of the corresponding support plan can quickly restore the normal operation of the power grid, ensuring its safety and stability and reducing the impact on power users. This method establishes a comprehensive emergency response mechanism, enabling the hydropower station to react quickly to power grid anomalies and improving the emergency response capability of the entire power system. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A schematic flowchart illustrating the load optimization and emergency support control method for an energy storage control system provided in this embodiment.

[0013] Figure 2 This is a flowchart illustrating the method for obtaining power generation duration and capacity provided in an embodiment of the present disclosure.

[0014] Figure 3 This is a flowchart illustrating the method for constructing a startup control strategy provided in an embodiment of the present disclosure.

[0015] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure. Detailed Implementation

[0016] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0017] Reference Figure 1This application discloses a load optimization and emergency support control method for an energy storage control system, including: S100 monitors head data in real time based on a multi-level head sensor network and dynamically determines power generation duration and capacity based on a hydraulic model. S200 uses an optimization algorithm to obtain the optimal load allocation scheme based on the power generation duration, capacity, and efficiency characteristic curves of each unit. S300, based on the optimal load allocation scheme, constructs a startup control strategy; S400, based on the startup control strategy, uses the automatic control algorithm of the computer monitoring platform to achieve unified management of unit startup and shutdown and load regulation, and obtain the unified management system operation status; S500 determines the support scheme for hydropower stations based on the operating status of the unified management system and through a capacity calculation model. When an abnormal state occurs in the power grid, the S600 invokes and executes the corresponding support scheme.

[0018] The energy storage control system load optimization and emergency support control method disclosed in this application improves the power generation efficiency of hydropower stations, reduces power generation costs, and increases power generation benefits through real-time monitoring, optimized load allocation, and reasonable control of unit start-up and shutdown. Accurate power generation duration and capacity calculation, optimized load allocation, and effective support schemes enable hydropower stations to better adapt to grid demands, providing a stable power supply during normal grid operation and timely support during grid anomalies, thus enhancing grid stability and reliability. The entire scheme achieves automated and intelligent management of hydropower stations, reducing manual intervention and improving management efficiency and accuracy. Simultaneously, through the analysis and optimization of operational data, operational strategies are continuously improved, enhancing the overall management level of hydropower stations. This method can rationally utilize water resources, improve power generation efficiency, reduce environmental impact, and meet the requirements of sustainable development. Furthermore, enhancing grid stability also helps promote the large-scale integration and consumption of clean energy, facilitating the optimization of the energy structure.

[0019] Reference Figure 2 For S100, "the method for obtaining the power generation capacity includes: real-time monitoring of head data based on a constructed multi-level head sensor network and dynamic determination of power generation duration based on a hydraulic model." S110 involves setting up high-precision pressure level gauges, ultrasonic level gauges, and radar level gauges at monitoring points in the upstream and downstream reservoir areas of the hydropower station in a gradient distribution, forming a multi-level head sensor network.

[0020] Among them, the high-precision pressure level gauge has an accuracy at the millimeter level; the ultrasonic level gauge has a measurement accuracy of ±0.5cm; the radar level gauge preferably adopts the non-contact radar level gauge, which has the advantage of not being affected by water quality, water temperature and other factors, and the measurement accuracy requirement is ±0.3cm. The range can be selected according to actual needs, such as 0-100 meters.

[0021] Specifically, in the upstream reservoir area, sensors are deployed at different elevations and flow cross-sections according to the reservoir's topography and flow characteristics, arranged in a gradient. For example, sensors are installed at the reservoir entrance, middle section, and near the dam to comprehensively monitor upstream water level changes. In the downstream reservoir area, sensors are similarly distributed in a gradient, placed at different distances and in different flow areas downstream of the dam, such as near the stilling basin and at different cross-sections of the downstream river channel, to ensure accurate acquisition of downstream water level information.

[0022] In this embodiment, pressure level gauges, ultrasonic level gauges, and radar level gauges are installed simultaneously at each monitoring point to achieve redundant data acquisition. When one sensor malfunctions or its data is abnormal, data from other sensors can be used for supplementation and verification. Each sensor is equipped with a wireless transmission module, such as a LoRa wireless module, which features low power consumption and long-distance transmission (up to several kilometers), enabling the real-time transmission of sensor-collected data to a data aggregation center. A receiving base station is set up at the data aggregation center to receive data from each sensor and perform preliminary data processing and storage.

[0023] S120 uses a multi-level head sensor network to monitor head data in real time and obtain a head dataset.

[0024] Specifically, the sampling frequency of all sensors is uniformly set to 1Hz, that is, water level data is collected once per second, to ensure that minute changes in water level can be captured in a timely manner; the collected water level data is sent to the data aggregation center in real time through the wireless transmission module, and the data aggregation center uniformly receives and stores the data from different types of sensors.

[0025] Furthermore, the collected data can be standardized to convert the measurements from different sensors into a unified head data format and output a standardized head dataset.

[0026] S130: Construct a four-dimensional hydraulic model based on the head dataset, which integrates reservoir capacity characteristic curves, dynamic discharge coefficients, seasonal hydrological fluctuations, and meteorological forecast data. Based on the four-dimensional hydraulic model, perform noise suppression and multi-step prediction on the head dataset, and output the net head time series and the predicted head time series.

[0027] The methods for obtaining the reservoir capacity characteristic curve include: establishing an accurate reservoir capacity characteristic curve by topographic surveying of the reservoir area of ​​the hydropower station and statistical analysis of historical water level-reservoir capacity data. Specifically, three-dimensional laser scanning technology can be used to conduct high-precision measurements of the reservoir area topography, and combined with historical water level data, the functional relationship between reservoir capacity V and water level H can be obtained by using numerical fitting methods.

[0028] The methods for obtaining the dynamic discharge coefficient include: establishing a dynamic discharge coefficient model based on the structure of the discharge structure and the water flow characteristics of the hydropower station; monitoring and analyzing discharge data under different operating conditions; and using machine learning algorithms (such as neural networks) to predict and update the discharge coefficient in real time.

[0029] The methods for obtaining seasonal hydrological fluctuations include: analyzing the patterns of water level and flow changes in different seasons by collecting historical hydrological data over many years, and using time series analysis methods (such as the ARIMA model) to model and predict seasonal hydrological fluctuations to obtain the corresponding seasonal hydrological fluctuations.

[0030] The methods for obtaining weather forecast data include: establishing a data interface with the local meteorological department to obtain weather forecast data in real time, including information such as rainfall, temperature, and wind speed.

