Intelligent control and power regulation method of string type network configuration type energy storage system

Abstract

This invention relates to the field of energy storage system control technology, and discloses an intelligent control and power regulation method for a string-type grid-connected energy storage system, comprising: S1, acquiring grid status and mode-defined demand; S2, acquiring cluster-level topology mapping; S3, collecting cluster state variables of each energy storage cluster; S4, determining capacity estimates and uncertainty metrics, and shrinking or correcting capacity boundary terms; S5, performing inter-cluster collaborative allocation based on support demand and available capacity to obtain support shares; S6, mapping control parameters and controlling converters; S7, event-triggered reconfiguration and reallocation. By introducing a safe availability capacity assessment mechanism based on cluster-level capacity estimation, uncertainty metrics, and boundary shrinkage, the method achieves precise on-demand supply of grid support while meeting current, voltage, and temperature constraints, significantly reducing the risks of overload, exceeding limits, and misjudgment, and improving system safety and support effectiveness.

Description

Technical Field

[0001] This invention relates to the field of energy storage system control technology, specifically to an intelligent control and power regulation method for a string-grid type energy storage system. Background Technology

[0002] String-type grid-connected energy storage systems typically consist of multiple energy storage clusters, grid-connected converters, and their controllable channels. They provide active and reactive power support to the grid or load through energy storage units, achieving stable voltage and frequency control. These systems are designed for scenarios such as weak grid support, islanded operation, and black start. They can provide rapid power response when there are external grid disturbances or load changes. To meet the support requirements under different operating conditions, it is usually necessary to assess the status and capabilities of each energy storage cluster and coordinate the power allocation of multiple clusters to complete power regulation and stable support during grid-connected operation. In the engineering implementation of existing string-type grid-connected energy storage systems, power limiting and allocation methods based on fixed thresholds or empirical margins are mostly adopted. The output capabilities of each energy storage cluster are estimated based on rated parameters or simplified models, and power commands are issued according to preset rules during operation.

[0003] However, in current technologies, power limiting and allocation methods based on fixed thresholds or empirical margins are difficult to reflect the actual available support capacity of each energy storage cluster under different conditions and rapid changes in operating conditions in a timely and accurate manner. This can easily lead to capacity assessment bias, resulting in overly conservative power regulation or the risk of exceeding limits when constraints are approaching, which affects the safety and effectiveness of grid support. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent control and power regulation method for string-type grid-type energy storage systems, which solves the problem that existing methods are unable to accurately and timely assess the available support capacity of energy storage clusters, leading to conservative power regulation or the risk of exceeding limits.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for intelligent control and power regulation of a string-grid energy storage system, comprising: S1. Obtain power grid operation status information and system operation mode information, and determine the grid construction support requirements accordingly; S2. Obtain cluster-level topology information of the string-type grid-type energy storage system to determine the correspondence between multiple energy storage clusters and grid-type converters and their independently controllable channels. S3. Collect the cluster state parameters of each energy storage cluster, including state of charge, health status, temperature, DC side voltage and equivalent internal resistance. S4. Evaluate the available support capacity of each energy storage cluster based on the cluster state variables. The evaluation includes determining the estimated support capacity and its corresponding uncertainty measure, and shrinking or correcting the capacity boundary term corresponding to the estimated value based on the uncertainty measure. S5. Based on the network support requirements and the assessment results of available support capabilities, perform inter-cluster collaborative allocation to determine the support share of each energy storage cluster. The support share includes the share of fast active power support, the share of fast reactive power support, the share of current limiting capability, and the share of recovery slope. S6. Map the support share of each energy storage cluster to the control parameters of the corresponding grid-type converter. The control parameters include current-limiting trajectory parameters and recovery trajectory parameters. Control the corresponding grid-type converter based on the control parameters. S7. Detect the operating events of each energy storage cluster and trigger reconfiguration control, update the available support capacity assessment results and the support share, and re-execute steps S5 to S6.

[0006] Preferably, S1 includes: Acquire grid frequency information, grid voltage information, and system operating mode identifier; Based on the system operation mode identifier, the target network construction support requirement is selected from the preset network construction support requirement set; The target network support requirements are parameterized to obtain network support requirement parameters, which include fast active power support requirement parameters, fast reactive power support requirement parameters, current limiting support requirement parameters, and recovery slope requirement parameters.

[0007] Preferably, S2 includes: Obtain the energy storage cluster identifier, the grid-type converter identifier, and the identifier of the independently controllable channel; Establish a correspondence table between energy storage clusters, independently controllable channels, and grid-type converters; The control object of each energy storage cluster is determined based on the correspondence table.

[0008] Preferably, S3 includes: Collect the state of charge, health status, temperature, DC side voltage, and equivalent internal resistance of each energy storage cluster; Perform time alignment processing on the collected data and remove outliers; Generate a set of cluster state variables for assessing available support capabilities.

[0009] Preferably, S4 includes: Based on the cluster state variables, the estimated support capacity of each energy storage cluster is calculated. The estimated support capacity includes the estimated fast active power support capacity, the estimated fast reactive power support capacity, the estimated current limiting capacity, and the estimated recovery slope capacity. Based on current constraints, DC voltage constraints, and temperature constraints, determine the capability boundary terms corresponding to the estimated support capability; Output the estimated support capacity of each energy storage cluster and the corresponding capacity boundary terms.

