Grid-side energy storage optimization configuration and operation control method and system

Abstract

The present invention provides a grid-side energy storage optimization configuration and operation control method and system, which relates to the technical field of grid-side storage optimization, including obtaining historical grid operation data, generating energy storage configuration parameters based on node voltage over-limit risks and line power over-limit risks, and constructing a multi-time-scale energy storage control framework based on the obtained energy storage configuration parameters by adopting a distributed model predictive control algorithm, dividing the control framework into a day-ahead planning layer, an hourly optimization layer, and a minute-level real-time control layer, performing a fault safety check on the initial energy storage group control instructions, establishing a distributed adaptive fault diagnosis mechanism, and adopting a distributed consistency algorithm to perform online optimization on the charging and discharging power of the energy storage unit, generating an energy storage control instruction that takes fault avoidance into consideration, and controlling the energy storage unit to perform energy storage control operations in an adaptive collaborative manner in response to the energy storage control instruction.

Description

Technical Field

[0001] The present invention relates to grid-side energy storage optimization technology, and in particular to a grid-side energy storage optimization configuration and operation control method and system. Background Art

[0002] As the scale of new energy grid connection continues to expand, grid load fluctuations intensify, and the safe and stable operation of the power system faces severe challenges. As an important means of regulation, energy storage systems play a key role in peak shaving, frequency regulation, and voltage support. However, the configuration schemes of existing energy storage systems often lack comprehensive consideration of load characteristics and new energy penetration, resulting in low efficiency of energy storage investment.

[0003] At present, the operation and control of energy storage systems generally adopt a centralized control architecture, which has problems such as heavy communication burden and poor system scalability. In addition, there is a lack of effective fault diagnosis and fault-tolerant control mechanisms, making it difficult to ensure the safe and stable operation of energy storage groups. In particular, when a storage unit fails, the existing control strategy cannot adjust the control structure and power allocation scheme in time, affecting the reliability and economy of the system. Summary of the invention

[0004] The embodiments of the present invention provide a grid-side energy storage optimization configuration and operation control method and system, which can solve the problems in the prior art.

[0005] According to a first aspect of the embodiments of the present invention,

[0006] Provided is a grid-side energy storage optimization configuration and operation control method, comprising:

[0007] Obtain historical data of power grid operation, extract time series features of the historical data, establish a mathematical model of energy storage configuration that takes into account load fluctuation rate and new energy penetration rate, including a power grid operation benefit sub-model and an energy storage cost sub-model, and construct a hierarchical progressive energy storage optimization configuration strategy based on the energy storage configuration mathematical model. Determine the target layout nodes of the energy storage system by calculating the voltage sensitivity of each region, divide the energy storage response partitions based on the load distribution density, and establish an energy storage capacity allocation model that takes into account the power fluctuation transfer characteristics within the energy storage response partitions. Generate energy storage configuration parameters based on the node voltage over-limit risk and the line power over-limit risk, and the energy storage configuration parameters include energy storage capacity parameters and energy storage power parameters;

[0008] Based on the obtained energy storage configuration parameters, a distributed model predictive control algorithm is used to construct a multi-time scale energy storage control framework, and the control framework is divided into a day-ahead planning layer, an hourly optimization layer, and a minute-level real-time control layer. In the day-ahead planning layer, an energy storage charging and discharging plan is formulated according to the comprehensive load forecast results to generate a planned power curve; in the hourly optimization layer, the planned power curve is decomposed into a basic power target and an adjustment power target, and the adjustment power target is corrected based on an adaptive dynamic programming algorithm to generate an energy storage power adjustment instruction; in the minute-level real-time control layer, a distributed collaborative control strategy is adopted, and an energy storage group intelligent control model is established according to the energy storage power adjustment instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints;

[0009] A fault safety check is performed on the initial energy storage group control instruction, a distributed adaptive fault diagnosis mechanism is established, a fault-tolerant control strategy for the energy storage group is constructed through real-time monitoring of the energy storage unit status and fault feature identification, the energy storage group control topology and power allocation weight are dynamically adjusted according to the energy storage group fault-tolerant control strategy, and a distributed consistency algorithm is used to perform online optimization of the energy storage unit charging and discharging power, generate energy storage control instructions that take fault avoidance into account, and in response to the energy storage control instructions, control the energy storage unit to perform energy storage control operations in an adaptive collaborative manner.

[0010] In an optional embodiment,

[0011] A hierarchical progressive energy storage optimization configuration strategy is constructed based on the energy storage configuration mathematical model. The target layout nodes of the energy storage system are determined by calculating the voltage sensitivity of each region. The energy storage response partitions are divided according to the load distribution density. An energy storage capacity allocation model considering the power fluctuation transfer characteristics is established within the energy storage response partitions. The energy storage configuration parameters generated based on the node voltage over-limit risk and the line power over-limit risk include:

[0012] Calculate the voltage sensitivity matrix between the power grid nodes based on the energy storage configuration mathematical model, calculate the node importance index according to the voltage sensitivity matrix, the node importance index is determined by the sum of the absolute values ​​of the elements in each row of the voltage sensitivity matrix, and select the node with the highest node importance index as the target layout node of the energy storage system;

[0013] A storage power injection model is constructed at the target deployment node, an adaptive time window method is used to analyze the power fluctuation transfer characteristics, and a power fluctuation transfer coefficient in the energy storage charging and discharging process is calculated. The power fluctuation transfer coefficient characterizes the line power fluctuation change characteristics before and after the energy storage is connected. An energy storage capacity allocation model based on the power fluctuation transfer coefficient is established, and a node voltage over-limit risk assessment module is constructed using a hierarchical neural network. The node voltage over-limit risk assessment module calculates the voltage over-limit risk based on the node voltage dynamic response curve. A line power over-limit risk assessment module is constructed using a fuzzy fault tree. The line power over-limit risk assessment module calculates the power over-limit risk based on the line power dynamic response curve.

[0014] For the energy storage capacity allocation model, the multi-objective optimization theory is used to construct an objective function that takes into account the voltage over-limit risk, power over-limit risk and power fluctuation transfer coefficient, and the energy storage power limit constraints, energy storage capacity limit constraints and charge state limit constraints are set. The multi-objective optimization algorithm is used to solve the objective function and finally generate the energy storage configuration parameters.

[0015] In an optional embodiment,

[0016] For the energy storage capacity allocation model, the multi-objective optimization theory is used to construct an objective function that considers the voltage over-limit risk, power over-limit risk and power fluctuation transfer coefficient, and the energy storage power limit constraint, energy storage capacity limit constraint and charge state limit constraint are set. The multi-objective optimization algorithm is used to solve the objective function, and the energy storage configuration parameters are finally generated, including:

[0017] Constructing a multi-objective optimization objective function of an energy storage capacity allocation model, wherein the multi-objective optimization objective function includes a voltage over-limit risk, a power over-limit risk, and a power fluctuation transfer coefficient, and using a neural network to calculate a first weight coefficient of the voltage over-limit risk, a second weight coefficient of the power over-limit risk, and a third weight coefficient of the power fluctuation transfer coefficient;

[0018] Inputting the voltage over-limit risk, the power over-limit risk and the power fluctuation transfer coefficient into a multi-layer feedforward neural network to generate a weight adjustment amount, and adjusting the first weight coefficient, the second weight coefficient and the third weight coefficient according to the weight adjustment amount to obtain an updated multi-objective optimization objective function;

[0019] A long short-term memory network is used to predict the voltage over-limit risk, and an upper threshold value and a lower threshold value of energy storage power are determined based on the predicted voltage over-limit risk, so as to construct an energy storage power limit constraint;

[0020] A graph neural network is used to construct a spatiotemporal correlation model of power over-limit risk, and the system nodes are represented as a graph structure with a node feature matrix. The topological correlation features between nodes are extracted through the graph convolution layer to predict the power over-limit risk. The upper and lower thresholds of the energy storage capacity are set according to the power over-limit risk, and the energy storage capacity limit constraint is constructed.

[0021] Constructing a distributed energy storage control unit, collaboratively optimizing the power fluctuation transfer coefficients between the distributed energy storage control units, and constructing a state of charge limit constraint based on the collaboratively optimized power fluctuation transfer coefficients;

[0022] Inputting the updated multi-objective optimization objective function, energy storage power limit constraint, energy storage capacity limit constraint and state of charge limit constraint into a non-dominated sorting genetic algorithm model, generating an initial population by a tournament selection method, and optimizing the initial population by simulating a binary crossover operation and a polynomial mutation operation;

[0023] The crowding distance of each solution in the optimized population is calculated, the solution set is optimized based on the crowding distance, the non-dominated solution set is determined using the Pareto dominance relationship, the optimal solution is selected from the non-dominated solution set using a pre-constructed comprehensive evaluation index, and finally the energy storage configuration parameters are obtained.