[0031] Furthermore, meteorological forecast data is fused with head data, reservoir capacity characteristic curves, dynamic discharge coefficients, and other data to construct a four-dimensional hydraulic model. Based on this model, the head dataset is preprocessed to remove obvious outliers and noise. Then, an extended Kalman filter (EPF) algorithm is used for preliminary filtering and state estimation of the head data, leveraging its linearization advantage to quickly track the changing trend of the head data. Next, the output of the EPF is used as the initial state for particle filtering. The particle filtering algorithm further suppresses noise and performs multi-step prediction on the head data. Particle filtering can handle nonlinear and non-Gaussian systems, and through random sampling and resampling of a large number of particles, it improves prediction accuracy. Through continuous iterative updates, the prediction accuracy reaches ±2 cm, outputting both net head time series and predicted head time series.

[0032] Among them, the net head time refers to the head that actually acts on the turbine after deducting the head loss caused by factors such as friction and local resistance in the water diversion system. It reflects the amount of head that the turbine can effectively utilize at the current moment.

[0033] S140, based on the net head time series and predicted head time series, combined with the flow characteristics of multiple units and nonlinear reservoir capacity function, adopts an iterative calculation method optimized by genetic algorithm to dynamically solve the optimal power generation duration under constraints, denoted as power generation duration capacity.

[0034] The method for obtaining the flow characteristics of multiple generating units includes: conducting performance tests on each generating unit of the hydropower station to obtain the unit flow data at different power levels, and plotting the flow characteristic curves of multiple generating units. Q i = f(P i ) ,in Q i Indicates the first i The flow rate of the Taiwanese unit P i Indicates the first i The power of the generator unit. The method for determining the nonlinear reservoir capacity function includes: based on the reservoir capacity characteristic curve (i.e., the reservoir capacity function) above, obtaining the nonlinear reservoir capacity function through mathematical fitting methods. Specifically, the nonlinear reservoir capacity function is the derivative of the reservoir capacity function with respect to the head h.

[0035] Specifically, 1) Determine the objective function: With the goal of maximizing the power generation benefits of the hydropower station, while considering constraints such as ecological flow requirements and power grid dispatch instructions, define the objective function. For example, the objective function can be expressed as the difference between power generation revenue and power generation costs and ecological compensation costs.

[0036] 2) Set constraints; 2.1) Water level constraint: Set minimum and maximum water head to ensure that the water level is within a safe range during power generation; 2.2) Flow constraint: Consider ecological flow requirements and set minimum flow limit to ensure downstream ecological water demand; 2.3) Unit power constraint: Set upper and lower power limits for each unit according to the rated power and operating restrictions of the unit.

[0037] The specific steps of "dynamically solving the optimal power generation duration under constraints using an iterative calculation method optimized by a genetic algorithm" include: 1) Initializing the population: Randomly generating a set of initial solutions as the population, each solution representing a set of power generation duration and unit power allocation schemes; 2) Fitness evaluation: Calculating the fitness value of each solution based on the objective function to evaluate its quality; 3) Selection operation: Using roulette wheel selection or tournament selection, selecting individuals with higher fitness from the current population as parents; 4) Crossover operation: Performing a crossover operation on parent individuals to generate new offspring individuals, increasing the diversity of the population; 5) Mutation operation: Performing a mutation operation on offspring individuals to introduce new genes and avoid the algorithm getting stuck in local optima; 6) Iterative update: Repeating the above steps to iteratively update the population until the termination condition is met (such as reaching the maximum number of iterations or the objective function converging). In each iteration, based on the current power generation duration and unit power allocation scheme, combined with the multi-unit flow characteristics and nonlinear storage capacity function, the optimal power generation duration under constraints is dynamically solved.

[0038] Among them, the optimal power generation duration (i.e., power generation duration capacity) is: : ,in, For the clean water head at time t, Minimum water head, The nonlinear reservoir capacity function represents the rate of change of reservoir capacity with water head. It reflects the change in reservoir capacity caused by a unit change in water head when the water head is h. To indicate that all units are at their respective power... The sum of the following flows, It is the flow characteristic function of the i-th generating unit, which describes the relationship between the unit's flow rate and power. Different generating units have different flow rates under different power levels. By summing the flow rates of each generating unit, the total power generation flow rate of the entire hydropower station under the current power allocation can be obtained.

[0039] The entire integral expression is represented by the head from Change to During the process, considering the reservoir capacity change rate and the total unit flow, the optimal power generation duration is calculated. Furthermore, numerical integration methods are preferred for calculation, such as the trapezoidal integration method and Simpson's integration method. These methods define the integration interval […]. , The function is divided into several small intervals. By approximating the function values ​​in each small interval and summing them, an approximate value of the integral is obtained, thus yielding the optimal power generation duration.

[0040] The S200 method, which "obtains the optimal load allocation scheme based on power generation duration, capacity, and efficiency characteristic curves of each unit through an optimization algorithm," includes: S210, based on power generation duration capacity and net head time series, constructs a unit efficiency model with time-varying head-flow-power three-dimensional mapping; Key points are identified from the unit efficiency model. Based on these key points, the nonlinear efficiency curves in the unit efficiency model are fitted using the Bezier surface fitting algorithm. After introducing correction parameters, a dynamic efficiency prediction model strongly coupled with the power generation duration is obtained.

[0041] Among these methods, key points are determined from the unit efficiency model using either uniform sampling or sampling based on the density of the data distribution. These key points reflect the main characteristics of the efficiency curve. A Bézier surface fitting algorithm is then used to fit the nonlinear efficiency curve based on the selected control points. The order of the Bézier surface can be adjusted according to the complexity of the curve to achieve a better fitting effect. The fitted surface can more accurately describe the nonlinear relationship between unit efficiency and head, flow rate, and power.

[0042] The correction parameters include one or more of the following: unit aging coefficient based on power generation duration, water temperature correction factor, and vibration state assessment parameters. Specifically, the method for obtaining the unit aging coefficient based on power generation duration includes: establishing a unit aging model that considers the impact of factors such as unit operating time and number of start-ups and shutdowns on unit performance. For example, by analyzing historical operating data, a functional relationship between the aging coefficient and power generation duration can be established. As the power generation duration increases, the unit aging coefficient gradually decreases, reflecting the decline in unit efficiency.

[0043] The method for obtaining the water temperature correction factor includes: collecting water temperature data under different water heads, analyzing the impact of water temperature on unit efficiency, establishing a functional relationship between the water temperature correction factor and the water head. Changes in water temperature affect the density and viscosity of water, thereby affecting the efficiency of the unit. The efficiency model is corrected by using the water temperature correction factor.

[0044] The methods for obtaining vibration status assessment parameters include: installing vibration sensors to monitor the vibration of the unit, establishing vibration status assessment indices, calculating vibration status assessment parameters based on the vibration indices and power generation duration, and correcting the efficiency model by using vibration status assessment parameters when the unit vibration is too large.