[0010] Preferably, the shrinking or correction of the capability boundary term based on the uncertainty metric includes: Calculate or obtain the uncertainty measure corresponding to the estimated support capacity of each energy storage cluster; The capability boundary term is shrunk or corrected based on the uncertainty metric to obtain the available support capability boundary of each energy storage cluster. The available support capability boundary is used as the constraint input for the inter-cluster collaborative allocation.

[0011] Preferably, S5 includes: The support share of each energy storage cluster is determined under the constraint that the support share of each energy storage cluster does not exceed the corresponding available support capacity boundary. The combined support of each energy storage cluster covers the network support requirements. When the synthetic support cannot cover the network support requirements, the support share of each energy storage cluster is determined according to the preset gap cost rule.

[0012] Preferably, S6 includes: Current-limiting trajectory parameters and recovery trajectory parameters are generated based on the support share of each energy storage cluster; wherein, the current-limiting trajectory parameters and recovery trajectory parameters are generated or updated based on the available support capacity boundary, so that the current-limiting trajectory parameters and recovery trajectory parameters satisfy the constraints of the available support capacity boundary; Current limiting control is implemented on the current reference of the corresponding grid-type converter based on the current limiting trajectory parameters; Based on the recovery trajectory parameters, recovery control is implemented for the active and reactive power references of the corresponding grid-type converter.

[0013] Preferably, S7 includes: For each energy storage cluster, detect at least one of the following events: continuous current limiting event, DC undervoltage approach event, temperature rise rate exceeding limit event, energy storage cluster exit event, and communication anomaly event; The reconstruction control is triggered when the detected quantity meets the preset threshold condition. Record the energy storage cluster identifier and trigger reason identifier of the triggering event, and update the uncertainty metric and the available support capacity boundary after triggering the reconfiguration control.

[0014] Preferably, the uncertainty measure is obtained using the Monte Carlo random deactivation method, and the support capability estimation model is a neural network model containing a random deactivation layer. The Monte Carlo random deactivation method includes: While keeping the structure of the support capability estimation model unchanged, the support capability estimation model is made to randomly deactivate some neural network units with a preset deactivation probability during multiple inference processes; Multiple inferences are performed on the same cluster of state variables to obtain multiple sets of support capability estimation outputs; The output dispersion is calculated based on the multiple sets of support capability estimation outputs, and the output dispersion is used as the uncertainty measure.

[0015] This invention provides an intelligent control and power regulation method for a string-grid energy storage system. It offers the following advantages: 1. This invention introduces a safe availability capability assessment mechanism based on cluster-level capability estimation, uncertainty measurement, and boundary contraction. Under the premise of meeting current, voltage, and temperature constraints, it enables precise on-demand supply of network support, significantly reducing the risks of overload, exceeding limits, and misjudgment, and improving system safety and support effectiveness.

[0016] 2. This invention establishes a topological correspondence between clusters, channels, and converters, and performs time alignment and outlier removal on multi-source state data to achieve unified perception and consistent control of multi-cluster states, thereby improving the executability, noise resistance, and overall robustness of control commands.

[0017] 3. The present invention adopts an inter-cluster collaborative power allocation and gap cost strategy, which prioritizes the protection of critical support needs and suppresses gap propagation when support capacity is insufficient, while taking into account the constraint margin of each cluster and balanced utilization, thereby improving the utilization rate of the system's available capacity and extending its lifespan.

[0018] 4. This invention converts the collaborative allocation result into control parameters for the current limiting trajectory and the recovery trajectory, and triggers online reconstruction and redistribution based on events such as continuous current limiting, undervoltage approach, excessive temperature rise rate, and cluster exit, so as to achieve rapid recovery and stable operation under disturbance. Attached Figure Description

[0019] Figure 1 This is a flowchart of an intelligent control and power regulation method for a string-type grid-type energy storage system according to the present invention. Detailed Implementation

[0020] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see the appendix Figure 1 This invention provides an intelligent control and power regulation method for a string-grid energy storage system, comprising: S1. Obtain power grid operation status information and system operation mode information, and determine the grid construction support requirements accordingly; Furthermore, S1 includes: Acquire grid frequency information, grid voltage information, and system operating mode identifier; Based on the system operation mode identifier, the target network construction support requirement is selected from the preset network construction support requirement set; The target network support requirements are parameterized to obtain network support requirement parameters, which include fast active power support requirement parameters, fast reactive power support requirement parameters, current limiting support requirement parameters, and recovery slope requirement parameters.

[0022] Specifically, the grid frequency and voltage information are first obtained, and the system operation mode identifier is obtained. The grid frequency and voltage information are used to characterize the current grid operation status, and the system operation mode identifier is used to characterize the current system operation scenario. Based on the above information, the process of determining the grid support requirements begins. Subsequently, based on the system operation mode identifier, the target network construction support requirement corresponding to the operation mode is selected from the preset network construction support requirement set. The network construction support requirement set is pre-configured with support requirement items under different operation modes, so that the selection of the target network construction support requirement is consistent with the system operation mode. After selecting the target network support requirements, the target network support requirements are parameterized to obtain network support requirement parameters. These parameters are used as input conditions for inter-cluster collaborative allocation in subsequent steps and to constrain the determination range and allocation direction of subsequent support shares. The network support requirement parameters include fast active power support requirement parameters, fast reactive power support requirement parameters, current limiting support requirement parameters, and recovery slope requirement parameters. For example, when the system operation mode identifier corresponds to the weak network support mode, after obtaining the grid frequency information and grid voltage information, the target network support requirement corresponding to this operation mode is selected from the network support requirement set, and the target network support requirement is parameterized into fast active power support requirement parameters, fast reactive power support requirement parameters, current limiting support requirement parameters, and recovery slope requirement parameters for use in subsequent steps.

