Distributed compressed air energy storage system cluster cooperative control method and system

Abstract

The invention belongs to the technical field of energy storage, and particularly discloses a distributed compressed air energy storage system cluster cooperative control method and system, and the method comprises the steps: a cloud end determines a real-time operation scene; when the real-time operation scene is in a normal state, the cloud end receives the spatial-temporal characteristic matrix of each energy storage node uploaded by the terminal and the load prediction curve uploaded by the edge end; based on the spatial-temporal characteristic matrix and the load prediction curve, aggregating the energy storage nodes by using the constructed global federated learning model to obtain the weight of each energy storage node, and generating a global control scheme according to the weight of each energy storage node; and issuing the global control scheme to the edge end and the terminal. And in an emergency state, the cloud carries out fault positioning and issues an emergency instruction to the edge end and the terminal according to a fault positioning result. According to the invention, efficient collaboration among the energy storage nodes is realized, and the overall performance and the operation efficiency of the system are improved.

Description

Technical Field

[0001] This invention belongs to the field of energy storage technology, specifically relating to a cluster collaborative control method and system for a distributed compressed air energy storage system. Background Technology

[0002] With the rapid development of renewable energy and the construction of smart grids, distributed compressed air energy storage systems (CASS) have emerged as a highly efficient energy storage technology. They offer advantages such as high energy density, low cost, and environmental friendliness, effectively addressing the intermittency and instability of renewable energy sources and improving the flexibility and stability of the power grid. This has made CASS an important demand-side regulation resource in the new power system. However, current distributed CASS systems still have some shortcomings in the coordinated control of their operational strategies.

[0003] Traditional distributed compressed air energy storage (CASS) systems lack the ability to analyze and process real-time data. The complex gas-electricity-thermal coupling relationships among multiple distributed CASS nodes make traditional models ill-suited to the complex grid operating environment and fluctuating load demands. Furthermore, the coordinated control among multiple energy storage devices is not precise or efficient enough, resulting in the system's overall performance not being fully realized. When facing large-scale renewable energy integration and rapid grid load changes, existing technologies exhibit lag and inconsistencies in energy storage device charging and discharging control, power regulation, and energy management, impacting grid stability and efficient energy utilization. With the widespread application of distributed CASS systems, the current centralized control method relying on a central server for scheduling suffers from high communication latency and a high risk of data privacy breaches. Summary of the Invention

[0004] The purpose of this invention is to provide a distributed compressed air energy storage system cluster collaborative control method and system, which realizes efficient collaboration between energy storage nodes and improves the overall performance and operating efficiency of the system.

[0005] To achieve the above objectives, the present invention employs the following technical solution: According to one aspect of the present invention, a method for coordinated control of a distributed compressed air energy storage system cluster is provided, comprising the following steps: Determine the real-time operating scenario; When the real-time operating scenario is in a normal state, the receiving terminal uploads the spatiotemporal feature matrix of each energy storage node and the load prediction curve uploaded by the edge terminal; where normal state is the state of stable grid operation; Based on the spatiotemporal feature matrix and load forecast curve, the energy storage nodes are aggregated using the constructed global federated learning model to obtain the weight of each energy storage node, and a global control scheme is generated based on the weight of each energy storage node. The global control scheme is distributed to the edge and terminal.

[0006] By adopting the above technical solution and utilizing the Dynamic Weighted Federated Learning (DWFL) algorithm, the real-time physical parameters (pressure, temperature, health status, etc.) of each energy storage system device are integrated into the model aggregation weight calculation, which effectively improves the practicality and reliability of cluster collaboration.

[0007] According to one embodiment of the present invention, the real-time operation scenario further includes an emergency state; the emergency state is the state when a power grid failure occurs; when the real-time operation scenario is in an emergency state, fault location is performed, and emergency instructions are issued to the edge end and the terminal based on the fault location result.

[0008] Furthermore, when the real-time operating scenario is in an emergency, the steps of fault location and issuing emergency commands to the edge and terminal based on the fault location results include: Determine the fault type and calculate the theoretical power change of each power grid node based on the fault type; Obtain the actual power change and rated power of each power grid node; The fault location accuracy of each power grid node is calculated based on the theoretical power change, actual power change, and rated power of each power grid node. Fault location is performed based on the fault location accuracy of each power grid node to obtain the fault location result; Emergency commands are issued to the edge devices and terminals based on the fault location results.

[0009] According to one embodiment of the present invention, the step of calculating the theoretical power change of each power grid node based on the fault type includes: Input the fault type and the location information of each power grid node into the constructed fault analysis mathematical model; Based on the power grid parameters under normal conditions, the power flow equations are solved to obtain the power distribution of the faulty power grid nodes before the fault. Based on the fault type and location information, modify the boundary conditions in the network equations and re-solve the power distribution after the faulty power grid node fails. The theoretical power change is calculated based on the power distribution before and after the fault in the power grid node.

[0010] Information on each power grid node includes the location of each power grid node.