[0045] The above-mentioned correction parameters are introduced into the fitted efficiency model to form a dynamic efficiency prediction model that is strongly coupled with the power generation duration.

[0046] S220, construct the objective function and constraints.

[0047] The objective function includes maximizing the overall power generation efficiency under power generation duration constraints, minimizing unit start-up and shutdown losses, and balancing unit operating time. Constraints include upper and lower limits of unit output, flow balance based on power generation duration, duration-dependent regulation rate limits, and vibration constraints.

[0048] Specifically, the overall power generation efficiency can be expressed as: Where n is the number of generator sets. For the first The power output of each generating unit is optimized to maximize overall power generation efficiency while meeting power generation duration constraints.

[0049] Minimize unit start-up and shutdown losses: Unit start-up and shutdown losses are related to the number of start-ups and shutdowns and the energy consumption during the start-up and shutdown process. A unit start-up and shutdown loss model can be established. ,in For the first Loss coefficient for each start-up and shutdown of the generator unit. For the first The goal of minimizing the number of start-ups and shutdowns of the generator units during the optimization process is to reduce start-up and shutdown losses.

[0050] Equalizing unit operating time: Defines the unit operating time equalization index ,in For the first The operating time of the unit The average operating time of all units is calculated by optimizing power distribution to make the operating time of each unit as close as possible to the average operating time, so as to balance the wear and maintenance needs of the units.

[0051] S230, based on the objective function, constraints, and dynamic efficiency prediction model, uses a particle swarm optimization algorithm combined with simulated annealing to solve for the Pareto optimal solution set containing power generation duration constraints.

[0052] Specifically, the particle swarm is initialized by randomly initializing a group of particles, each representing a power allocation scheme for a group of generators. Each particle has two attributes: velocity and position. Velocity represents the particle's direction and speed of movement in the search space, and position represents the particle's current power allocation scheme. Based on the objective function and constraints, the fitness value of each particle is calculated. The fitness value is used to evaluate the merits of each power allocation scheme; a higher fitness value indicates a better scheme. In each iteration, particles update their velocity and position based on their historical best position and global best position. Simultaneously, a simulated annealing strategy is introduced to accept poorer solutions with a certain probability, thus preventing the algorithm from getting trapped in local optima. After each iteration, the Pareto optimal solution set is updated. If a particle's solution is better than a solution in the Pareto optimal solution set, that solution is added to the Pareto optimal solution set, and the dominated solution is removed. When the preset number of iterations is reached or other termination conditions are met, the algorithm stops and outputs a Pareto optimal solution set containing the power generation duration constraint.

[0053] S240 uses the Pareto optimal solution set as the decision space and constructs a multi-criteria decision matrix based on fuzzy hierarchical analysis, which integrates generation duration, real-time electricity price, unit health status index optimized based on duration, and grid dispatch priority weight.

[0054] Specifically, based on the fuzzy hierarchical analysis method, experts were invited to assess the importance of indicators such as power generation duration, real-time electricity price, unit health status index optimized based on duration, and grid dispatch priority weight. A judgment matrix was constructed to calculate the weight of each indicator. Data for indicators such as power generation duration, real-time electricity price, unit health status index, and grid dispatch priority weight were standardized to eliminate dimensional differences between different indicators. Commonly used standardization methods include linear transformation and logarithmic transformation. Using the Pareto optimal solution set as the decision space, each solution corresponds to a decision scheme. The indicator values ​​of each decision scheme were arranged into a matrix according to the standardized results. Each row of the matrix represents a decision scheme, and each column represents an indicator.

[0055] In this embodiment, the multi-criteria decision matrix contains quantitative information for each scheme in the Pareto optimal solution set under multiple criteria such as generation duration, real-time electricity price, time-optimized unit health status index, and grid dispatch priority weight. This information is presented in the form of a matrix, with each row representing a scheme corresponding to a Pareto optimal solution and each column representing the standardized value of a criterion.

[0056] S250 obtains the load allocation scheme that best matches the generation duration and capacity from the Pareto optimal solution set based on the multi-criteria decision matrix.

[0057] Specifically, historical data is collected, including the Pareto optimal solution set, various index values, and corresponding optimal load allocation schemes. This data is used to train a neural network. The input to the neural network is the index values ​​in the multi-criteria decision matrix, and the output is the optimal load allocation scheme. The current multi-criteria decision matrix is ​​input into the trained neural network, and the neural network outputs a load allocation scheme that best matches the generation duration and capacity. Through the learning and prediction capabilities of the neural network, the optimal load allocation scheme can be selected more accurately from the Pareto optimal solution set.

[0058] Specifically, the data in the multi-criteria decision matrix is ​​divided into a training set, a validation set, and a test set. The training set is used to train the neural network, the validation set is used to adjust the hyperparameters of the model during training to prevent overfitting, and the test set is used to evaluate the performance of the trained neural network. The preferred division ratio is 70%, 15%, and 15%. For each sample in the training set and the validation set (i.e., each row of the multi-criteria decision matrix), it is labeled with the corresponding optimal load allocation scheme. These labels can be the optimal scheme actually used in historical data or the optimal scheme determined by expert evaluation.

[0059] Then, based on the complexity of the problem and the characteristics of the data, a suitable neural network architecture is selected. For multi-criteria decision problems, multilayer perceptrons (MLPs) are a commonly used choice because they can handle nonlinear relationships and learn the complex correlations between factors in the multi-criteria decision matrix well. The structural parameters of the neural network, such as the number of layers and the number of neurons in each layer, are determined. The number of neurons in the input layer should be the same as the number of columns in the multi-criteria decision matrix (i.e., the number of criteria), and the number of neurons in the output layer should be the same as the dimension of the load allocation scheme (i.e., the number of units). For example, if the multi-criteria decision matrix has 4 criteria and 5 units need to be allocated load, then the input layer has 4 neurons and the output layer has 5 neurons. The number of layers and neurons in the intermediate hidden layers can be adjusted experimentally to achieve optimal performance.

[0060] Choosing an appropriate loss function to measure the difference between the neural network's predictions and the actual labels is crucial. For load allocation prediction, mean squared error (MSE) is a commonly used loss function, measuring the average error between the predicted load allocation and the actual optimal load allocation. Optimization algorithms are then used to minimize the loss function; common algorithms include stochastic gradient descent (SGD) and Adam.