[0023] S2. Obtain cluster-level topology information of the string-type grid-type energy storage system to determine the correspondence between multiple energy storage clusters and grid-type converters and their independently controllable channels. Furthermore, S2 includes: Obtain the energy storage cluster identifier, the grid-type converter identifier, and the identifier of the independently controllable channel; Establish a correspondence table between energy storage clusters, independently controllable channels, and grid-type converters; The control objects of each energy storage cluster are determined based on the correspondence table.

[0024] Specifically, after the grid support requirements are determined, in order to ensure that the subsequent status acquisition, support capability assessment and collaborative allocation of each energy storage cluster can be implemented on specific execution objects, it is necessary to first identify the grid-type converter and its independently controllable channel corresponding to each energy storage cluster, thereby forming the object relationship of cluster-level control. First, obtain the energy storage cluster identifier, the grid-type converter identifier, and the independently controllable channel identifier. The energy storage cluster identifier is used to distinguish different energy storage clusters, the grid-type converter identifier is used to distinguish different grid-type converters, and the independently controllable channel identifier is used to distinguish channels that can be controlled separately within the same grid-type converter. The above identifiers can be provided by pre-configured system parameters or by existing equipment information records in the system, and are used as the basic data for establishing the corresponding relationship in the future.

[0025] Subsequently, a correspondence table is established based on the identifiers between energy storage clusters, independently controllable channels, and grid-type converters. The correspondence table contains at least the associated record of "energy storage cluster identifier - independently controllable channel identifier - grid-type converter identifier", which is used to record which converter and which channel controls each energy storage cluster. Through the correspondence table, the support share allocation result of a certain energy storage cluster can be consistently mapped with the corresponding converter control parameter generation process in subsequent steps, avoiding inconsistencies between the cluster allocation result and the execution object. After establishing the correspondence table, the control objects of each energy storage cluster are determined based on the correspondence table. The control objects include at least the grid-type converter identifier and the independently controllable channel identifier corresponding to the energy storage cluster. This enables subsequent steps to issue the current limiting trajectory parameters and recovery trajectory parameters corresponding to the support share for each energy storage cluster, and to accurately locate the energy storage cluster and its corresponding control channel when the operation event triggers reconfiguration.

[0026] For example, a certain string-grid energy storage system includes three energy storage clusters, denoted as cluster 1, cluster 2, and cluster 3. The system includes one grid-type converter with three independently controllable channels, denoted as channel A, channel B, and channel C. The obtained identifiers are: energy storage cluster identifiers are {cluster 1, cluster 2, cluster 3}, grid-type converter identifiers are {converter 1}, and independently controllable channel identifiers are {channel A, channel B, channel C}. Based on this, a corresponding relationship table is established: cluster 1—channel A—converter 1, cluster 2—channel B—converter 1, cluster 3—channel C—converter 1. When the support share is subsequently allocated collaboratively, the support share of cluster 1 is used to generate the control parameters of channel A, the support share of cluster 2 is used to generate the control parameters of channel B, and the support share of cluster 3 is used to generate the control parameters of channel C, thereby ensuring the consistency between cluster-level allocation and channel-level control.

[0027] S3. Collect the cluster state parameters of each energy storage cluster, including state of charge, health status, temperature, DC side voltage and equivalent internal resistance. Furthermore, S3 includes: Collect the state of charge, health status, temperature, DC side voltage, and equivalent internal resistance of each energy storage cluster; Perform time alignment processing on the collected data and remove outliers; Generate a set of cluster state variables for assessing available support capabilities.

[0028] Specifically, after the control objects of each energy storage cluster have been identified, in order to carry out the subsequent assessment of available support capabilities, it is necessary to collect and organize the cluster state variables of each energy storage cluster so that they can be used as a unified input for support capability estimation and uncertainty measurement. First, the state of charge, health status, temperature, DC-side voltage, and equivalent internal resistance of each energy storage cluster are collected. The state of charge and health status are used to characterize the energy status and aging degree of the energy storage cluster, the temperature is used to characterize the thermal status, the DC-side voltage is used to characterize the available DC-side voltage conditions, and the equivalent internal resistance is used to characterize the output capability and loss characteristics. The above cluster state quantities can be obtained from the existing cluster-level information records in the system or from the state estimation results corresponding to the energy storage cluster. The original data sequences are formed according to the energy storage cluster identifier so that the various state quantities of the same energy storage cluster can be processed consistently in the future. Subsequently, the collected data is time-aligned and outliers are removed. Time alignment is used to pair the state of charge, health, temperature, DC voltage and equivalent internal resistance of the same energy storage cluster at the same evaluation time to the same time reference, so as to avoid inconsistencies in input caused by different sampling times. Outlier removal is used to exclude data points that deviate significantly from the normal range or have abnormal jumps, so as to prevent them from participating in the subsequent assessment of available support capabilities. Time alignment can be achieved by synchronous reading under a unified sampling period or by interpolation / holding after timestamp matching. Outlier removal can be achieved by threshold discrimination or rate of change discrimination. For example, when a certain state quantity exceeds the preset physical range or the change in adjacent periods exceeds the preset threshold, the data is marked as an anomaly and removed or replaced with the most recent valid value. After completing time alignment and outlier processing, a cluster state input set is generated for assessing available support capacity. The cluster state input set is organized according to the energy storage cluster identifier. The state of charge, health status, temperature, DC side voltage and equivalent internal resistance of the same energy storage cluster at the same assessment time are combined into a set of inputs. These are used in subsequent steps to calculate the estimated support capacity and further perform uncertainty measurement and capacity boundary term shrinkage / correction, so that the subsequent assessment and allocation process adopts a consistent data caliber. For example, the system includes cluster 1 and cluster 2. At a certain evaluation time, the state of charge of cluster 1 is read as 0.62, the health status as 0.95, the temperature as 35°C, the DC side voltage as 720V, and the equivalent internal resistance as 2.1mΩ. At the same time, the state of charge of cluster 2 is read as 0.58, the health status as 0.90, the temperature as 40°C, the DC side voltage as 715V, and the equivalent internal resistance as 2.6mΩ. If a sudden change in the temperature of cluster 2 exceeding a preset change threshold is detected in an adjacent period, the temperature value is marked as abnormal and replaced with the most recent effective temperature value. Then, the processed state variables of cluster 1 and cluster 2 are respectively composed into cluster state variable input sets for subsequent use in the evaluation of available support capabilities.