[0024] In an optional embodiment,

[0025] A distributed collaborative control strategy is adopted in the minute-level real-time control layer, an energy storage group intelligent control model is established according to the energy storage power adjustment instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints, including:

[0026] The energy storage units are represented as vertex sets, the communication links between the energy storage units are represented as edge sets, and a communication topology weight matrix is ​​constructed based on the electrical distance between the energy storage units.

[0027] Calculating a Laplace matrix according to the communication topology weight matrix, obtaining a maximum eigenvalue of the Laplace matrix, and calculating a consistency gain coefficient based on the maximum eigenvalue;

[0028] A distributed consistency protocol is constructed using the communication topology weight matrix and the consistency gain coefficient, a control amount difference of adjacent energy storage units is calculated according to the distributed consistency protocol, and the control amount difference is combined with the target power deviation to generate the energy storage unit control amount;

[0029] A fuzzy neural network including an input layer, a fuzzy layer, a rule layer, a normalization layer and an output layer is constructed. The state of charge signal and the power change rate signal of the energy storage unit are taken as inputs, and the intermediate characteristic quantities are obtained by calculating the fuzzy neural network.

[0030] A power distribution coefficient is calculated based on the intermediate characteristic quantity, and a power instruction of the energy storage unit is obtained by multiplying the target power by the power distribution coefficient;

[0031] Determine whether the state of charge signal is within a preset charge range and whether the power change rate signal is less than a preset change threshold, generate a dispatchable state signal of the energy storage unit according to the judgment result, and combine the energy storage unit power instruction with the dispatchable state signal of the energy storage unit to generate a final control instruction of the energy storage unit.

[0032] In an optional embodiment,

[0033] Determining whether the state of charge signal is within a preset charge range and whether the power change rate signal is less than a preset change threshold, and generating a dispatchable state signal of the energy storage unit according to the determination result includes:

[0034] Acquire the operation time information of the energy storage unit, and calculate the upper limit and lower limit of the dynamic threshold of the state of charge according to the operation time information;

[0035] Collecting a real-time state of charge signal of the energy storage unit, and comparing the real-time state of charge signal with an upper limit of a dynamic state of charge threshold and a lower limit of a dynamic state of charge threshold, respectively, to obtain a first judgment result;

[0036] Calculate the state of charge deviation value and the state of charge change rate of the energy storage unit based on the real-time state of charge signal, and calculate the power change rate limit value according to the state of charge deviation value and the state of charge change rate;

[0037] Collecting a real-time power change rate signal of the energy storage unit, comparing the real-time power change rate signal with the power change rate limit value, and obtaining a second judgment result;

[0038] Establishing a fuzzy judgment rule based on the first judgment result and the second judgment result, and calculating the initial dispatchable state of the energy storage unit according to the fuzzy judgment rule;

[0039] Obtaining the ideal operating point parameters of the energy storage unit, calculating the deviation between the real-time state of charge signal and the ideal state of charge, and the deviation between the real-time power change rate signal and the ideal power change rate;

[0040] The state of charge deviation and the power change rate deviation are compared with the preset dead zone parameters, and it is determined whether to maintain the dispatchable state at the previous moment according to the comparison result, and a final dispatchable state signal of the energy storage unit is generated.

[0041] In an optional embodiment,

[0042] Perform a fault safety check on the initial energy storage group control command, establish a distributed adaptive fault diagnosis mechanism, build a fault-tolerant control strategy for the energy storage group through real-time monitoring of the energy storage unit status and fault feature identification, dynamically adjust the energy storage group control topology and power allocation weight according to the energy storage group fault-tolerant control strategy, and use a distributed consistency algorithm to optimize the charging and discharging power of the energy storage unit online, and generate energy storage control instructions that take fault avoidance into account, including:

[0043] Collecting a voltage signal, a current signal and a temperature signal of an energy storage unit, constructing a state observer based on the voltage signal, the current signal and the temperature signal, and obtaining a state observation value of the energy storage unit;

[0044] Calculating a residual vector according to the deviation between the energy storage unit state observation value and the measured signal, and extracting features from the residual vector to obtain fault features;

[0045] Acquire the real-time control quantity and output quantity of the energy storage unit, establish an adaptive threshold criterion according to the real-time control quantity and output quantity, compare the fault feature with the adaptive threshold criterion, and output a fault alarm signal;

[0046] Setting a fault confirmation time window, accumulating and counting the fault alarm signal within the fault confirmation time window, and confirming the occurrence of a fault and generating a fault flag signal when the accumulated count value exceeds a preset threshold;

[0047] Update the energy storage group adjacency matrix according to the location information of the fault sign signal, calculate the shortest path distance between each energy storage unit and the faulty unit based on the updated adjacency matrix, substitute the shortest path distance into the characteristic attenuation function, and calculate the power allocation weight coefficient in combination with the fault sign signal;

[0048] A distributed consistency optimization problem is constructed based on the updated adjacency matrix and the power allocation weight coefficient, and the optimized power instruction of each energy storage unit is obtained through iterative solution. The optimized power instruction is multiplied by the fault flag signal of the corresponding energy storage unit to generate an energy storage control instruction that takes fault avoidance into account.

[0049] In an optional embodiment,

[0050] A distributed consistency optimization problem is constructed based on the updated adjacency matrix and the power allocation weight coefficient, and the formula for the optimized power instruction of each energy storage unit is obtained by iterative solution as follows:

[0051] ;

[0052] in, represents the updated power instruction of energy storage unit i in the t+1th iteration, represents the power instruction of energy storage unit i in the tth iteration, α represents the iteration step, N(i) represents the neighbor set of energy storage unit i, and w ij represents the power distribution weight between energy storage unit i and energy storage unit j, represents the power command of energy storage unit j in the tth iteration, β i represents the adjustment factor of energy storage unit i, represents the gradient of the power optimization objective function J at the tth iteration.

[0053] According to a second aspect of the embodiments of the present invention,

[0054] Provided is a grid-side energy storage optimization configuration and operation control system, including:

[0055] The first unit is used to obtain historical data of power grid operation, extract time series features of the historical data, establish a mathematical model of energy storage configuration that takes into account load fluctuation rate and new energy penetration rate, including a power grid operation benefit sub-model and an energy storage cost sub-model, construct a hierarchical progressive energy storage optimization configuration strategy based on the energy storage configuration mathematical model, determine the target layout nodes of the energy storage system by calculating the voltage sensitivity of each region, divide the energy storage response partitions according to the load distribution density, establish an energy storage capacity allocation model that takes into account the power fluctuation transfer characteristics within the energy storage response partitions, and generate energy storage configuration parameters based on the node voltage over-limit risk and the line power over-limit risk, wherein the energy storage configuration parameters include energy storage capacity parameters and energy storage power parameters;

[0056] The second unit is used to construct a multi-time-scale energy storage control framework based on the obtained energy storage configuration parameters by using a distributed model predictive control algorithm, and divide the control framework into a day-ahead planning layer, an hourly optimization layer, and a minute-level real-time control layer. In the day-ahead planning layer, an energy storage charging and discharging plan is formulated according to the comprehensive load forecast results to generate a planned power curve; in the hourly optimization layer, the planned power curve is decomposed into a basic power target and an adjustment power target, and the adjustment power target is corrected based on an adaptive dynamic programming algorithm to generate an energy storage power adjustment instruction; in the minute-level real-time control layer, a distributed collaborative control strategy is adopted, an energy storage group intelligent control model is established according to the energy storage power adjustment instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints;

[0057] The third unit is used to perform a fault safety check on the initial energy storage group control instruction, establish a distributed adaptive fault diagnosis mechanism, build an energy storage group fault-tolerant control strategy through real-time monitoring of the energy storage unit status and fault feature identification, dynamically adjust the energy storage group control topology and power allocation weight according to the energy storage group fault-tolerant control strategy, and use a distributed consistency algorithm to perform online optimization of the energy storage unit charging and discharging power, generate energy storage control instructions that take fault avoidance into account, and in response to the energy storage control instructions, control the energy storage unit to perform energy storage control operations in an adaptive collaborative manner.