[0061] The training data is input into the neural network. Forward propagation is used to calculate the prediction result, then the error between the prediction result and the label is calculated based on the loss function. The gradient is then calculated using the backpropagation algorithm. Finally, an optimization algorithm is used to update the neural network parameters. This process is repeated until the loss function converges or a preset number of training epochs are reached. The trained neural network is evaluated using validation and test sets. Evaluation metrics may include mean squared error (MSE) and mean absolute error (MAE). These metrics determine whether the model's performance meets the requirements. If the model's performance is unsatisfactory, the neural network's hyperparameters, such as the learning rate, the number of neurons in the hidden layers, and the regularization parameter, can be adjusted. Grid search, random search, and other methods can be used to find the optimal combination of hyperparameters.

[0062] The current multi-criteria decision matrix is ​​input into the trained neural network. The neural network calculates the output result through forward propagation. This output result is the load allocation scheme that best matches the generation duration and capacity. In S250, the multi-criteria decision matrix is ​​used as input, and the decision engine trained by the neural network obtains the load allocation scheme that best matches the generation duration and capacity from the Pareto optimal solution set.

[0063] Furthermore, a rolling optimization mechanism linked to head forecasting can be established, which recalculates load allocation every 3 minutes based on updated power generation duration and capacity, reserves a 15% adjustment margin, and outputs the duration-load coupling optimization results to the lower-level control system to ensure that the system response time does not exceed 30 seconds.

[0064] Reference Figure 3 The method for "constructing a startup control strategy based on the optimal load allocation scheme" in S300, i.e., the method for constructing the startup control strategy, includes: S310 uses the optimal load allocation scheme as input and determines the multi-level startup sequence based on the power allocation and startup power requirements of each unit. The startup sequence matrix for all units is determined based on the multi-level startup sequence.

[0065] In this embodiment, a three-stage startup sequence is preferred, wherein the first stage selects and allocates power to satisfy... Small-capacity generator units with a capacity of ≤5MW and self-starting capability are used as starting power sources, and the secondary power allocation meets the requirements of 5MW < Medium-capacity units ≤20MW are used to provide stable voltage and frequency support, enhancing system stability; the third-level selection of power allocation meets... Large-capacity generating units of >20MW are used to undertake the main load and meet the power demand of the power grid.

[0066] The startup sequence matrix includes the allocated power, startup time, and startup priority of each unit. In this embodiment, the startup timing optimization problem is transformed into a path planning problem, with the startup time of each unit as a node on the path. The goal is to find an optimal path that minimizes the total startup time while satisfying the constraint that the startup interval at each stage is no less than 45 seconds.

[0067] Furthermore, an improved ant colony algorithm is preferably used to optimize the startup sequence based on the load allocation matrix to obtain the startup sequence matrix of all units. Specifically, 1) Initialize the number of ants, usually set to 20-50, randomly initialize the initial path of each ant, i.e., randomly assign the startup order of each unit, and initialize the pheromone matrix. The elements of the pheromone matrix represent the pheromone concentration from the startup of one unit to the startup of another unit. The initial value can be set to a small constant, such as 0.1; 2) Iterative search: Each ant selects the next unit to start based on the pheromone matrix and heuristic information (such as the startup power requirement of the unit, the allocated power, etc.) and constructs its own startup path; calculate the fitness value of the path constructed by each ant. The fitness value can be defined as the total startup time plus the penalty term for violating the startup interval constraint; update the pheromone matrix and update the pheromone concentration on the path according to the fitness value. The smaller the fitness value, the more the pheromone concentration increases. 3) Repeat the above steps until the termination condition is met, such as reaching the maximum number of iterations (usually set to 100-200 times) or the fitness value no longer improves. 4) Select the path with the minimum fitness value as the optimal startup sequence. Tstartup=[T 1 , T 2 ,⋯,T n ] 5) Based on the unit stratification results and the optimized startup sequence, generate a startup sequence matrix. Sequence_matrix =[Unit Number, Allocated Power, Start-up Time, Start-up Priority]. The start-up priority can be sorted according to the start-up sequence, with the earlier the start-up time, the higher the priority.

[0068] Suppose a power system has 5 generating units (A, B, C, D, and E). Based on the optimal load allocation scheme, the system requires 100MW of power in the initial startup phase, 200MW in the mid-term, and 300MW in the later stage. The rated powers of each unit are A-50MW, B-80MW, C-100MW, D-120MW, and E-150MW, respectively, with startup power requirements of A-10MW, B-15MW, C-20MW, D-25MW, and E-30MW. Based on this information, a multi-stage startup sequence is determined: units A and B are started in the initial stage because they have low startup power requirements and can quickly provide a certain amount of power; unit C is started in the mid-term; and units D and E are started in the later stage. Determining the multi-stage startup sequence allows for a reasonable arrangement of the unit startup order based on the power requirements of different system stages, avoiding system impact caused by starting too many units simultaneously, and improving system startup stability. The startup sequence matrix integrates key unit information, facilitating subsequent analysis and processing, and providing a clear data foundation for constructing startup control strategies.

[0069] S320 constructs a black-start strategy optimization model based on deep reinforcement learning, and trains the black-start strategy optimization model based on historical start sequence data related to power grid black-start.

[0070] Specifically, a black-start strategy optimization model was constructed using the Deep Q-Network (DQN) algorithm from deep reinforcement learning. Historical black-start sequence data from the past 10 years for this power grid was collected, including information on unit startup order, power allocation, and system response under different fault scenarios. This historical data was divided into training and validation sets for model training. During training, the model continuously tried different startup strategies and adjusted the strategy according to a preset reward function (e.g., higher rewards for shorter system recovery time and smaller frequency and voltage fluctuations) until the model converged. Deep reinforcement learning models possess powerful learning and decision-making capabilities, enabling them to learn the optimal startup strategy from a large amount of historical data, adapting to different fault scenarios and system states. By training the model, startup strategies can be continuously optimized, improving the efficiency and reliability of black starts and reducing system recovery time and losses.

[0071] S330 establishes a multi-dimensional decision matrix that includes water level, start-up sequence matrix, power grid topology, and target load demand.

[0072] Considering the presence of multiple hydropower stations in the power grid, water level information significantly impacts the startup and power generation capacity of the hydropower units. This study obtains the current water level data for each hydropower station and, combined with the previously determined startup sequence matrix, the power grid topology (such as line connection methods and substation locations), and target load demand (such as user electricity demand at different times), establishes a multi-dimensional decision matrix. This multi-dimensional decision matrix comprehensively considers multiple factors influencing black start, enabling the model to gain a more comprehensive understanding of the system's state and demands, thereby formulating a more reasonable and realistic startup strategy. The inclusion of water level information is particularly important for power grids containing hydropower plants, as it allows for the full utilization of water resources, improving the power generation efficiency of hydropower plants and the overall stability of the system.

[0073] S340: Input the multidimensional decision matrix into the trained black-start strategy optimization model to obtain the start control strategy.