[0029] S4. Evaluate the available support capacity of each energy storage cluster based on the cluster state variables. The evaluation includes determining the estimated support capacity and its corresponding uncertainty measure, and shrinking or correcting the capacity boundary terms corresponding to the estimated value based on the uncertainty measure. Furthermore, S4 includes: The estimated support capacity of each energy storage cluster is calculated based on the cluster state variables. The estimated support capacity includes the estimated fast active power support capacity, the estimated fast reactive power support capacity, the estimated current limiting capacity, and the estimated recovery slope capacity. The capability boundary terms corresponding to the estimated support capability are determined based on current constraints, DC voltage constraints, and temperature constraints. Output the estimated support capacity of each energy storage cluster and the corresponding capacity boundary terms.

[0030] Specifically, after obtaining the cluster state input set for evaluation, the available support capacity of each energy storage cluster is evaluated in order to provide a basic quantity of the support capacity of each energy storage cluster in the current state for subsequent inter-cluster collaborative allocation. Calculate the estimated support capacity of each energy storage cluster based on cluster state variables: For each energy storage cluster Using the cluster state variable input set as input, the estimated support capacity is calculated. This estimated support capacity includes fast active power support capacity, fast reactive power support capacity, current limiting capacity, and recovery slope capacity. The fast active power support capacity estimate characterizes the active power support capacity that the energy storage cluster can provide during short-term support; the fast reactive power support capacity estimate characterizes the reactive power support capacity that the energy storage cluster can provide; the current limiting capacity estimate characterizes the upper limit of current support for the corresponding channel of the energy storage cluster; and the recovery slope capacity estimate characterizes the permissible rate of change of active / reactive power for the energy storage cluster during the recovery phase. The estimated support capacity can be calculated by a pre-established capacity estimation model. The input to the capacity estimation model includes at least the state of charge, health state, temperature, DC-side voltage, and equivalent internal resistance. The output is the above four types of capacity estimates, thus relating the support capacity estimate to the energy state, thermal state, and electrical characteristics of the energy storage cluster. The cluster state variable vector is as follows: ; in, Indicates energy storage cluster The vector of estimated support capabilities, This is a rapid estimate of active power support capacity. This is an estimate of the rapid reactive power support capability. This is an estimate of the current limiting capacity. , These are the estimated values ​​of the active and reactive power recovery slope capabilities, respectively. For the capacity estimation model, The cluster state vector, The charged state is derived from the cluster state estimation results. The health status is derived from the health assessment results. Temperature, derived from temperature measurements. This is the DC-side voltage, derived from a DC voltage measurement. The equivalent internal resistance is derived from the internal resistance estimation result; Among them, the capacity estimation model For a pre-defined neural network model, the model input uses the aforementioned cluster state vector. The model output uses the aforementioned support capacity estimate vector. The parameters of the neural network model can be obtained through offline training. The training samples can come from historical running data, experimental calibration data or simulation data. After training, the model parameters are fixed and used for inference output during the running phase. In order to cooperate with the subsequent acquisition of uncertainty measurement, when the Monte Carlo random deactivation method is adopted, a random deactivation layer is set in the neural network model and the random deactivation layer is kept in the active state during the inference phase. Capacity boundary terms corresponding to the estimated support capacity are determined based on current constraints, DC voltage constraints, and temperature constraints: After obtaining the estimated support capacity, to ensure that the subsequent allocation process meets the safety operation constraints of the energy storage cluster, capacity boundary terms are further determined based on current constraints, DC voltage constraints, and temperature constraints. Current constraints limit the maximum allowable current of the corresponding channel of the energy storage cluster; DC voltage constraints limit the DC side voltage to a preset threshold; and temperature constraints limit the temperature to a preset upper limit and can further limit the temperature rise rate. The above constraints form the upper bounds of the boundaries corresponding to rapid active power support, rapid reactive power support, current limiting capability, and recovery slope capability, denoted as the capacity boundary term vector, as shown in the following equation: ; in, Indicates energy storage cluster The capability boundary term vector determined by current constraints, DC voltage constraints, and temperature constraints; , , , , These correspond to the upper bounds of the boundary terms for fast active power, fast reactive power, current limiting capability, active power recovery slope, and reactive power recovery slope, respectively. Their values ​​are calculated based on the equipment's rated parameters, protection thresholds, and corrections based on cluster state variables. Output the estimated support capacity of each energy storage cluster and the corresponding capability boundary terms: In this embodiment, the estimated support capacity vector of each energy storage cluster is output. and its corresponding capability boundary term vector The output is used as input for uncertainty measurement and capacity boundary term contraction / correction in subsequent steps, so that subsequent steps can form an available support capacity boundary under the joint constraints of "estimated capacity" and "constraint boundary", and further be used for inter-cluster collaborative allocation. Capability Boundary Term Vector The current-limiting capacity boundary term is determined by a combination of current constraints, DC voltage constraints, and temperature constraints. The current constraint uses the maximum allowable current threshold of the corresponding channel of the energy storage cluster. The DC voltage constraint uses the constraint that the DC voltage is not lower than a preset threshold. The temperature constraint uses the constraint that the temperature does not exceed a preset upper limit and the optional temperature rise rate does not exceed a preset upper limit. Based on the above constraints, the current-limiting capacity boundary term can be directly determined by the maximum allowable current threshold. The fast active and fast reactive power support capacity boundary terms can be obtained by converting the maximum allowable current threshold and the DC voltage condition, and the smaller value is taken from the rated capacity upper limit. The recovery slope capacity boundary term can be determined by the allowable rate of change threshold or the allowable rate of change derived from the temperature constraint. This gives each type of capacity boundary term a clear threshold source and calculation rule.