[0058] According to a third aspect of the embodiments of the present invention,

[0059] An electronic device is provided, comprising:

[0060] processor;

[0061] a memory for storing processor-executable instructions;

[0062] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0063] According to a fourth aspect of the embodiments of the present invention,

[0064] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0065] In this embodiment, the accurate arrangement of the energy storage system is achieved through a hierarchical and progressive energy storage optimization configuration strategy, the investment efficiency of energy storage is improved, a multi-time scale collaborative control framework is adopted to effectively cope with load and new energy fluctuations, and the safe and reliable operation of the energy storage group is guaranteed based on distributed fault diagnosis and fault-tolerant control strategies. The distributed consistency optimization algorithm is used to achieve rapid redistribution of energy storage power, shorten the fault response time and improve the regulation performance, thereby improving the economy and reliability of the energy storage system and providing effective technical support for large-scale energy storage access to the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A schematic diagram of a flow chart of a method for optimizing configuration and operation control of grid-side energy storage according to an embodiment of the present invention;

[0067] Figure 2 Schematic diagram of the structure of the grid-side energy storage optimization configuration and operation control system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0070] Figure 1 Schematic diagram of the flow of the grid-side energy storage optimization configuration and operation control method according to an embodiment of the present invention, such as Figure 1 As shown, the method includes:

[0071] S101. Obtain historical data on power grid operation, extract time series features from the historical data, establish a mathematical model for energy storage configuration that takes into account load fluctuation rate and new energy penetration rate, including a power grid operation benefit sub-model and an energy storage cost sub-model, and construct a hierarchical and progressive energy storage optimization configuration strategy based on the energy storage configuration mathematical model. Determine the target layout nodes of the energy storage system by calculating the voltage sensitivity of each region, divide the energy storage response partitions based on the load distribution density, and establish an energy storage capacity allocation model that takes into account the power fluctuation transfer characteristics within the energy storage response partitions. Generate energy storage configuration parameters based on the node voltage over-limit risk and the line power over-limit risk, and the energy storage configuration parameters include energy storage capacity parameters and energy storage power parameters.

[0072] First, historical data of power grid operation are obtained, including load curves, voltage curves, power curves, etc. Time series features are extracted from these historical data to analyze load fluctuation characteristics, voltage fluctuation characteristics, and power fluctuation characteristics. Based on the extracted time series features, a mathematical model of energy storage configuration that takes into account load fluctuation rate and new energy penetration rate is established. The model contains two sub-models: a power grid operation benefit sub-model and an energy storage cost sub-model. The power grid operation benefit sub-model considers the voltage regulation benefits, power regulation benefits, and peak shaving and valley filling benefits brought by energy storage. The energy storage cost sub-model considers the investment cost, operation and maintenance cost, and life loss cost of energy storage equipment.

[0073] According to the established mathematical model of energy storage configuration, a hierarchical progressive energy storage optimization configuration strategy is constructed. First, the target layout nodes of the energy storage system are determined by calculating the voltage sensitivity of each region. The specific method is to calculate the sensitivity of the voltage of each node to active / reactive injection, and select nodes with higher sensitivity as candidate nodes for energy storage deployment. Then, combined with the load distribution density, the power grid is divided into multiple energy storage response partitions. In each energy storage response partition, an energy storage capacity allocation model that considers the power fluctuation transfer characteristics is established. The model calculates the energy storage capacity required for each node based on the transfer characteristics of power fluctuations.

[0074] Finally, based on the node voltage over-limit risk and line power over-limit risk, energy storage configuration parameters are generated, including energy storage capacity parameters and energy storage power parameters. For example, the energy storage capacity parameter of a 110kV substation is 10MWh, and the power parameter is 5MW. These parameters serve as an important basis for the subsequent design and operation control of the energy storage system.

[0075] In an optional embodiment,

[0076] A hierarchical progressive energy storage optimization configuration strategy is constructed based on the energy storage configuration mathematical model. The target layout nodes of the energy storage system are determined by calculating the voltage sensitivity of each region. The energy storage response partitions are divided according to the load distribution density. An energy storage capacity allocation model considering the power fluctuation transfer characteristics is established within the energy storage response partitions. The energy storage configuration parameters generated based on the node voltage over-limit risk and the line power over-limit risk include:

[0077] Calculate the voltage sensitivity matrix between the power grid nodes based on the energy storage configuration mathematical model, calculate the node importance index according to the voltage sensitivity matrix, the node importance index is determined by the sum of the absolute values ​​of the elements in each row of the voltage sensitivity matrix, and select the node with the highest node importance index as the target layout node of the energy storage system;

[0078] A storage power injection model is constructed at the target deployment node, an adaptive time window method is used to analyze the power fluctuation transfer characteristics, and a power fluctuation transfer coefficient in the energy storage charging and discharging process is calculated. The power fluctuation transfer coefficient characterizes the line power fluctuation change characteristics before and after the energy storage is connected. An energy storage capacity allocation model based on the power fluctuation transfer coefficient is established, and a node voltage over-limit risk assessment module is constructed using a hierarchical neural network. The node voltage over-limit risk assessment module calculates the voltage over-limit risk based on the node voltage dynamic response curve. A line power over-limit risk assessment module is constructed using a fuzzy fault tree. The line power over-limit risk assessment module calculates the power over-limit risk based on the line power dynamic response curve.

[0079] For the energy storage capacity allocation model, the multi-objective optimization theory is used to construct an objective function that takes into account the voltage over-limit risk, power over-limit risk and power fluctuation transfer coefficient, and the energy storage power limit constraints, energy storage capacity limit constraints and charge state limit constraints are set. The multi-objective optimization algorithm is used to solve the objective function and finally generate the energy storage configuration parameters.

[0080] According to the energy storage configuration mathematical model, a hierarchical progressive energy storage optimization configuration strategy is constructed. First, the target layout nodes of the energy storage system are determined by calculating the voltage sensitivity of each region. Specifically, the voltage sensitivity matrix between the nodes of the power grid is calculated based on the energy storage configuration mathematical model. The matrix characterizes the sensitivity of the voltage of each node to power injection. Then, the node importance index is calculated according to the voltage sensitivity matrix. The index is determined by the sum of the absolute values ​​of the elements in each row of the voltage sensitivity matrix, which reflects the degree of influence of the node on the system voltage. The nodes with the highest node importance index are selected as the target layout nodes of the energy storage system. These nodes are usually located in weak areas of the power grid. The energy storage system is connected here to effectively improve the system voltage quality. For example, a 10kV distribution network has a total of 50 nodes. The calculated voltage sensitivity matrix is ​​a 50×50 matrix. The absolute values ​​of the elements in each row are summed to obtain the importance index of 50 nodes. The importance index of node 25 is the highest, which is 0.85. Therefore, node 25 is selected as the target layout node of the energy storage system.

[0081] Next, the energy storage response partitions are divided according to the load distribution density. According to the load distribution characteristics of the distribution network, the network is divided into several energy storage response partitions, and the load density in each partition is similar. Clustering algorithms such as the K-means algorithm can be used for partitioning, and the clustering features include node coordinates, load size, etc. For example, 50 nodes are divided into 5 energy storage response partitions, where the partition where node 25 is located contains 10 nodes and the load density is 2MW / km².

[0082] Within the determined energy storage response partition, an energy storage capacity allocation model that considers the power fluctuation transfer characteristics is established. First, an energy storage power injection model is constructed at the target layout node to describe the charging and discharging process of the energy storage system. Then, the adaptive time window method is used to analyze the power fluctuation transfer characteristics, and the power fluctuation transfer coefficient during the energy storage charging and discharging process is calculated. This coefficient characterizes the line power fluctuation change characteristics before and after the energy storage is connected, and reflects the energy storage system's suppression effect on power fluctuations. For example, a 1-hour power injection simulation is performed on node 25, and a 5-minute sliding time window is used. The power fluctuation transfer coefficient is calculated to be 0.75, indicating that the energy storage system can suppress 75% of power fluctuations.

[0083] Based on the power fluctuation transfer coefficient, an energy storage capacity allocation model is established. The model aims to minimize the system voltage deviation and power fluctuation, and the constraints include energy storage power limit, capacity limit and state of charge limit. In order to evaluate the risk of energy storage configuration scheme, a hierarchical neural network is used to construct a node voltage over-limit risk assessment module. This module calculates the voltage over-limit risk based on the node voltage dynamic response curve, reflecting the probability and degree of voltage exceeding the allowable range. For example, a 24-hour voltage simulation is performed on node 25, and a 3-layer LSTM network is used for voltage prediction. The calculated voltage over-limit risk is 0.05, indicating that the voltage over-limit risk is low.

[0084] At the same time, a fuzzy fault tree is used to construct a line power over-limit risk assessment module. This module calculates the power over-limit risk based on the line power dynamic response curve, reflecting the possibility of line overload. For example, a 24-hour power simulation is performed on the line connected to node 25, and a three-layer fuzzy fault tree model is constructed. The power over-limit risk is calculated to be 0.08, indicating that the line overload risk is low.

[0085] Finally, for the energy storage capacity allocation model, the multi-objective optimization theory is used to construct an objective function that considers the voltage over-limit risk, power over-limit risk and power fluctuation transfer coefficient. Set energy storage power limit constraints, energy storage capacity limit constraints and state of charge limit constraints to ensure that the energy storage system operates within a safe range. Use a multi-objective optimization algorithm such as NSGA-II to solve the objective function and obtain a set of Pareto optimal solutions. Select the solution with the best comprehensive performance as the final energy storage configuration plan to generate energy storage configuration parameters. For example, for the energy storage system of node 25, the final configuration parameters are: rated power 1MW, rated capacity 2MWh, maximum charge and discharge power ±1.2MW, and state of charge range 20%-90%.