[0074] The constructed multidimensional decision matrix is ​​input into the trained black-start strategy optimization model. The model reasons and makes decisions based on the information in the matrix, outputting a specific start-up control strategy. For example, the strategy includes: given a current water level of 50m, a certain grid topology, and a target load demand of 100MW, start generating units A and B sequentially according to the start-up sequence matrix, while adjusting the power output of the units to meet the load demand. By inputting the multidimensional decision matrix into the trained model, a start-up control strategy adapted to the current system state can be obtained quickly and accurately, improving the scientific nature and efficiency of decision-making. This strategy is based on a large amount of historical data and deep reinforcement learning model training, possessing high reliability and optimization capabilities, and effectively ensuring the smooth black start of the power grid.

[0075] In this embodiment, by rationally determining the multi-level startup sequence and constructing the startup sequence matrix, excessive shocks during system startup are avoided. Simultaneously, the combination of a deep reinforcement learning model and a multi-dimensional decision matrix enables the formulation of optimal strategies based on the actual system state, reducing system frequency and voltage fluctuations and improving system stability during startup. The model trained using historical data can learn the optimal startup strategy and make rapid decisions based on the current system state, thereby reducing the time of black starts in the power grid, restoring power supply as quickly as possible, and minimizing economic losses caused by power outages. The multi-dimensional decision matrix comprehensively considers multiple factors such as water level, grid topology, and target load demand, enabling the startup control strategy to adapt to different fault scenarios and system operating conditions, improving the flexibility and adaptability of the strategy. When formulating the strategy, the allocated power and startup priority of the generating units, as well as resource information such as water level, are fully considered, enabling the rational allocation and effective utilization of resources, improving the operating efficiency and economic benefits of the power grid.

[0076] The S400 method of "achieving unified management of unit start-up and shutdown and load regulation through the automatic control algorithm of the computer monitoring platform according to the start-up control strategy, and obtaining the operating status of the unified management system" specifically includes: S410 constructs a three-layer monitoring architecture based on cloud-edge collaboration, and obtains real-time monitoring datasets of the units based on the three-layer monitoring architecture.

[0077] The three-tier monitoring architecture includes a field control layer, an edge computing layer, and a cloud decision-making layer; the real-time monitoring dataset includes unit operating status, startup progress, and system response parameters.

[0078] Specifically, sensors such as temperature sensors, pressure sensors, and vibration sensors are installed at various key parts of the unit to collect real-time operating data. These sensors transmit the data to a local controller, such as a programmable logic controller (PLC), which performs preliminary data processing and storage. Edge computing devices receive data from the PLC at the field control layer and perform further analysis and processing, such as data cleaning and feature extraction. Simultaneously, edge computing devices can make local decisions based on preset rules, such as issuing an alarm when an abnormality in a unit parameter is detected. A cloud server connects to the edge computing devices via a network, receives the processed data, and uses big data analytics and machine learning algorithms in the cloud to perform in-depth data mining to obtain a real-time monitoring dataset for the unit, including unit operating status, startup progress, and system response parameters.

[0079] The cloud-edge collaborative three-layer monitoring architecture enables layered data processing and analysis, reducing the burden on cloud servers and improving system response speed; the sensors in the field control layer can collect unit operating data in real time and accurately, providing a reliable basis for subsequent analysis and decision-making; the local decision-making function of the edge computing layer can take timely measures in case of emergencies to ensure the safe operation of the unit.

[0080] S420 obtains unit status information based on real-time monitoring datasets, including unit temperature, vibration, efficiency, and availability.

[0081] Specifically, data collected by temperature sensors is extracted from the real-time monitoring dataset. By analyzing the temperature of different parts, the thermal status of the unit is determined. For example, excessively high generator winding temperature may indicate overload or poor heat dissipation. Vibration sensor data is used to analyze the unit's vibration frequency and amplitude. By comparing this data with vibration data during normal operation, mechanical faults such as bearing wear or rotor imbalance can be identified. The unit's efficiency is calculated based on its input and output power data. For example, for a hydroelectric turbine, the efficiency can be calculated by measuring water flow and head, as well as the generator's output power. Finally, the unit's availability is assessed by combining maintenance records and fault history. For instance, if the unit has recently undergone maintenance and all parameters are normal, its availability is considered high.

[0082] A comprehensive understanding of the unit's status information can help identify potential problems in a timely manner, providing a basis for unit maintenance and repair, preventing major failures, and accurate unit status information helps optimize unit operating parameters, thereby improving unit operating efficiency and reliability.

[0083] S430 establishes an adaptive start-stop timing optimization model based on reservoir operating conditions, power grid load forecasts, start-up control strategies, and unit status information.

[0084] Specifically, information such as reservoir water level and flow rate is collected, and the reservoir's storage capacity and flow conditions are analyzed to determine the hydropower resources that the reservoir can provide to the generating units. Historical grid load data and meteorological data are used to predict future grid load using time series analysis, machine learning, and other methods. Based on grid demand and generating unit characteristics, a unit start-up control strategy is formulated, such as prioritizing the start-up of high-efficiency units and avoiding frequent unit start-ups and shutdowns. Taking into account reservoir operating conditions, grid load forecasts, start-up control strategies, and generating unit status information, an adaptive start-up and shutdown timing optimization model is established. For example, linear programming and dynamic programming methods can be used to build the model to determine the optimal start-up and shutdown timing of the units.

[0085] In this step, the adaptive start-stop timing optimization model can dynamically adjust the start-stop timing of the units according to different reservoir operating conditions and power grid load requirements, thereby improving the operating efficiency and economic benefits of the units. At the same time, considering the unit status information can avoid starting the units when there are faults or maintenance needs, thus ensuring the safe operation of the units.

[0086] S440 takes minimizing start-up and shutdown costs and maximizing unit efficiency as objective functions, and uses the unit's preheating and cooling times as constraints. It uses a genetic algorithm to optimize and solve the adaptive start-up and shutdown timing model, obtains the optimal start-up and shutdown timing, power setpoints and control parameters that satisfy the objective functions, and generates a start-up and shutdown control instruction set.

[0087] Specifically, an objective function is defined to minimize start-up and shutdown costs and maximize unit efficiency. Start-up and shutdown costs include energy consumption and equipment wear and tear during unit startup and shutdown. Unit efficiency can be represented by the ratio of output power to input power. Preheating and cooling times are set as constraints to ensure sufficient time for preheating and cooling during startup and shutdown to avoid damage to the unit. A genetic algorithm is used to optimize the adaptive start-up and shutdown timing model. The genetic algorithm iteratively searches for the optimal solution by simulating biological evolution. In each iteration, individuals are evaluated and selected based on the objective function and constraints. After multiple iterations, the optimal solution satisfying the objective function is obtained, namely the optimal start-up and shutdown timing, power setpoint, and control parameters. Based on the optimal start-up and shutdown timing, power setpoint, and control parameters obtained from the optimization, a start-up and shutdown control instruction set is generated, including the unit's start-up time, stop time, and power setpoint.