[0031] Furthermore, shrinking or correcting the capacity boundary term based on uncertainty measures includes: Calculate or obtain the uncertainty measure corresponding to the estimated support capacity of each energy storage cluster; The capacity boundary terms are shrunk or corrected based on uncertainty measures to obtain the available supporting capacity boundary of each energy storage cluster. Use the available support capacity boundary as a constraint input for inter-cluster collaborative allocation.

[0032] Specifically, after obtaining the estimated support capacity of each energy storage cluster and its corresponding capacity boundary term, the uncertainty of the capacity boundary term is further corrected to form an available support capacity boundary that can be used for subsequent inter-cluster collaborative allocation. First, calculate or obtain the uncertainty measure corresponding to the estimated support capacity of each energy storage cluster. In specific implementation, for each energy storage cluster... and each type of support capacity component Uncertainty measures are determined separately. These measures characterize the fluctuation or estimation error level of the support capability estimate under the current cluster state input conditions. In one implementation, the uncertainty measure is obtained from the dispersion of multiple inference outputs of the support capability estimation model under the same cluster state input, thus making the uncertainty measure correspond to the current state of the cluster. The uncertainty measure is determined as follows: ; in, Indicates the energy storage cluster The cluster state input is used for the first time. The first reasoning yielded the... Class support capability estimation output, For the number of inferences, For the first The mean of the output of the class support capability estimation is used as the estimated value of the support capability for that class. For the first The standard deviation of the support capability estimate output is used as the corresponding uncertainty measure. It can provide rapid active power support, rapid reactive power support, current limiting, and recovery slope capability. Subsequently, the capacity boundary terms are shrunk or corrected based on uncertainty metrics to obtain the available supporting capacity boundary for each energy storage cluster, i.e., for each energy storage cluster Bound its capability boundary terms to the upper limit With uncertainty measurement By combining these methods, we obtain the upper bound of the available support capacity boundary. To ensure that the available support capacity boundary reflects the uncertainty constraints of the capacity estimation, the boundary terms can be contracted using a confidence boundary contraction method, as shown in the following equation: ; in, For energy storage clusters In the Upper bound of available support capacity in class support capability This is the upper bound of the capability boundary term, determined by current constraints, DC voltage constraints, and temperature constraints. This corresponds to the estimated support capacity. To correspond to the uncertainty measure, This is the boundary contraction coefficient, used to set the contraction range for uncertainty, and can be given by preset parameters; After obtaining the available support capacity boundary of each energy storage cluster, the available support capacity boundary is used as the constraint input for inter-cluster collaborative allocation. First, each energy storage cluster... Upper bound of available support capacity for each of the various support capacity components This is provided to subsequent collaborative allocation steps to limit the share of support that can be allocated to the energy storage cluster to no more than its available support capacity boundary, thereby enabling collaborative allocation to be carried out under boundary constraints; For example, for energy storage cluster 1, under the same cluster state input, The next inference yields the fast active power support capability output sequence, and the mean is calculated. with standard deviation Simultaneously, the upper bound of the boundary terms for the cluster's rapid active power support capability is determined by current constraints, DC voltage constraints, and temperature constraints. In taking At that time, , and Substituting into the boundary contraction formula yields the usable boundary. and will This serves as a constraint input when the cluster participates in inter-cluster collaborative allocation.

[0033] Furthermore, the uncertainty measure is obtained through the Monte Carlo random deactivation method, and the support capability estimation model is a neural network model containing a random deactivation layer. The Monte Carlo random deactivation method includes: While keeping the structure of the support capability estimation model unchanged, the support capability estimation model is made to randomly deactivate some neural network units with a preset deactivation probability during multiple inference processes; Multiple inferences are performed on the same cluster of state variables to obtain multiple sets of support capability estimation outputs; The output dispersion is calculated based on multiple sets of support capacity estimation outputs, and the output dispersion is used as an uncertainty measure.