[0086] Through the above-mentioned hierarchical and progressive optimization configuration strategy, the reasonable layout and capacity allocation of energy storage systems in the distribution network can be achieved, effectively improving the voltage quality and power balance capability of the system, and reducing the risk of voltage over-limit and line overload. This method fully considers the grid topology, load distribution characteristics and power fluctuation transfer rules, and can adapt to the energy storage optimization configuration requirements of different types of distribution networks.

[0087] In an optional embodiment,

[0088] For the energy storage capacity allocation model, the multi-objective optimization theory is used to construct an objective function that considers the voltage over-limit risk, power over-limit risk and power fluctuation transfer coefficient, and the energy storage power limit constraint, energy storage capacity limit constraint and charge state limit constraint are set. The multi-objective optimization algorithm is used to solve the objective function, and the energy storage configuration parameters are finally generated, including:

[0089] Constructing a multi-objective optimization objective function of an energy storage capacity allocation model, wherein the multi-objective optimization objective function includes a voltage over-limit risk, a power over-limit risk, and a power fluctuation transfer coefficient, and using a neural network to calculate a first weight coefficient of the voltage over-limit risk, a second weight coefficient of the power over-limit risk, and a third weight coefficient of the power fluctuation transfer coefficient;

[0090] Inputting the voltage over-limit risk, the power over-limit risk and the power fluctuation transfer coefficient into a multi-layer feedforward neural network to generate a weight adjustment amount, and adjusting the first weight coefficient, the second weight coefficient and the third weight coefficient according to the weight adjustment amount to obtain an updated multi-objective optimization objective function;

[0091] A long short-term memory network is used to predict the voltage over-limit risk, and an upper threshold value and a lower threshold value of energy storage power are determined based on the predicted voltage over-limit risk, so as to construct an energy storage power limit constraint;

[0092] A graph neural network is used to construct a spatiotemporal correlation model of power over-limit risk, and the system nodes are represented as a graph structure with a node feature matrix. The topological correlation features between nodes are extracted through the graph convolution layer to predict the power over-limit risk. The upper and lower thresholds of the energy storage capacity are set according to the power over-limit risk, and the energy storage capacity limit constraint is constructed.

[0093] Constructing a distributed energy storage control unit, collaboratively optimizing the power fluctuation transfer coefficients between the distributed energy storage control units, and constructing a state of charge limit constraint based on the collaboratively optimized power fluctuation transfer coefficients;

[0094] Inputting the updated multi-objective optimization objective function, energy storage power limit constraint, energy storage capacity limit constraint and state of charge limit constraint into a non-dominated sorting genetic algorithm model, generating an initial population by a tournament selection method, and optimizing the initial population by simulating a binary crossover operation and a polynomial mutation operation;

[0095] The crowding distance of each solution in the optimized population is calculated, the solution set is optimized based on the crowding distance, the non-dominated solution set is determined using the Pareto dominance relationship, the optimal solution is selected from the non-dominated solution set using a pre-constructed comprehensive evaluation index, and finally the energy storage configuration parameters are obtained.

[0096] In a specific implementation, a multi-objective optimization objective function of the energy storage capacity allocation model is first constructed. The objective function includes three indicators: voltage over-limit risk, power over-limit risk, and power fluctuation transfer coefficient. A neural network is used to calculate the weight coefficients of these three indicators. Specifically, a three-layer feedforward neural network is constructed, the input layer includes voltage, power, and fluctuation-related features in historical data, the hidden layer uses a ReLU activation function, and the output layer corresponds to three weight coefficients. The network is trained by a back-propagation algorithm to obtain the final weight coefficient.

[0097] Next, the voltage over-limit risk, power over-limit risk and power fluctuation transfer coefficient are input into the multi-layer feedforward neural network to generate weight adjustment. The network contains 5 hidden layers, with 64, 32, 16, 8 and 4 neurons in each layer. The previously calculated weight coefficient is fine-tuned according to the generated adjustment to obtain the updated objective function. For example, if the adjustment is [-0.05, 0.03, 0.02], the corresponding weight coefficient is adjusted to [w1-0.05, w2+0.03, w3+0.02].

[0098] Then, a long short-term memory network (LSTM) is used to predict the voltage over-limit risk. The LSTM network consists of an LSTM layer (128 units) and a fully connected layer. The input is the historical voltage data of the sliding window, and the output is the predicted value of the voltage over-limit risk in the next 24 hours. Based on the prediction results, the upper and lower limits of the energy storage power are set, such as the upper limit is 1.2 times the predicted maximum value, and the lower limit is 0.8 times the predicted minimum value. In this way, the energy storage power limit constraint is constructed.

[0099] A graph neural network is used to construct a spatiotemporal correlation model of power over-limit risk. Each node of the system is represented as a graph structure, and the node features include power, voltage and other information. A two-layer graph convolutional network is constructed, with 64 convolution kernels in each layer, to extract the topological correlation features between nodes. The last layer is fully connected to output the power over-limit risk prediction value. The upper and lower limits of the energy storage capacity are set according to the prediction results, such as the upper limit is 1.5 times the predicted maximum value, and the lower limit is 0.5 times the predicted minimum value, to construct the energy storage capacity limit constraint.

[0100] Construct a distributed energy storage control unit to collaboratively optimize the power fluctuation transfer coefficient. The control unit adopts the Actor-Critic architecture. The Actor network outputs the power regulation strategy of each unit, and the Critic network evaluates the strategy value. Through reinforcement learning, the strategy is continuously optimized, and finally the power fluctuation transfer coefficient after collaborative optimization is obtained. Based on this, the state of charge limit constraint is constructed, such as setting the upper and lower limits of the state of charge to 95% and 5%.

[0101] The updated objective function and constraints are input into the non-dominated sorting genetic algorithm (NSGA-II) model. Binary coding is used, the population size is set to 100, and the evolutionary generations are 200. The initial population is generated by tournament selection. The simulated binary crossover probability is set to 0.9, and the polynomial mutation probability is 0.1. The population is optimized and iterated, and the crowding distance of each solution is calculated. Excellent individuals are selected based on non-dominated sorting and crowding comparison. Finally, comprehensive evaluation indicators such as energy storage configuration cost and system reliability are constructed to select the optimal solution from the non-dominated solution set to obtain the final energy storage configuration parameters.

[0102] In this embodiment, firstly, through the hierarchical and progressive energy storage optimization configuration strategy, the stability and safety of the power grid can be effectively improved, and the risk of voltage and power over-limit can be reduced. Secondly, the accurate voltage sensitivity calculation and the evaluation of the node importance index make the layout of the energy storage system more reasonable, giving priority to the nodes with the greatest impact on the power grid. In addition, the adaptive time window method analyzes the power fluctuation transfer characteristics, which helps to dynamically adjust the energy storage charging and discharging strategy and optimize the power fluctuation management. Finally, combined with the multi-objective optimization theory, the voltage and power over-limit risks and power fluctuation characteristics are comprehensively considered to ensure that the energy storage system can maximize the economic benefits while meeting the technical constraints, thereby improving the overall operation efficiency and reliability of the power grid.

[0103] S102. Based on the obtained energy storage configuration parameters, a distributed model predictive control algorithm is used to construct a multi-time scale energy storage control framework, and the control framework is divided into a day-ahead planning layer, an hourly optimization layer and a minute-level real-time control layer. In the day-ahead planning layer, an energy storage charging and discharging plan is formulated according to the comprehensive load forecast results to generate a planned power curve; in the hourly optimization layer, the planned power curve is decomposed into a basic power target and a regulated power target, and the regulated power target is corrected based on an adaptive dynamic programming algorithm to generate an energy storage power regulation instruction; in the minute-level real-time control layer, a distributed collaborative control strategy is adopted, and an energy storage group intelligent control model is established according to the energy storage power regulation instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints.

[0104] After obtaining the energy storage configuration parameters, a distributed model predictive control algorithm is used to construct a multi-time scale energy storage control framework. The framework is divided into three layers: day-ahead planning layer, hourly optimization layer, and minute-level real-time control layer. In the day-ahead planning layer, an energy storage charging and discharging plan is formulated based on the comprehensive load forecast results to generate a planned power curve. For example, if the load peak is predicted to occur between 14:00 and 16:00 on a certain day, energy storage discharge is arranged during this period, and charging is arranged at other times.