[0011] According to one embodiment of the present invention, the step of determining the real-time running scene includes: Obtain the frequency change, voltage change, and actual power change of each power grid node; The comprehensive fault assessment criterion value is calculated based on the frequency change, voltage change, and actual power change of each power grid node; The comprehensive fault assessment criterion value is compared with the preset value; if the comprehensive fault assessment criterion value is not greater than the preset value, it is a normal state; if the comprehensive fault assessment criterion value is greater than the preset value, it is an emergency state.

[0012] According to another aspect of the present invention, a method for coordinated control of a distributed compressed air energy storage system cluster is provided, comprising the following steps: The system receives a global control scheme from the cloud and updates the local model parameters according to the global control scheme. The global control scheme is generated as follows: the cloud aggregates the energy storage nodes based on the spatiotemporal feature matrix and load prediction curve of each energy storage node uploaded by the terminal, obtains the weight of each energy storage node, and generates a global control scheme based on the weight of each energy storage node. Obtain real-time status information of each energy storage node in the terminal; Based on the real-time status information of each energy storage node in the terminal, the load forecast curve is obtained through the updated local model and then uploaded to the cloud. Based on the overall control scheme and the real-time status information of each energy storage node in the terminal, control commands are issued to the terminal.

[0013] According to one embodiment of the present invention, the local model is an LSTM time series prediction model; The real-time status information of each energy storage node includes load electricity price, grid electricity price, energy storage system status parameters, and ambient temperature; the energy storage system status parameters include compressed gas pressure and compressed gas temperature. The training methods for the local model include: Obtain a data training set; the data training set includes the historical state information of each energy storage node of the terminal and the actual future load value corresponding to the historical state information of each energy storage node; Using the local model to be trained, the predicted load value for future moments is obtained based on the historical state information of each energy storage node; The objective function is established to minimize the mean squared error between the predicted future load value and the actual future load value; the objective function is solved, the local model parameters are adjusted, and the trained local model is obtained.

[0014] According to one embodiment of the present invention, a distributed compressed air energy storage system cluster collaborative control method further includes the following steps: Receive emergency instructions from the cloud; Emergency control instructions are issued to the terminal according to the emergency instructions.

[0015] According to another aspect of the present invention, a method for cluster collaborative control of a distributed compressed air energy storage system is provided, comprising the following steps: The spatiotemporal feature matrix and real-time status information of each energy storage node are collected; wherein, the spatiotemporal feature matrix includes: gas storage tank pressure, temperature, compressor remaining life, current charge and discharge efficiency, photovoltaic output and load demand of each energy storage node; the real-time status information includes photovoltaic output, load demand, gas storage tank pressure, grid frequency deviation and equipment health status of each energy storage node; The spatiotemporal feature matrix is ​​uploaded to the cloud, and the real-time status information of the energy storage nodes is uploaded to the edge. Receive and execute global control schemes issued from the cloud; Receive and execute control commands issued from the edge device.

[0016] According to one embodiment of the present invention, a distributed compressed air energy storage system cluster collaborative control method further includes the following steps: The real-time operating scenario is determined based on the spatiotemporal feature matrix and real-time status information. The real-time operating scenario includes normal state and emergency state. When in an emergency state, wherein the emergency state is the state when a power grid failure occurs; Energy storage nodes closer to the fault point release electrical energy first to provide power support to the grid and send fault signals to the edge.

[0017] According to a fourth aspect of the present invention, a distributed compressed air energy storage system cluster collaborative control system is provided, comprising: In the cloud, the system determines the real-time operating scenario. When the real-time operating scenario is in a normal state, it receives the spatiotemporal feature matrix of each energy storage node uploaded by the terminal and the load prediction curve uploaded by the edge terminal. The normal state refers to the stable operation of the power grid. Based on the spatiotemporal feature matrix and the load prediction curve, the system aggregates the energy storage nodes using a pre-constructed global federated learning model to obtain the weight of each energy storage node. A global control scheme is then generated based on the weight of each energy storage node. The global control scheme is then distributed to the edge terminal and the terminal. At the edge, the system receives the global control scheme from the cloud and updates the local model parameters according to the global control scheme. The global control scheme is generated by the cloud based on the spatiotemporal feature matrix and load prediction curve of each energy storage node uploaded by the terminal, and aggregates the energy storage nodes using a pre-constructed global federated learning model to obtain the weight of each energy storage node. Obtain real-time status information of each energy storage node in the terminal; based on the real-time status information of each energy storage node in the terminal, obtain the load forecast curve through the updated local model, and upload the load forecast curve to the cloud; based on the global control scheme and the real-time status information of each energy storage node in the terminal, issue control commands to the terminal. The terminal is used to collect the spatiotemporal feature matrix and real-time status information of each energy storage node. The spatiotemporal feature matrix includes: gas storage tank pressure, temperature, compressor remaining life, current charge / discharge efficiency, photovoltaic output, and load demand of each energy storage node. The real-time status information includes photovoltaic output, load demand, gas storage tank pressure, grid frequency deviation, and equipment health status of each energy storage node. The terminal uploads the spatiotemporal feature matrix to the cloud and uploads the real-time status information of the energy storage nodes to the edge. It receives and executes the global control scheme issued by the cloud and receives and executes the control commands issued by the edge.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects: 1. This invention combines federated learning with deep learning, which enables efficient collaboration between energy storage nodes while ensuring data security, thereby improving the overall performance and operating efficiency of the system. It has significant innovation and practicality.