[0088] In this step, minimizing start-up and shutdown costs and maximizing unit efficiency are the objective functions, which can reduce the operating costs of the unit and improve its economic benefits. Genetic algorithms have strong global search capabilities and can find the optimal solution in a complex solution space, thereby improving the optimization effect of the model.

[0089] S450 outputs the unified management system operation status corresponding to different types of power grid dispatching instructions based on the start-stop control instruction set. The unified management system operation status includes load allocation execution status, unit coordination operation parameters, and system stability indicators.

[0090] Specifically, based on the start-stop control command set, the grid load is distributed to various generating units, and the load distribution execution status of the units is monitored in real time. For example, by monitoring the output power of the units, it is determined whether the units are distributing load according to the requirements of the command set; according to the start-stop control command set, the operating parameters of the units, such as speed and flow rate, are adjusted to ensure coordinated operation between the units. For example, when multiple units are operating in parallel, the governor parameters of the units are adjusted to keep the output power of each unit balanced; the stability indicators of the system, such as voltage and frequency, are monitored to determine whether the system is operating stably. If the system becomes unstable, the operating parameters of the units are adjusted in a timely manner to ensure the stability of the system.

[0091] In this step, outputting the unified management system operation status corresponding to different types of power grid dispatching instructions can realize unified management and dispatching of generating units, improve the operating efficiency and stability of the power grid, and at the same time, real-time monitoring of the system's operating status can promptly detect and resolve problems in the system, ensuring the safe and reliable operation of the power grid.

[0092] The S500 method of "determining the support scheme for hydropower stations based on the operating status of the unified management system and through a capacity calculation model" specifically includes: S510 constructs a three-dimensional capacity assessment matrix that integrates reservoir energy storage level and power grid security constraints based on the operating status of the unified management system.

[0093] Specifically, the three dimensions of the matrix are set as time, reservoir storage level, and power grid safety constraints. The time dimension can be divided into time intervals such as hours and days. The reservoir storage level can be quantitatively graded according to indicators such as reservoir water level and storage capacity. Power grid safety constraints can include voltage stability range and frequency fluctuation range. Relevant data, such as reservoir water level at different time points, real-time voltage and frequency of the power grid, are obtained from the unified management system. This data is categorized and organized according to the dimensions of the matrix. Based on the collected data, the supporting capacity that the hydropower station can provide under the corresponding reservoir storage level and power grid safety constraints at each time point is calculated and filled into the corresponding position of the three-dimensional capacity assessment matrix. For example, at a specific moment, when the reservoir storage is at a medium level, the power grid voltage is within the safe range, and the frequency is close to the rated value, the generating capacity that the hydropower station can provide is calculated as an element of the matrix.

[0094] This step integrates two key factors: reservoir storage capacity and grid security constraints. It enables a more comprehensive and accurate assessment of the support capacity of hydropower stations under different conditions, avoiding the limitations of assessment based on a single factor.

[0095] S520 constructs a capacity calculation model based on a three-dimensional capacity assessment matrix.

[0096] Specifically, based on the data characteristics of the three-dimensional capacity assessment matrix, a suitable mathematical model structure is selected, such as a linear regression model or a neural network model. If the matrix data exhibits a relatively linear relationship, a linear regression model can be used; if the data relationship is complex, a neural network model can be considered. The model is trained using historical data from the three-dimensional capacity assessment matrix, and the model parameters are determined through optimization algorithms (such as least squares method, gradient descent method, etc.) so that the model can accurately calculate the support capacity based on the input time, reservoir storage level, and grid safety constraints. The constructed model is validated using a portion of data not used in training to evaluate its accuracy and reliability. If the model's prediction results deviate significantly from the actual situation, the model's structure or parameters are adjusted until a satisfactory result is achieved.

[0097] By building a model, the support capacity of a hydropower station can be accurately calculated based on different input conditions, providing a scientific basis for formulating support plans. The model can be adjusted and optimized according to the actual situation to adapt to the characteristics of different hydropower stations and changing operating conditions.

[0098] S530 employs a sliding time window technique in its capacity calculation model to perform time-series analysis of the unified management system's operational status across three time scales, outputting multi-scenario support solutions, including a rapid response solution, a standard support solution, and a maximum support solution.

[0099] Specifically, the time scale can be divided into short-term (e.g., the next 1-2 hours), medium-term (e.g., the next 1-2 days), and long-term (e.g., the next 1-2 weeks). At each time scale, a sliding time window technique is used to analyze the operational status data of the unified management system. For example, at the short-term time scale, a sliding time window of 1 hour is set, sliding across the time series data in 15-minute increments. The data within each window is used as input, and the support capacity calculation model is used to calculate the support capacity within that time period. Based on the analysis results at different time scales and combined with different grid needs, multi-scenario support solutions are generated. The rapid response solution is mainly for short-term, sudden grid emergency needs, utilizing the rapid adjustment capabilities of hydropower stations to provide immediate support. The standard support solution is suitable for medium-term routine grid dispatch needs, ensuring the stable operation of the grid. The maximum support solution fully utilizes the maximum power generation capacity of hydropower stations during long-term grid load peaks or special circumstances.

[0100] By conducting time-series analysis at three time scales and generating multi-scenario solutions, the power grid's support needs can be met under different time periods and conditions, thereby improving the power grid's response capabilities. The sliding time window technology can track changes in the system's operating status in real time, enabling support solutions to be dynamically adjusted according to actual conditions, thus enhancing the effectiveness and adaptability of the solutions.

[0101] The S600 method for "calling and executing the corresponding support scheme when an abnormal state occurs in the power grid" includes: The S610 can detect the power grid status in real time and obtain abnormal power grid conditions.

[0102] Abnormal states include significant power loss and system frequencies severely below the normal range. Specifically, numerous monitoring devices, including voltage sensors, current sensors, and frequency sensors, are installed at key nodes in the power grid, such as substations and transmission lines. These sensors can collect parameters such as voltage, current, and frequency of the power grid in real time. The monitoring devices transmit the collected data to the power grid monitoring center in real time via communication networks. The computer system at the monitoring center processes and analyzes this data in real time, comparing it with pre-set normal parameter ranges. When certain parameters are detected to be outside the normal range, the system determines that the power grid is in an abnormal state. For example, when the system frequency is detected to be severely below the normal range (e.g., below 49.5Hz), or when significant power loss occurs (e.g., a sudden and substantial drop in power in a certain area), the system will promptly identify and record these abnormal states.