[0105] The hourly optimization layer decomposes the planned power curve into basic power targets and adjustment power targets. The adjustment power target is modified based on the adaptive dynamic programming algorithm to generate energy storage power adjustment instructions. For example, the original plan was to discharge 2MW of energy storage from 14:00 to 15:00, but the actual load was lower than expected, so it was corrected to discharge 1.5MW. The minute-level real-time control layer adopts a distributed collaborative control strategy to establish an intelligent control model for the energy storage group based on the energy storage power adjustment instructions. The model takes into account the state of charge and power change rate constraints of the energy storage unit to generate the initial energy storage group control instructions.

[0106] In an optional embodiment,

[0107] A distributed collaborative control strategy is adopted in the minute-level real-time control layer, an energy storage group intelligent control model is established according to the energy storage power adjustment instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints, including:

[0108] The energy storage units are represented as vertex sets, the communication links between the energy storage units are represented as edge sets, and a communication topology weight matrix is ​​constructed based on the electrical distance between the energy storage units.

[0109] Calculating a Laplace matrix according to the communication topology weight matrix, obtaining a maximum eigenvalue of the Laplace matrix, and calculating a consistency gain coefficient based on the maximum eigenvalue;

[0110] A distributed consistency protocol is constructed using the communication topology weight matrix and the consistency gain coefficient, a control amount difference of adjacent energy storage units is calculated according to the distributed consistency protocol, and the control amount difference is combined with the target power deviation to generate the energy storage unit control amount;

[0111] A fuzzy neural network including an input layer, a fuzzy layer, a rule layer, a normalization layer and an output layer is constructed. The state of charge signal and the power change rate signal of the energy storage unit are taken as inputs, and the intermediate characteristic quantities are obtained by calculating the fuzzy neural network.

[0112] A power distribution coefficient is calculated based on the intermediate characteristic quantity, and a power instruction of the energy storage unit is obtained by multiplying the target power by the power distribution coefficient;

[0113] Determine whether the state of charge signal is within a preset charge range and whether the power change rate signal is less than a preset change threshold, generate a dispatchable state signal of the energy storage unit according to the judgment result, and combine the energy storage unit power instruction with the dispatchable state signal of the energy storage unit to generate a final control instruction of the energy storage unit.

[0114] The distributed collaborative control strategy is adopted at the minute-level real-time control layer. The intelligent control model of the energy storage group is established according to the energy storage power regulation instruction. The specific implementation method of generating the initial energy storage group control instruction based on the energy storage unit charge state and power change rate constraint is as follows:

[0115] First, each energy storage unit in the energy storage system is represented as a vertex set, and the communication link between the energy storage units is represented as an edge set. The communication topology weight matrix is ​​constructed based on the electrical distance between the energy storage units. Specifically, the electrical distance between any two energy storage units can be calculated based on the physical position relationship of the energy storage units in the power grid, and the inverse of the electrical distance can be used as the weight of the communication link to obtain the communication topology weight matrix. For example, for a system containing 5 energy storage units, a 5x5 weight matrix can be obtained.

[0116] Next, the Laplace matrix is ​​calculated based on the above communication topology weight matrix, the maximum eigenvalue of the Laplace matrix is ​​obtained, and the consistency gain coefficient is calculated based on the maximum eigenvalue. The Laplace matrix can be obtained by subtracting the degree matrix from the weight matrix, and then the power iteration method is used to solve the maximum eigenvalue. The consistency gain coefficient can be set to the inverse of the maximum eigenvalue.

[0117] Then, a distributed consistency protocol is constructed using the communication topology weight matrix and the consistency gain coefficient. The control quantity difference of adjacent energy storage units is calculated according to the protocol, and the control quantity difference is combined with the target power deviation to generate the energy storage unit control quantity. Specifically, each energy storage unit calculates the control quantity difference between itself and the adjacent unit based on the communication information with the adjacent unit, and performs a weighted summation with the target power deviation issued by the system to obtain its own control quantity.

[0118] Next, a fuzzy neural network consisting of an input layer, a fuzzy layer, a rule layer, a normalization layer, and an output layer is constructed. The state of charge signal and the power change rate signal of the energy storage unit are used as inputs, and the intermediate feature quantity is calculated through the fuzzy neural network. The input layer receives two input signals, the state of charge and the power change rate. The fuzzy layer performs fuzzification on the input signal. The rule layer performs reasoning according to the fuzzy rules. The normalization layer normalizes the output of the rule layer. Finally, the output layer obtains the intermediate feature quantity.

[0119] The power allocation coefficient is calculated based on the above intermediate characteristic quantity, and the target power is multiplied by the power allocation coefficient to obtain the power instruction of the energy storage unit. The power allocation coefficient can be obtained by nonlinear mapping of the intermediate characteristic quantity to ensure that the power allocation of each energy storage unit is reasonable.

[0120] Finally, determine whether the state of charge signal is within the preset charge range and whether the power change rate signal is less than the preset change threshold, and generate the energy storage unit dispatchable state signal based on the judgment result. For example, the state of charge range can be set to 20%-80%, and the power change rate threshold can be set to 10% / min of the rated power. Combine the energy storage unit power command and the dispatchable state signal to generate the final control command of the energy storage unit. If the energy storage unit is in a dispatchable state, it is executed according to the power command; otherwise, the control command is set to 0.

[0121] In this embodiment, first, the distributed collaborative control strategy enables the energy storage unit to respond in real time at the minute level, improving the flexibility and dynamic adjustment capability of the system. By constructing an intelligent control model, the state of charge and power change rate of the energy storage unit are effectively considered, thereby optimizing the overall performance and coordination of the energy storage group. Secondly, the application of the communication topology weight matrix and the consistency protocol ensures the efficient transmission of information between the energy storage units, promotes the consistency of the control quantity, and improves the stability of the system. The introduction of fuzzy neural networks enhances the processing capability of the state of charge and power change rate signals, allowing more accurate power allocation and further improving the scheduling efficiency. Finally, by judging the dispatchable state signal, it is ensured that the energy storage unit operates within a safe and effective operating range, maximizing energy utilization and improving the economy and reliability of the entire energy storage system.

[0122] In an optional embodiment,

[0123] Determining whether the state of charge signal is within a preset charge range and whether the power change rate signal is less than a preset change threshold, and generating a dispatchable state signal of the energy storage unit according to the determination result includes:

[0124] Acquire the operation time information of the energy storage unit, and calculate the upper limit and lower limit of the dynamic threshold of the state of charge according to the operation time information;

[0125] Collecting a real-time state of charge signal of the energy storage unit, and comparing the real-time state of charge signal with an upper limit of a dynamic state of charge threshold and a lower limit of a dynamic state of charge threshold, respectively, to obtain a first judgment result;

[0126] Calculate the state of charge deviation value and the state of charge change rate of the energy storage unit based on the real-time state of charge signal, and calculate the power change rate limit value according to the state of charge deviation value and the state of charge change rate;

[0127] Collecting a real-time power change rate signal of the energy storage unit, comparing the real-time power change rate signal with the power change rate limit value, and obtaining a second judgment result;

[0128] Establishing a fuzzy judgment rule based on the first judgment result and the second judgment result, and calculating the initial dispatchable state of the energy storage unit according to the fuzzy judgment rule;

[0129] Obtaining the ideal operating point parameters of the energy storage unit, calculating the deviation between the real-time state of charge signal and the ideal state of charge, and the deviation between the real-time power change rate signal and the ideal power change rate;

[0130] The state of charge deviation and the power change rate deviation are compared with the preset dead zone parameters, and it is determined whether to maintain the dispatchable state at the previous moment according to the comparison result, and a final dispatchable state signal of the energy storage unit is generated.

[0131] In a specific implementation, the following steps may be used to determine the dispatchable state of the energy storage unit:

[0132] First, obtain the operating time information of the energy storage unit, including the cumulative operating time, the number of charge and discharge cycles, etc. Based on this time information and the performance curve of the energy storage unit, calculate the upper and lower limits of the dynamic threshold of the state of charge at the current moment. For example, for a lithium battery energy storage system that has been running for 2 years and cycled 500 times, the upper limit of the dynamic threshold of the state of charge may be 95% and the lower limit may be 10%.

[0133] Next, the real-time state of charge signal of the energy storage unit is collected, and it is assumed that the current state of charge is 80%. 80% is compared with 95% and 10% respectively, and the first judgment result is obtained: the state of charge is within the dynamic threshold range.

[0134] Then, based on the real-time state of charge of 80%, the state of charge deviation and change rate are calculated. Assuming the ideal state of charge is 50%, the deviation value is 30%. The state of charge change rate is calculated by continuous sampling to be 2% / minute. Based on these two parameters, the current power change rate limit is 5% / minute.

[0135] Collect the real-time power change rate signal of the energy storage unit, assuming it is 3% / minute. Compare 3% / minute with 5% / minute, and get the second judgment result: the power change rate is less than the limit value.

[0136] Based on the above two judgment results, a fuzzy judgment rule is established. For example: if the state of charge is within the range and the power change rate is less than the limit, the dispatchability is "high"; if only one condition is met, it is "medium"; if neither condition is met, it is "low". According to the rule calculation, the initial dispatchability state is "high".