[0034] Specifically, the uncertainty measure is obtained using the Monte Carlo random deactivation method. The support capability estimation model employs a neural network model with a random deactivation layer. The model takes cluster state variables as input and outputs the support capability estimation result. The random deactivation layer is used to randomly deactivate some neural network units during the inference phase with a preset deactivation probability, thereby enabling the same input to produce different outputs in different inference rounds. While maintaining the structure of the support capability estimation model, some neural network units are randomly deactivated with a preset deactivation probability during multiple inference processes. Specifically, the deactivation probability is pre-set, and the randomly deactivated layers remain active during the inference phase. Each inference iteration generates a different deactivation mask based on the deactivation probability, ensuring different effective network paths in different inference rounds. This results in output fluctuations that reflect the model's uncertainty. Multiple inferences are performed on the same cluster of state variables to obtain multiple sets of support capacity estimates. In specific implementation, for the same energy storage cluster... Cluster state inputs at the same evaluation time Cluster state vector, execution Through this reasoning, we obtain... The group of support capacity estimation outputs, each of which may include fast active power support capacity estimation output, fast reactive power support capacity estimation output, current limiting capacity estimation output, and recovery slope capacity estimation output, are used to characterize the sample set of support capacity estimation results for this cluster at the evaluation time. The output dispersion is calculated based on multiple sets of support capacity estimation outputs, and the output dispersion is used as an uncertainty measure. Specifically, a dispersion index is calculated for each type of support capacity output, and the dispersion index can be one of standard deviation, variance, or range. In the implementation method of using standard deviation as dispersion, the standard deviation of multiple inference outputs is used as the uncertainty measure of the corresponding support capacity estimate, and this uncertainty measure is input into the subsequent capacity boundary term contraction or correction step to form the available support capacity boundary. For example, for energy storage clusters The cluster state input at a certain evaluation time is executed. Through this reasoning, 20 sets of estimated output values ​​for rapid active power support capacity are obtained. The standard deviation of these 20 output values ​​is calculated and used as an uncertainty measure of rapid active power support capacity. Similarly, the uncertainty measures of rapid reactive power support capacity, current limiting capacity, and recovery slope capacity can be obtained respectively, thus forming a set of uncertainty measures for the energy storage cluster to be used for boundary contraction or correction at the current moment.

[0035] S5. Based on the assessment results of network support requirements and available support capabilities, perform inter-cluster collaborative allocation to determine the support share of each energy storage cluster. The support share includes the share of fast active power support, the share of fast reactive power support, the share of current limiting capacity, and the share of recovery slope. Furthermore, S5 includes: The support share of each energy storage cluster is determined under the constraint that the support share of each energy storage cluster does not exceed the corresponding available support capacity boundary. The combined support from each energy storage cluster covers the grid support requirements. When the combined support cannot cover the grid support requirements, the support share of each energy storage cluster is determined according to the preset gap cost rule.

[0036] Specifically, after obtaining the grid support demand parameters and the available support capacity boundary of each energy storage cluster, inter-cluster collaborative allocation is performed to determine the support share of each energy storage cluster. The support share includes the share of fast active power support, the share of fast reactive power support, the share of current limiting capacity, and the share of recovery slope, which are used as inputs for subsequent control parameter generation and grid control execution. Under the constraint that the support share of each energy storage cluster does not exceed the corresponding available support capacity boundary, the support share of each energy storage cluster is determined. In specific implementation, four types of support shares are set for each energy storage cluster, and an upper bound constraint is applied to each type of support share to ensure that it does not exceed the available support capacity boundary of the energy storage cluster in the corresponding support type. The available support capacity boundary is obtained by shrinking or correcting the capacity boundary term based on uncertainty measurement in the previous steps, thereby realizing the uncertainty constraint as the constraint condition of the allocation stage. The composite support of each energy storage cluster's support share is made to cover the grid construction support requirements. That is, under the premise of meeting the upper limit constraints of each energy storage cluster's share, the shares of the same support type are aggregated so that the composite value of the fast active power support share covers the fast active power support requirement parameters, the composite value of the fast reactive power support share covers the fast reactive power support requirement parameters, the composite value of the current limiting capacity share covers the current limiting support requirement parameters, and the composite value of the recovery slope share covers the recovery slope requirement parameters, so as to establish a correspondence between the support share allocation and the grid construction support requirement parameters. When the combined support cannot cover the grid support requirements, the support share of each energy storage cluster is determined according to the preset gap cost rule. When the combined share of a certain support type is insufficient to cover the corresponding requirements due to the limitation of available support capacity boundary, the insufficient part is identified as a gap and the allocation is completed according to the preset gap cost rule. The gap cost rule can adopt a preset priority method, giving priority to allocating the share of fast active power support and fast reactive power support, and then allocating the share of current limiting capacity and recovery slope. Within the same support type, the allocation can be carried out proportionally according to the relative size of the available support capacity boundary of each energy storage cluster, or according to the preset weight, so as to obtain the support share allocation result that meets the boundary constraint conditions. For example, if the system contains cluster 1 and cluster 2, and the demand for fast active power support is greater than the sum of the available fast active power boundaries of cluster 1 and cluster 2, then the difference is determined as a fast active power gap. When the gap cost rule is set to "fast active power priority and allocation of the same type according to the proportion of available boundaries", the fast active power support shares of cluster 1 and cluster 2 are first allocated to the allowable range of their respective available boundaries, and the gap is recorded for subsequent control process processing. At the same time, the allocation of low priority support types is adjusted according to preset rules, thereby forming the support share allocation result for this time. In one embodiment, the gap cost rule processes four types of support requirements in a preset priority order: fast active power support, fast reactive power support, current limiting capacity, and recovery slope. For any support type, allocation is first carried out under the condition that it does not exceed the available support capacity boundary of each energy storage cluster. When an energy storage cluster reaches its available support capacity boundary, the allocation of that type of share to that energy storage cluster is stopped, and the remaining demand is continued to be allocated among the energy storage clusters that have not reached the boundary. When all energy storage clusters have reached the boundary and there is still an unmet portion, the unmet portion is recorded as the support gap of that type, and the allocation process of the next priority support type is carried out according to a preset strategy.