[0103] By monitoring the power grid status in real time, abnormal situations in the power grid can be detected immediately, buying valuable time for subsequent handling and preventing the abnormal situation from deteriorating further and causing more serious consequences. A large number of monitoring devices and real-time data processing can accurately obtain specific information about the abnormal state of the power grid, such as the time, location, and type of the abnormality, providing a reliable basis for subsequent decision-making.

[0104] S620 determines the early warning level and support capacity requirements based on abnormal power grid conditions.

[0105] Specifically, different warning level classification standards are established based on the severity of abnormal grid conditions. For example, for large power loss situations, a yellow warning is set when the lost power accounts for 5%-10% of the total load; a red warning is set when the lost power accounts for more than 10%. For situations where the system frequency is significantly lower than the normal range, a yellow warning is set when the frequency is between 49-49.5Hz, and a red warning is set when it is below 49Hz. Based on the type and severity of the abnormal condition, combined with the grid's operating parameters and historical data, the required support capacity is calculated. For example, in the event of a large power loss, the required supplementary power capacity is calculated based on the magnitude of the lost power and the reserve capacity of other units; for situations where the system frequency is too low, the required additional generating power to restore the frequency is calculated based on parameters such as the magnitude of the frequency drop and the system's inertia constant.

[0106] In this step, determining the warning level allows for tiered management of abnormal states of varying severity, enabling staff to take appropriate measures based on the warning level and improving response efficiency. Accurately calculating support capacity requirements ensures the deployment of suitable support solutions, avoiding insufficient or excessive support and improving the effectiveness and cost-effectiveness of support.

[0107] S630 invokes and executes the corresponding support plan based on the warning level and support capacity requirements.

[0108] A detailed support plan library is pre-established, and corresponding support plans are formulated for different warning levels and support capacity requirements. For example, for a yellow warning, some standby units may be called up for power generation support; for a red warning, more units need to be mobilized or even load shedding measures may be taken. Once the warning level and support capacity requirements are determined, the system automatically selects the corresponding support plan from the support plan library and sends the execution instructions to the relevant equipment and departments. For example, if a support plan for a hydropower station needs to be called, the system will send instructions to the hydropower station's monitoring system to start the units and adjust the power output, and the hydropower station will execute the corresponding operations according to the instructions. During the execution of the support plan, the execution status is monitored in real time to ensure the effective implementation of the plan. If problems are found during the execution process, adjustments and corrections are made in a timely manner.

[0109] The ability to quickly invoke corresponding support solutions based on early warning levels and support capacity requirements enables responses to abnormal power grid conditions in the shortest possible time, reducing the impact of abnormal conditions on the power grid. The establishment of a support solution library and the automatic invocation of solutions make the operation of responding to abnormal power grid conditions more standardized and regulated, improving the reliability and accuracy of the processing.

[0110] By detecting, accurately judging, and responding promptly to abnormal grid conditions in real time, effective support measures can be taken quickly to maintain the stability of grid voltage and frequency, prevent serious accidents such as grid collapse, and ensure the safe and reliable operation of the grid. The hierarchical management and precise support plan call mechanism enable staff to respond quickly and accurately to different types and degrees of abnormal grid conditions, improving the efficiency of responding to abnormal grid conditions. Timely and effective support measures can reduce the impact of abnormal grid conditions on power users, reduce economic losses caused by power outages and other accidents, and at the same time avoid energy waste and increased costs caused by excessive support.

[0111] Secondly, this application discloses a load optimization and emergency support control system for an energy storage control system, used to execute the load optimization and emergency support control method for an energy storage control system disclosed in the first aspect of this application. The system includes: The power generation duration and capacity acquisition module is used to monitor head data in real time based on the constructed multi-level head sensor network and dynamically determine the power generation duration and capacity based on the hydraulic model. The optimal load allocation scheme determination module is used to obtain the optimal load allocation scheme based on the power generation duration, capacity, and efficiency characteristic curves of each unit through an optimization algorithm. The startup control strategy construction module is used to build startup control strategies based on the optimal load allocation scheme. The unified management system operation status acquisition module is used to obtain the unified management system operation status by realizing the unified management of unit start-up and shutdown and load regulation through the automatic control algorithm of the computer monitoring platform according to the start-up control strategy. The support scheme acquisition module is used to determine the support scheme for the hydropower station based on the operating status of the unified management system and the capacity calculation model. The trigger module is used to invoke and execute the corresponding support scheme when an abnormal state occurs in the power grid.

[0112] A computer device according to embodiments of the present disclosure includes a memory and a processor. The memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), a hard disk, flash memory, etc. The processor may be a central processing unit (CPU) or other forms of processing unit having data processing capabilities and / or instruction execution capabilities, and may control other components in the computer device to perform desired functions.

[0113] like Figure 4 This is a schematic diagram of a computer device provided for an embodiment of the present disclosure. It illustrates a structural schematic diagram suitable for implementing the computer device in the embodiments of the present disclosure. Figure 4 The computer device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0114] like Figure 4 As shown, a computer device may include a processor (such as a central processing unit, graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) or programs loaded from storage devices into random access memory (RAM). The RAM also stores various programs and data required for the operation of the computer device. The processor, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0115] Typically, the following devices can be connected to the I / O interface: input devices, such as sensors or visual information acquisition devices; output devices, such as displays; storage devices, such as magnetic tapes or hard drives; and communication devices. Communication devices allow the computer device to communicate wirelessly or wiredly with other devices (such as edge computing devices) to exchange data. Although Figure 4A computer apparatus with various devices is shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or included alternatively.

[0116] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A method for load optimization and emergency support control of an energy storage control system, characterized in that, include: The head data is monitored in real time by constructing a multi-level head sensor network, and the power generation duration and capacity are dynamically determined based on the hydraulic model. Based on the power generation duration and capacity and the efficiency characteristic curves of each unit, the optimal load allocation scheme is obtained through an optimization algorithm. Based on the optimal load allocation scheme, a startup control strategy is constructed; According to the aforementioned startup control strategy, the automatic control algorithm of the computer monitoring platform is used to achieve unified management of unit startup and shutdown and load regulation, thereby obtaining the operating status of the unified management system. Based on the operational status of the unified management system, a support plan for the hydropower station is determined using a capacity calculation model; When the power grid experiences an abnormal state, the corresponding support scheme is invoked and executed.