[0137] Get the ideal operating point parameters of the energy storage unit, for example, the ideal state of charge is 50%, and the ideal power change rate is 1% / minute. Calculate the deviation between the actual value and the ideal value, the state of charge deviation is 30%, and the power change rate deviation is 2% / minute.

[0138] The two deviation values ​​are compared with the preset dead zone parameters (such as 10% dead zone for state of charge and 1% / minute dead zone for power change rate). Since the deviation exceeds the dead zone, the state at the previous moment is not maintained, and the final dispatchable state signal is generated as "high".

[0139] In this embodiment, first, by setting the dynamic threshold, the system can flexibly adapt to the change of the operating time of the energy storage unit, thereby improving the real-time monitoring and management capabilities of the state of charge. Secondly, the comparison of the real-time state of charge signal with the dynamic threshold ensures that the energy storage unit operates within a safe range and reduces the risk of over-charging and discharging. In addition, the calculation of the power change rate limit and the comparison of the real-time signal help to timely identify power fluctuations and enhance the responsiveness of the system. The application of fuzzy judgment rules improves the processing accuracy of complex states and makes the generation of the initial dispatchable state more reasonable. Finally, by calculating the ideal operating point parameters and comparing the dead zone parameters, the system can effectively determine whether to maintain the dispatchable state, ensure that the energy storage unit operates under optimized operating conditions, and improve the stability and economy of the overall system. This series of measures jointly promotes the intelligence and efficiency of the energy storage system.

[0140] S103. Perform a fault safety check on the initial energy storage group control instruction, establish a distributed adaptive fault diagnosis mechanism, build an energy storage group fault-tolerant control strategy through real-time monitoring of the energy storage unit status and fault feature identification, dynamically adjust the energy storage group control topology and power allocation weight according to the energy storage group fault-tolerant control strategy, and use a distributed consistency algorithm to perform online optimization of the energy storage unit charging and discharging power, generate energy storage control instructions that take fault avoidance into account, and in response to the energy storage control instructions, control the energy storage unit to perform energy storage control operations in an adaptive collaborative manner.

[0141] Perform fault safety check on the initial energy storage group control instructions and establish a distributed adaptive fault diagnosis mechanism. Build a fault-tolerant control strategy for the energy storage group through real-time monitoring of the energy storage unit status and fault feature identification. This strategy can dynamically adjust the control topology and power allocation weight of the energy storage group. For example, if a certain energy storage unit fails, it will be removed from the control network and its power task will be reallocated. A distributed consensus algorithm is used to optimize the charging and discharging power of the energy storage unit online, and generate energy storage control instructions that take fault avoidance into account.

[0142] Finally, in response to the generated energy storage control instructions, the energy storage units are controlled to perform energy storage control operations in an adaptive collaborative manner. For example, the control instructions generated at a certain moment are: energy storage unit 1 charges 0.5MW, energy storage unit 2 discharges 1MW, and energy storage unit 3 is on standby. Each energy storage unit performs the corresponding charging and discharging operations according to the instructions to achieve the optimized operation of the grid-side energy storage system.

[0143] In an optional embodiment,

[0144] Perform a fault safety check on the initial energy storage group control command, establish a distributed adaptive fault diagnosis mechanism, build a fault-tolerant control strategy for the energy storage group through real-time monitoring of the energy storage unit status and fault feature identification, dynamically adjust the energy storage group control topology and power allocation weight according to the energy storage group fault-tolerant control strategy, and use a distributed consistency algorithm to optimize the charging and discharging power of the energy storage unit online, and generate energy storage control instructions that take fault avoidance into account, including:

[0145] Collecting a voltage signal, a current signal and a temperature signal of an energy storage unit, constructing a state observer based on the voltage signal, the current signal and the temperature signal, and obtaining a state observation value of the energy storage unit;

[0146] Calculating a residual vector according to the deviation between the energy storage unit state observation value and the measured signal, and extracting features from the residual vector to obtain fault features;

[0147] Acquire the real-time control quantity and output quantity of the energy storage unit, establish an adaptive threshold criterion according to the real-time control quantity and output quantity, compare the fault feature with the adaptive threshold criterion, and output a fault alarm signal;

[0148] Setting a fault confirmation time window, accumulating and counting the fault alarm signal within the fault confirmation time window, and confirming the occurrence of a fault and generating a fault flag signal when the accumulated count value exceeds a preset threshold;

[0149] Update the energy storage group adjacency matrix according to the location information of the fault sign signal, calculate the shortest path distance between each energy storage unit and the faulty unit based on the updated adjacency matrix, substitute the shortest path distance into the characteristic attenuation function, and calculate the power allocation weight coefficient in combination with the fault sign signal;

[0150] A distributed consistency optimization problem is constructed based on the updated adjacency matrix and the power allocation weight coefficient, and the optimized power instruction of each energy storage unit is obtained through iterative solution. The optimized power instruction is multiplied by the fault flag signal of the corresponding energy storage unit to generate an energy storage control instruction that takes fault avoidance into account.

[0151] In this embodiment, the voltage signal, current signal and temperature signal of the energy storage unit are first collected. The voltage, current and temperature data of the energy storage unit can be collected by a voltage sensor, a current sensor and a temperature sensor respectively. For example, for a lithium-ion battery energy storage unit, its terminal voltage, charge and discharge current and battery surface temperature can be collected. The sampling frequency can be set to 1Hz.

[0152] Next, a state observer is constructed based on the collected voltage, current and temperature signals to obtain the state observation value of the energy storage unit. The state observer can be designed using an extended Kalman filter, with the battery equivalent circuit model as the observer model. The state variables of the observer include the state of charge (SOC), internal resistance, etc. The estimated values ​​of these state variables, namely the state observation values, can be obtained through iterative calculation.

[0153] Then, the state observation value is compared with the measured signal to calculate the residual vector. For example, the difference between the SOC estimate and the SOC value obtained by the ampere-hour integration method can be calculated as the SOC residual. The residual vector is feature extracted to obtain the fault feature. The feature extraction method can use the principal component analysis method to extract the main feature vector as the fault feature.

[0154] Next, the real-time control quantity (such as charging and discharging power instructions) and output quantity (such as actual output power) of the energy storage unit are obtained, and an adaptive threshold judgment criterion is established based on the control quantity and output quantity. The sliding time window method can be used to calculate the mean and variance of the control quantity and output quantity within a certain time window, and the adaptive threshold is set accordingly. The extracted fault features are compared with the adaptive threshold, and a fault alarm signal is output if the threshold is exceeded.

[0155] In order to avoid false alarms, a fault confirmation time window is set, such as 60s. The fault alarm signal is counted cumulatively within this time window. When the cumulative count value exceeds the preset threshold (such as 50 times), the fault is confirmed to have occurred and a fault flag signal is generated. The fault flag signal contains the number information of the faulty unit.

[0156] Update the energy storage group adjacency matrix according to the fault flag signal. For example, if unit 1 fails, set the element related to unit 1 in the adjacency matrix to 0. Based on the updated adjacency matrix, use the Dijkstra algorithm to calculate the shortest path distance between each energy storage unit and the faulty unit. Substitute the shortest path distance into the characteristic attenuation function (such as the exponential attenuation function) to obtain the distance attenuation coefficient. Multiply the distance attenuation coefficient by the fault flag signal to obtain the power allocation weight coefficient.

[0157] Finally, based on the updated adjacency matrix and power allocation weight coefficient, a distributed consistency optimization problem is constructed. The optimization goal is minimizing Σ(Pi - P_ref)^2, where Pi is the power of each energy storage unit and P_ref is the reference power. Constraints include power balance constraints and unit power limits. By iteratively solving the optimization problem, the optimized power instructions for each energy storage unit are obtained. The optimized power instructions are multiplied by the fault flag signal of the corresponding unit to generate the final energy storage control instructions that take fault avoidance into account.

[0158] In an optional embodiment,

[0159] A distributed consistency optimization problem is constructed based on the updated adjacency matrix and the power allocation weight coefficient, and the formula for the optimized power instruction of each energy storage unit is obtained by iterative solution as follows:

[0160] ;

[0161] in, represents the updated power instruction of energy storage unit i in the t+1th iteration, represents the power instruction of energy storage unit i in the tth iteration, α represents the iteration step, N(i) represents the neighbor set of energy storage unit i, and w ij represents the power distribution weight between energy storage unit i and energy storage unit j, represents the power command of energy storage unit j in the tth iteration, β i represents the adjustment factor of energy storage unit i, represents the gradient of the power optimization objective function J at the tth iteration.