[0037] S6. Map the support share of each energy storage cluster to the control parameters of the corresponding grid-type converter. The control parameters include current limiting trajectory parameters and recovery trajectory parameters. Control the corresponding grid-type converter based on the control parameters. Furthermore, S6 includes: The current-limiting trajectory parameters and recovery trajectory parameters are generated based on the support share of each energy storage cluster; the current-limiting trajectory parameters and recovery trajectory parameters are generated or updated based on the available support capacity boundary, so that the current-limiting trajectory parameters and recovery trajectory parameters meet the constraints of the available support capacity boundary; Current limiting control is implemented on the current reference of the corresponding grid-type converter based on the current limiting trajectory parameters; Recovery control is implemented for the active and reactive power references of the corresponding grid-type converter based on the recovery trajectory parameters.

[0038] Specifically, after obtaining the support share of each energy storage cluster, the support share is converted into the control parameters of the corresponding grid-type converter, and the corresponding independently controllable channels are controlled accordingly, so that the allocation result can be implemented into executable current limiting control and recovery control. For each energy storage cluster, current limiting trajectory parameters and recovery trajectory parameters are generated based on its fast active power support share, fast reactive power support share, current limiting capacity share, and recovery slope share. The current limiting trajectory parameters are used to constrain the change of the current reference as it approaches the current limiting limit, and the recovery trajectory parameters are used to constrain the change of the active power reference and reactive power reference during the recovery process. At the same time, the current limiting trajectory parameters and recovery trajectory parameters are generated or updated based on the available support capacity boundary of the energy storage cluster, so that the current reference limit, power reference change amplitude and change rate do not exceed the range allowed by the available support capacity boundary. Then, the current limit of the current reference is determined based on the current limiting capacity share, and the current reference is limited and updated according to the current limiting trajectory parameters so that the current reference does not exceed the current limiting limit. Then, based on the recovery trajectory parameters, recovery control is implemented on the active and reactive power references of the corresponding grid-type converters: the target values ​​of the active and reactive power references are determined according to the fast active power support share and the fast reactive power support share, and the active and reactive power references are updated according to the recovery trajectory parameters and the recovery slope share, so that the active and reactive power references change according to the preset recovery method. For example, for energy storage cluster 1, after determining its current limiting capacity share and recovery slope share through collaborative allocation, the corresponding current limit value and its current limiting trajectory parameters are generated, and the recovery trajectory parameters of the active and reactive power recovery process are generated; when performing grid control, the current reference is first limited according to the current limiting trajectory parameters, and then the active and reactive power references are restored and updated according to the recovery trajectory parameters.

[0039] S7. Detect the operating events of each energy storage cluster and trigger reconfiguration control, update the available support capacity assessment results and support share, and re-execute steps S5 to S6.

[0040] Furthermore, S7 includes: For each energy storage cluster, detect at least one of the following events: continuous current limiting event, DC undervoltage approach event, temperature rise rate exceeding limit event, energy storage cluster exit event, and communication anomaly event; Reconstruction control is triggered when the detection quantity meets the preset threshold condition; Record the energy storage cluster identifier and trigger cause identifier of the triggering event, and update the uncertainty metric and available support capacity boundary after triggering reconfiguration control.

[0041] Specifically, during execution, the operating status of each energy storage cluster is continuously monitored, and detection quantities are set and updated periodically for different anomaly types. Detection quantities may include the number of consecutive cycles in which the current reaches the limit, the margin between the DC voltage and the undervoltage threshold, the temperature change rate, the exit flag, and the data update timeout / communication status, etc., to reflect whether the current operating status deviates from the executable range. When any detection quantity reaches the corresponding preset threshold condition, the reconstruction process is initiated. This threshold condition can be set separately according to the anomaly type, so that the triggering criteria for entering the reconstruction process have a definite source and a unified judgment method. After entering the reconfiguration process, the identifiers of the energy storage clusters that experienced anomalies and their cause identifiers are recorded for object location in the subsequent update process. Then, the uncertainty measure is recalculated based on the updated cluster state variables, and the shrinkage or correction results of the capacity boundary terms are updated accordingly to obtain the updated available support capacity boundary. Based on the updated results, the subsequent collaborative allocation and control parameter generation process is re-executed to complete the reconfiguration. For example, during a grid construction support process, if the DC voltage margin of a certain energy storage cluster continuously decreases and reaches a preset threshold, it will enter the reconfiguration process. The identifier of the energy storage cluster and the undervoltage cause identifier are recorded. Then, the uncertainty metric and available support capacity boundary corresponding to the energy storage cluster are updated, and the support share and control parameters are recalculated accordingly.