2. The load optimization and emergency support control method for energy storage control system according to claim 1, characterized in that, The process of monitoring head data in real time based on a constructed multi-level head sensor network and dynamically determining power generation duration and capacity based on a hydraulic model includes: Pressure level gauges, ultrasonic level gauges, and radar level gauges are installed at monitoring points in the upstream and downstream reservoir areas of the hydropower station in a gradient distribution to form a multi-level head sensor network. Based on the multi-level head sensor network, head data is monitored in real time to obtain a head dataset; Based on the head dataset, a four-dimensional hydraulic model is constructed that integrates reservoir capacity characteristic curves, dynamic discharge coefficients, seasonal hydrological fluctuations, and meteorological forecast data. Based on the four-dimensional hydraulic model, noise suppression and multi-step prediction are performed on the head dataset, and the net head time series and predicted head time series are output. Based on the net head time series and the predicted head time series, combined with the multi-unit flow characteristics and nonlinear reservoir capacity function, a genetic algorithm is used to optimize the optimal power generation duration under dynamic constraints, denoted as the power generation duration capacity.

3. The load optimization and emergency support control method for energy storage control system according to claim 2, characterized in that, The process of obtaining the optimal load allocation scheme based on the power generation duration and capacity and the efficiency characteristic curves of each unit through an optimization algorithm includes: Based on the power generation duration and capacity and the net head time series, a unit efficiency model with time-varying head-flow-power three-dimensional mapping is constructed. Key points are determined from the unit efficiency model. Based on these key points, the nonlinear efficiency curves in the unit efficiency model are fitted using a Bézier surface fitting algorithm. After introducing correction parameters, a dynamic efficiency prediction model strongly coupled with the power generation duration is obtained. Construct the objective function and constraints; The objective function includes maximizing the overall power generation efficiency under the constraint of power generation duration, minimizing unit start-up and shutdown losses, and balancing unit operating time. The constraints include upper and lower limits of unit output, flow balance based on power generation duration, duration-related regulation rate limit, and vibration constraint. Based on the objective function, the constraints, and the dynamic efficiency prediction model, the Pareto optimal solution set containing the power generation duration constraint is solved. Using the Pareto optimal solution set as the decision space, a multi-criteria decision matrix is ​​constructed that integrates generation duration, real-time electricity price, unit health status index based on duration optimization, and grid dispatch priority weight. Based on the multi-criteria decision matrix, a load allocation scheme that best matches the generation duration and capacity is obtained from the Pareto optimal solution set.

4. The load optimization and emergency support control method for energy storage control system according to claim 1, characterized in that, The startup control strategy, based on the optimal load allocation scheme, includes: A multi-level startup sequence is determined based on the optimal load allocation scheme, the power allocation of each unit, and the startup power requirements. The startup sequence matrix of all units is determined based on the multi-level startup sequence, and the startup sequence matrix includes the allocated power, startup time and startup priority of each unit; A black start strategy optimization model is constructed based on deep reinforcement learning, and the black start strategy optimization model is trained based on historical start sequence data related to power grid black start. Establish a multi-dimensional decision matrix that includes water level, the start-up sequence matrix, power grid topology, and target load demand; The multidimensional decision matrix is ​​input into the trained black-start strategy optimization model to obtain the startup control strategy.

5. The load optimization and emergency support control method for the energy storage control system according to claim 4, characterized in that, The multi-level startup sequence includes: a first-level startup sequence, a second-level startup sequence, and a third-level startup sequence, including: The first-level startup sequence includes units whose allocated power meets a first preset range and which have self-starting capability; The second-level startup sequence includes units whose allocated power meets the second preset range and which have self-starting capability; The third-level startup sequence includes units whose allocated power meets the third preset range and which have self-starting capability; The third preset range is greater than the second preset range, and the second preset range is greater than the first preset range.

6. The load optimization and emergency support control method for the energy storage control system according to claim 4, characterized in that, Based on the aforementioned startup control strategy, the automatic control algorithm of the computer monitoring platform enables unified management of unit start-up and shutdown and load regulation, thereby obtaining the unified management system's operating status, including: A three-layer monitoring architecture based on cloud-edge collaboration is constructed, and a real-time monitoring dataset of the unit is obtained based on the three-layer monitoring architecture. The unit status information is obtained based on the real-time monitoring dataset, and the unit status information includes the unit's temperature, vibration, efficiency, and availability. Based on the reservoir operating conditions, power grid load forecasts, the aforementioned startup control strategy, and the aforementioned unit status information, an adaptive start-stop timing optimization model is established. The adaptive start-stop timing optimization model is optimized using a genetic algorithm with the objective functions of minimizing start-stop costs and maximizing unit efficiency, and with the preheating and cooling times of the unit as constraints. The optimal start-stop timing, power setpoints and control parameters that satisfy the objective functions are obtained, and a start-stop control instruction set is generated. Based on the start-stop control instruction set, the unified management system operation status corresponding to different types of power grid dispatch instructions is output. The unified management system operation status includes load allocation execution status, unit coordination operation parameters, and system stability indicators.

7. The load optimization and emergency support control method for an energy storage control system according to claim 6, characterized in that, The step of determining the support plan for the hydropower station based on the operating status of the unified management system and through a capacity calculation model includes: A three-dimensional capacity assessment matrix integrating reservoir energy storage level and power grid security constraints is constructed based on the operating status of the unified management system. A support capacity calculation model is constructed based on the aforementioned three-dimensional capacity assessment matrix; The support capacity calculation model employs a sliding time window technique to perform time-series analysis of the unified management system's operational status across three time scales, outputting support schemes for multiple scenarios. These support schemes include a rapid response scheme, a standard support scheme, and a maximum support scheme.

8. The load optimization and emergency support control method for energy storage control system according to claim 1, characterized in that, When an abnormal state occurs in the power grid, the corresponding support scheme is invoked and executed, including: Real-time monitoring of power grid status and acquisition of abnormal power grid conditions; The warning level and support capacity requirements are determined based on the abnormal power grid condition. The corresponding support plan is invoked and executed based on the warning level and the support capacity requirement.

9. A load optimization and emergency support control system for an energy storage control system, characterized in that, include: The power generation duration and capacity acquisition module is used to monitor head data in real time based on the constructed multi-level head sensor network and dynamically determine the power generation duration and capacity based on the hydraulic model. The optimal load allocation scheme determination module is used to obtain the optimal load allocation scheme based on the power generation duration capacity and the efficiency characteristic curves of each unit through an optimization algorithm. A startup control strategy construction module is used to construct a startup control strategy based on the optimal load allocation scheme. The unified management system operation status acquisition module is used to obtain the unified management system operation status by realizing the unified management of unit start-up and shutdown and load regulation through the automatic control algorithm of the computer monitoring platform according to the startup control strategy. The support scheme acquisition module is used to determine the support scheme for the hydropower station based on the operating status of the unified management system and through the capacity calculation model. The trigger module is used to invoke and execute the corresponding support scheme when an abnormal state occurs in the power grid.

10. A computer device, characterized in that, The computer device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the load optimization and emergency support control method of the energy storage control system according to any one of claims 1-8.

Images