[0162] In this embodiment, first, the distributed adaptive fault diagnosis mechanism can monitor the state of the energy storage unit in real time, identify potential faults in a timely manner, and enhance the safety and reliability of the system. Secondly, by constructing a state observer and analyzing the residual vector, the system can efficiently extract fault features, achieve accurate fault alarms, and prevent fault expansion. In addition, the set fault confirmation time window and cumulative counting mechanism ensure the accuracy of fault confirmation, thereby reducing the risk of false alarms. The generation of fault flag signals and the updating of the adjacency matrix support the dynamic adjustment of the control topology of the energy storage group, optimize power distribution, and ensure the stable operation of the system in the event of a fault. Finally, the charging and discharging power is optimized online through a distributed consistency algorithm, and the generated control instructions can effectively avoid the impact of faults, improve the overall dispatching efficiency and emergency response capabilities of the energy storage system, and thus optimize the reliability and performance of the power grid.

[0163] Figure 2 Schematic diagram of the structure of the grid-side energy storage optimization configuration and operation control system according to an embodiment of the present invention. Figure 2As shown, the system comprises:

[0164] The first unit is used to obtain historical data of power grid operation, extract time series features of the historical data, establish a mathematical model of energy storage configuration that takes into account load fluctuation rate and new energy penetration rate, including a power grid operation benefit sub-model and an energy storage cost sub-model, construct a hierarchical progressive energy storage optimization configuration strategy based on the energy storage configuration mathematical model, determine the target layout nodes of the energy storage system by calculating the voltage sensitivity of each region, divide the energy storage response partitions according to the load distribution density, establish an energy storage capacity allocation model that takes into account the power fluctuation transfer characteristics within the energy storage response partitions, and generate energy storage configuration parameters based on the node voltage over-limit risk and the line power over-limit risk, wherein the energy storage configuration parameters include energy storage capacity parameters and energy storage power parameters;

[0165] The second unit is used to construct a multi-time-scale energy storage control framework based on the obtained energy storage configuration parameters by using a distributed model predictive control algorithm, and divide the control framework into a day-ahead planning layer, an hourly optimization layer, and a minute-level real-time control layer. In the day-ahead planning layer, an energy storage charging and discharging plan is formulated according to the comprehensive load forecast results to generate a planned power curve; in the hourly optimization layer, the planned power curve is decomposed into a basic power target and an adjustment power target, and the adjustment power target is corrected based on an adaptive dynamic programming algorithm to generate an energy storage power adjustment instruction; in the minute-level real-time control layer, a distributed collaborative control strategy is adopted, an energy storage group intelligent control model is established according to the energy storage power adjustment instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints;

[0166] The third unit is used to perform a fault safety check on the initial energy storage group control instruction, establish a distributed adaptive fault diagnosis mechanism, build an energy storage group fault-tolerant control strategy through real-time monitoring of the energy storage unit status and fault feature identification, dynamically adjust the energy storage group control topology and power allocation weight according to the energy storage group fault-tolerant control strategy, and use a distributed consistency algorithm to perform online optimization of the energy storage unit charging and discharging power, generate energy storage control instructions that take fault avoidance into account, and in response to the energy storage control instructions, control the energy storage unit to perform energy storage control operations in an adaptive collaborative manner.

[0167] According to a third aspect of the embodiments of the present invention,

[0168] An electronic device is provided, comprising:

[0169] processor;

[0170] a memory for storing processor-executable instructions;

[0171] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0172] According to a fourth aspect of the embodiments of the present invention,

[0173] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0174] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0175] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A grid-side energy storage optimization configuration and operation control method, characterized in that: include: Obtain historical data of power grid operation, extract time series features of the historical data, establish a mathematical model of energy storage configuration that takes into account load fluctuation rate and new energy penetration rate, including a power grid operation benefit sub-model and an energy storage cost sub-model, and construct a hierarchical progressive energy storage optimization configuration strategy based on the energy storage configuration mathematical model. Determine the target layout nodes of the energy storage system by calculating the voltage sensitivity of each region, divide the energy storage response partitions based on the load distribution density, and establish an energy storage capacity allocation model that takes into account the power fluctuation transfer characteristics within the energy storage response partitions. Generate energy storage configuration parameters based on the node voltage over-limit risk and the line power over-limit risk, and the energy storage configuration parameters include energy storage capacity parameters and energy storage power parameters; Based on the obtained energy storage configuration parameters, a distributed model predictive control algorithm is used to construct a multi-time scale energy storage control framework, and the control framework is divided into a day-ahead planning layer, an hourly optimization layer, and a minute-level real-time control layer. In the day-ahead planning layer, an energy storage charging and discharging plan is formulated according to the comprehensive load forecast results to generate a planned power curve; in the hourly optimization layer, the planned power curve is decomposed into a basic power target and an adjustment power target, and the adjustment power target is corrected based on an adaptive dynamic programming algorithm to generate an energy storage power adjustment instruction; in the minute-level real-time control layer, a distributed collaborative control strategy is adopted, and an energy storage group intelligent control model is established according to the energy storage power adjustment instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints; A fault safety check is performed on the initial energy storage group control instruction, a distributed adaptive fault diagnosis mechanism is established, a fault-tolerant control strategy for the energy storage group is constructed through real-time monitoring of the energy storage unit status and fault feature identification, the energy storage group control topology and power allocation weight are dynamically adjusted according to the energy storage group fault-tolerant control strategy, and a distributed consistency algorithm is used to perform online optimization of the energy storage unit charging and discharging power, generate energy storage control instructions that take fault avoidance into account, and in response to the energy storage control instructions, control the energy storage unit to perform energy storage control operations in an adaptive collaborative manner.

2. The method according to claim 1, characterized in that A hierarchical progressive energy storage optimization configuration strategy is constructed based on the energy storage configuration mathematical model. The target layout nodes of the energy storage system are determined by calculating the voltage sensitivity of each region. The energy storage response partitions are divided according to the load distribution density. An energy storage capacity allocation model considering the power fluctuation transfer characteristics is established within the energy storage response partitions. The energy storage configuration parameters generated based on the node voltage over-limit risk and the line power over-limit risk include: Calculate the voltage sensitivity matrix between the power grid nodes based on the energy storage configuration mathematical model, calculate the node importance index according to the voltage sensitivity matrix, the node importance index is determined by the sum of the absolute values ​​of the elements in each row of the voltage sensitivity matrix, and select the node with the highest node importance index as the target layout node of the energy storage system; Constructing an energy storage power injection model at the target deployment node, using an adaptive time window method to analyze the power fluctuation transfer characteristics, calculating the power fluctuation transfer coefficient during the energy storage charging and discharging process, the power fluctuation transfer coefficient characterizing the line power fluctuation change characteristics before and after the energy storage is connected, establishing an energy storage capacity allocation model based on the power fluctuation transfer coefficient, using a hierarchical neural network to construct a node voltage over-limit risk assessment module, the node voltage over-limit risk assessment module calculates the voltage over-limit risk based on the node voltage dynamic response curve, and using a fuzzy fault tree to construct a line power over-limit risk assessment module, the line power over-limit risk assessment module calculates the power over-limit risk based on the line power dynamic response curve; For the energy storage capacity allocation model, the multi-objective optimization theory is used to construct an objective function that takes into account the voltage over-limit risk, power over-limit risk and power fluctuation transfer coefficient, and the energy storage power limit constraints, energy storage capacity limit constraints and charge state limit constraints are set. The multi-objective optimization algorithm is used to solve the objective function and finally generate the energy storage configuration parameters.

3. The method according to claim 2, characterized in that For the energy storage capacity allocation model, the multi-objective optimization theory is used to construct an objective function that considers the voltage over-limit risk, power over-limit risk and power fluctuation transfer coefficient, and the energy storage power limit constraint, energy storage capacity limit constraint and charge state limit constraint are set. The multi-objective optimization algorithm is used to solve the objective function, and the energy storage configuration parameters are finally generated, including: Constructing a multi-objective optimization objective function of an energy storage capacity allocation model, wherein the multi-objective optimization objective function includes a voltage over-limit risk, a power over-limit risk, and a power fluctuation transfer coefficient, and using a neural network to calculate a first weight coefficient of the voltage over-limit risk, a second weight coefficient of the power over-limit risk, and a third weight coefficient of the power fluctuation transfer coefficient; Inputting the voltage over-limit risk, the power over-limit risk and the power fluctuation transfer coefficient into a multi-layer feedforward neural network to generate a weight adjustment amount, and adjusting the first weight coefficient, the second weight coefficient and the third weight coefficient according to the weight adjustment amount to obtain an updated multi-objective optimization objective function; A long short-term memory network is used to predict the voltage over-limit risk, and an upper threshold value and a lower threshold value of energy storage power are determined based on the predicted voltage over-limit risk, so as to construct an energy storage power limit constraint; A graph neural network is used to construct a spatiotemporal correlation model of power over-limit risk, and the system nodes are represented as a graph structure with a node feature matrix. The topological correlation features between nodes are extracted through the graph convolution layer to predict the power over-limit risk. The upper and lower thresholds of the energy storage capacity are set according to the power over-limit risk, and the energy storage capacity limit constraint is constructed. Constructing a distributed energy storage control unit, collaboratively optimizing the power fluctuation transfer coefficients between the distributed energy storage control units, and constructing a state of charge limit constraint based on the collaboratively optimized power fluctuation transfer coefficients; Inputting the updated multi-objective optimization objective function, energy storage power limit constraint, energy storage capacity limit constraint and state of charge limit constraint into a non-dominated sorting genetic algorithm model, generating an initial population by a tournament selection method, and optimizing the initial population by simulating a binary crossover operation and a polynomial mutation operation; The crowding distance of each solution in the optimized population is calculated, the solution set is optimized based on the crowding distance, the non-dominated solution set is determined using the Pareto dominance relationship, the optimal solution is selected from the non-dominated solution set using a pre-constructed comprehensive evaluation index, and finally the energy storage configuration parameters are obtained.