[0042] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent control and power regulation of a string-grid energy storage system, characterized in that, include: S1. Obtain power grid operation status information and system operation mode information, and determine the grid construction support requirements accordingly; S2. Obtain cluster-level topology information of the string-type grid-type energy storage system to determine the correspondence between multiple energy storage clusters and grid-type converters and their independently controllable channels. S3. Collect the cluster state parameters of each energy storage cluster, including state of charge, health status, temperature, DC side voltage and equivalent internal resistance. S4. Evaluate the available support capacity of each energy storage cluster based on the cluster state variables. The evaluation includes determining the estimated support capacity and its corresponding uncertainty measure, and shrinking or correcting the capacity boundary term corresponding to the estimated value based on the uncertainty measure. S5. Based on the network support requirements and the assessment results of available support capabilities, perform inter-cluster collaborative allocation to determine the support share of each energy storage cluster. The support share includes the share of fast active power support, the share of fast reactive power support, the share of current limiting capability, and the share of recovery slope. S6. Map the support share of each energy storage cluster to the control parameters of the corresponding grid-type converter. The control parameters include current-limiting trajectory parameters and recovery trajectory parameters. Control the corresponding grid-type converter based on the control parameters. S7. Detect the operating events of each energy storage cluster and trigger reconfiguration control, update the available support capacity assessment results and the support share, and re-execute steps S5 to S6.

2. The intelligent control and power regulation method for a string-grid energy storage system according to claim 1, characterized in that, S1 includes: Acquire grid frequency information, grid voltage information, and system operating mode identifier; Based on the system operation mode identifier, the target network construction support requirement is selected from the preset network construction support requirement set; The target network support requirements are parameterized to obtain network support requirement parameters, which include fast active power support requirement parameters, fast reactive power support requirement parameters, current limiting support requirement parameters, and recovery slope requirement parameters.

3. The intelligent control and power regulation method for a string-grid energy storage system according to claim 1, characterized in that, S2 includes: Obtain the energy storage cluster identifier, the grid-type converter identifier, and the identifier of the independently controllable channel; Establish a correspondence table between energy storage clusters, independently controllable channels, and grid-type converters; The control object of each energy storage cluster is determined based on the correspondence table.

4. The intelligent control and power regulation method for a string-grid energy storage system according to claim 1, characterized in that, S3 includes: Collect the state of charge, health status, temperature, DC side voltage, and equivalent internal resistance of each energy storage cluster; Perform time alignment processing on the collected data and remove outliers; Generate a set of cluster state variables for assessing available support capabilities.

5. The intelligent control and power regulation method for a string-grid energy storage system according to claim 1, characterized in that, S4 includes: Based on the cluster state variables, the estimated support capacity of each energy storage cluster is calculated. The estimated support capacity includes the estimated fast active power support capacity, the estimated fast reactive power support capacity, the estimated current limiting capacity, and the estimated recovery slope capacity. Based on current constraints, DC voltage constraints, and temperature constraints, determine the capability boundary terms corresponding to the estimated support capability; Output the estimated support capacity of each energy storage cluster and the corresponding capacity boundary terms.

6. The intelligent control and power regulation method for a string-grid energy storage system according to claim 5, characterized in that, The shrinking or correction of the capability boundary term based on uncertainty measurement includes: Calculate or obtain the uncertainty measure corresponding to the estimated support capacity of each energy storage cluster; The capability boundary term is shrunk or corrected based on the uncertainty metric to obtain the available support capability boundary of each energy storage cluster. The available support capability boundary is used as the constraint input for the inter-cluster collaborative allocation.

7. The intelligent control and power regulation method for a string-grid energy storage system according to claim 1, characterized in that, S5 includes: The support share of each energy storage cluster is determined under the constraint that the support share of each energy storage cluster does not exceed the corresponding available support capacity boundary. The combined support of each energy storage cluster covers the network support requirements. When the synthetic support cannot cover the network support requirements, the support share of each energy storage cluster is determined according to the preset gap cost rule.

8. The intelligent control and power regulation method for a string-grid energy storage system according to claim 1, characterized in that, S6 includes: Current-limiting trajectory parameters and recovery trajectory parameters are generated based on the support share of each energy storage cluster; wherein, the current-limiting trajectory parameters and recovery trajectory parameters are generated or updated based on the available support capacity boundary, so that the current-limiting trajectory parameters and recovery trajectory parameters satisfy the constraints of the available support capacity boundary; Current limiting control is implemented on the current reference of the corresponding grid-type converter based on the current limiting trajectory parameters; Based on the recovery trajectory parameters, recovery control is implemented for the active and reactive power references of the corresponding grid-type converter.

9. The intelligent control and power regulation method for a string-grid energy storage system according to claim 1, characterized in that, S7 includes: For each energy storage cluster, detect at least one of the following events: continuous current limiting event, DC undervoltage approach event, temperature rise rate exceeding limit event, energy storage cluster exit event, and communication anomaly event; The reconstruction control is triggered when the detected quantity meets the preset threshold condition. Record the energy storage cluster identifier and trigger reason identifier of the triggering event, and update the uncertainty metric and the available support capacity boundary after triggering the reconfiguration control.

10. The intelligent control and power regulation method for a string-grid energy storage system according to claim 6, characterized in that, The uncertainty measure is obtained through the Monte Carlo random deactivation method, and the support capability estimation model is a neural network model containing a random deactivation layer. The Monte Carlo random deactivation method includes: While keeping the structure of the support capability estimation model unchanged, the support capability estimation model is made to randomly deactivate some neural network units with a preset deactivation probability during multiple inference processes; Multiple inferences are performed on the same cluster of state variables to obtain multiple sets of support capability estimation outputs; The output dispersion is calculated based on the multiple sets of support capability estimation outputs, and the output dispersion is used as the uncertainty measure.

Images