4. The method according to claim 1, characterized in that: A distributed collaborative control strategy is adopted in the minute-level real-time control layer, an energy storage group intelligent control model is established according to the energy storage power adjustment instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints, including: The energy storage units are represented as vertex sets, the communication links between the energy storage units are represented as edge sets, and a communication topology weight matrix is ​​constructed based on the electrical distance between the energy storage units. Calculating a Laplace matrix according to the communication topology weight matrix, obtaining a maximum eigenvalue of the Laplace matrix, and calculating a consistency gain coefficient based on the maximum eigenvalue; A distributed consistency protocol is constructed using the communication topology weight matrix and the consistency gain coefficient, a control amount difference of adjacent energy storage units is calculated according to the distributed consistency protocol, and the control amount difference is combined with the target power deviation to generate the energy storage unit control amount; A fuzzy neural network including an input layer, a fuzzy layer, a rule layer, a normalization layer and an output layer is constructed. The state of charge signal and the power change rate signal of the energy storage unit are taken as inputs, and the intermediate characteristic quantities are obtained by calculating the fuzzy neural network. A power distribution coefficient is calculated based on the intermediate characteristic quantity, and a power instruction of the energy storage unit is obtained by multiplying the target power by the power distribution coefficient; Determine whether the state of charge signal is within a preset charge range and whether the power change rate signal is less than a preset change threshold, generate a dispatchable state signal of the energy storage unit according to the judgment result, and combine the energy storage unit power instruction with the dispatchable state signal of the energy storage unit to generate a final control instruction of the energy storage unit.

5. The method according to claim 4, characterized in that Determining whether the state of charge signal is within a preset charge range and whether the power change rate signal is less than a preset change threshold, and generating a dispatchable state signal of the energy storage unit according to the determination result includes: Acquire the operation time information of the energy storage unit, and calculate the upper limit and lower limit of the dynamic threshold of the state of charge according to the operation time information; Collecting a real-time state of charge signal of the energy storage unit, and comparing the real-time state of charge signal with an upper limit of a dynamic state of charge threshold and a lower limit of a dynamic state of charge threshold, respectively, to obtain a first judgment result; Calculate the state of charge deviation value and the state of charge change rate of the energy storage unit based on the real-time state of charge signal, and calculate the power change rate limit value according to the state of charge deviation value and the state of charge change rate; Collecting a real-time power change rate signal of the energy storage unit, comparing the real-time power change rate signal with the power change rate limit value, and obtaining a second judgment result; Establishing a fuzzy judgment rule based on the first judgment result and the second judgment result, and calculating the initial dispatchable state of the energy storage unit according to the fuzzy judgment rule; Obtaining the ideal operating point parameters of the energy storage unit, calculating the deviation between the real-time state of charge signal and the ideal state of charge, and the deviation between the real-time power change rate signal and the ideal power change rate; The state of charge deviation and the power change rate deviation are compared with the preset dead zone parameters, and it is determined whether to maintain the dispatchable state at the previous moment according to the comparison result, and a final dispatchable state signal of the energy storage unit is generated.

6. The method according to claim 1, characterized in that Perform a fault safety check on the initial energy storage group control command, establish a distributed adaptive fault diagnosis mechanism, build a fault-tolerant control strategy for the energy storage group through real-time monitoring of the energy storage unit status and fault feature identification, dynamically adjust the energy storage group control topology and power allocation weight according to the energy storage group fault-tolerant control strategy, and use a distributed consistency algorithm to optimize the charging and discharging power of the energy storage unit online, and generate energy storage control instructions that take fault avoidance into account, including: Collecting a voltage signal, a current signal and a temperature signal of an energy storage unit, constructing a state observer based on the voltage signal, the current signal and the temperature signal, and obtaining a state observation value of the energy storage unit; Calculating a residual vector according to the deviation between the energy storage unit state observation value and the measured signal, and performing feature extraction on the residual vector to obtain a fault feature; Acquire the real-time control quantity and output quantity of the energy storage unit, establish an adaptive threshold criterion according to the real-time control quantity and output quantity, compare the fault feature with the adaptive threshold criterion, and output a fault alarm signal; Setting a fault confirmation time window, accumulating and counting the fault alarm signal within the fault confirmation time window, and confirming the occurrence of a fault and generating a fault flag signal when the accumulated count value exceeds a preset threshold; Update the energy storage group adjacency matrix according to the location information of the fault sign signal, calculate the shortest path distance between each energy storage unit and the faulty unit based on the updated adjacency matrix, substitute the shortest path distance into the characteristic attenuation function, and calculate the power allocation weight coefficient in combination with the fault sign signal; A distributed consistency optimization problem is constructed based on the updated adjacency matrix and the power allocation weight coefficient, and the optimized power instruction of each energy storage unit is obtained through iterative solution. The optimized power instruction is multiplied by the fault flag signal of the corresponding energy storage unit to generate an energy storage control instruction that takes fault avoidance into account.

7. The method according to claim 6, characterized in that A distributed consistency optimization problem is constructed based on the updated adjacency matrix and the power allocation weight coefficient, and the formula for the optimized power instruction of each energy storage unit is obtained by iterative solution as follows: ； in, represents the updated power instruction of energy storage unit i in the t+1th iteration, represents the power instruction of energy storage unit i in the tth iteration, α represents the iteration step, N(i) represents the neighbor set of energy storage unit i, and w ij represents the power distribution weight between energy storage unit i and energy storage unit j, represents the power command of energy storage unit j in the tth iteration, β i represents the adjustment factor of energy storage unit i, represents the gradient of the power optimization objective function J at the tth iteration.

8. A grid-side energy storage optimization configuration and operation control system, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to obtain historical data of power grid operation, extract time series features of the historical data, establish a mathematical model of energy storage configuration that takes into account load fluctuation rate and new energy penetration rate, including a power grid operation benefit sub-model and an energy storage cost sub-model, construct a hierarchical progressive energy storage optimization configuration strategy based on the energy storage configuration mathematical model, determine the target layout nodes of the energy storage system by calculating the voltage sensitivity of each region, divide the energy storage response partitions according to the load distribution density, establish an energy storage capacity allocation model that takes into account the power fluctuation transfer characteristics within the energy storage response partitions, and generate energy storage configuration parameters based on the node voltage over-limit risk and the line power over-limit risk, wherein the energy storage configuration parameters include energy storage capacity parameters and energy storage power parameters; The second unit is used to construct a multi-time-scale energy storage control framework based on the obtained energy storage configuration parameters by using a distributed model predictive control algorithm, and divide the control framework into a day-ahead planning layer, an hourly optimization layer, and a minute-level real-time control layer. In the day-ahead planning layer, an energy storage charging and discharging plan is formulated according to the comprehensive load forecast results to generate a planned power curve; in the hourly optimization layer, the planned power curve is decomposed into a basic power target and an adjustment power target, and the adjustment power target is corrected based on an adaptive dynamic programming algorithm to generate an energy storage power adjustment instruction; in the minute-level real-time control layer, a distributed collaborative control strategy is adopted, an energy storage group intelligent control model is established according to the energy storage power adjustment instruction, and an initial energy storage group control instruction is generated based on the energy storage unit charge state and power change rate constraints; The third unit is used to perform a fault safety check on the initial energy storage group control instruction, establish a distributed adaptive fault diagnosis mechanism, build an energy storage group fault-tolerant control strategy through real-time monitoring of the energy storage unit status and fault feature identification, dynamically adjust the energy storage group control topology and power allocation weight according to the energy storage group fault-tolerant control strategy, and use a distributed consistency algorithm to perform online optimization of the energy storage unit charging and discharging power, generate energy storage control instructions that take fault avoidance into account, and in response to the energy storage control instructions, control the energy storage unit to perform energy storage control operations in an adaptive collaborative manner.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images