[0019] 2. This invention utilizes the distributed architecture of federated learning to make the system easy to expand. Newly added energy storage nodes can be easily added to the cooperative control network without requiring large-scale modifications to the entire system.

[0020] 3. This invention optimizes the traditional "cloud-edge-device" collaborative control mechanism and proposes a multi-objective dynamic priority collaborative control mechanism with "security-efficiency-economy" corresponding to the "cloud-edge-device" layer. This solves the problem of "fixed objectives and insufficient flexibility" in the traditional layered control and realizes the synergy of "safe operation-efficiency optimization-economic benefits" of distributed energy storage clusters. Attached Figure Description

[0021] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of the distributed compressed air energy storage system cluster collaborative control method according to Embodiment 1 of the present invention. Detailed Implementation

[0022] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0023] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0024] Example 1 A method for coordinated control of a distributed compressed air energy storage system cluster includes the following steps: Determine the real-time operating scenario; When the real-time operating scenario is in a normal state, the receiving terminal uploads the spatiotemporal feature matrix of each energy storage node and the load prediction curve uploaded by the edge terminal; where normal state is the state of stable grid operation; Based on the spatiotemporal feature matrix and load forecast curve, the energy storage nodes are aggregated using the constructed global federated learning model to obtain the weight of each energy storage node, and a global control scheme is generated based on the weight of each energy storage node. The global control scheme is distributed to the edge and terminal.

[0025] It should be noted that an energy storage node refers to a node in a distributed compressed air energy storage system that is connected to an energy storage device. In this embodiment, the energy storage device is a compressed air energy storage device.

[0026] By adopting the above technical solution, the cloud, as the core decision-making layer of the distributed compressed air energy storage system, aggregates the various energy storage nodes at the terminal by constructing a global federated learning model. Based on the aggregation results, a global control scheme is generated to optimize the power allocation of the distributed compressed air energy storage system, achieving a high proportion of distributed clean energy consumption on the user side and helping to ensure the safe and stable operation of the power grid. The cloud utilizes the spatiotemporal feature matrix of the terminal and the load forecast curve obtained from the edge to aggregate the various energy storage nodes, enabling the analysis and processing of the real-time physical state of the system and the operating conditions of the power grid. This improves the lag and blindness in the grid's charging and discharging control, power regulation, and energy management processes, helping to improve the accuracy and efficiency of coordinated control among multiple energy storage nodes and fully leveraging the overall performance of the system. Furthermore, the distributed architecture of the global federated learning model allows newly added energy storage units to be easily integrated into the coordinated control network without requiring large-scale modifications to the entire system.

[0027] Specifically, the distributed compressed air energy storage system cluster collaborative control method includes the following steps: S1. Determine the real-time operating scenario.

[0028] The frequency change, voltage change, and actual power change of each grid node are obtained from WAMS data. The frequency change, voltage change, and actual power change of each grid node are information obtained from the grid. The voltage (including voltage amplitude and voltage phase, etc.), current (including current amplitude and current phase, etc.), frequency, and power of each grid node in the grid are monitored in real time.

[0029] It should be noted that a grid node refers to the connection point of an electrical component or branch that reflects the actual operating conditions and topology of the power grid, including but not limited to nodes of distributed compressed air energy storage systems, such as hub substations.

[0030] The comprehensive fault assessment criterion value is calculated based on the frequency change, voltage change, and actual power change at each power grid node. The comprehensive fault assessment criterion value is calculated according to the following formula: ; Where n is the number of grid nodes in the distributed compressed air energy storage system; Δf i ΔV represents the frequency change at grid node i. i ΔP represents the voltage change at node i in the power grid. i λ represents the actual power change at grid node i; f0 represents the frequency reference value at grid node i; V0 represents the voltage reference value at grid node i; P0 represents the power reference value at grid node i; λ i μ i and ν i These are the weighting coefficients.

[0031] Comprehensive evaluation criteria value Compared with preset value The comparison is performed; if the comprehensive evaluation criterion value is not greater than the preset value, it is a normal state, which is the state of stable operation of the power grid; if the comprehensive evaluation criterion value is greater than the preset value, it is an emergency state, which is the state when a power grid failure occurs.

[0032] S2. When the real-time operating scenario is in a normal state, receive the spatiotemporal feature matrix of each energy storage node uploaded by the receiving terminal and the load prediction curve uploaded by the edge terminal.

[0033] The spatiotemporal feature matrix includes real-time physical parameters such as gas storage tank pressure, temperature, compressor remaining life, current charge / discharge efficiency, photovoltaic output, and load demand for each energy storage node.

[0034] The cloud communicates with multiple edge devices to collect load forecast curves obtained by the edge devices using local models.

[0035] S3. Based on the spatiotemporal feature matrix and load prediction curve, the energy storage nodes are aggregated using the constructed global federated learning model to obtain the weights of each energy storage node, and a global control scheme is generated based on the weights of each energy storage node.

[0036] The spatiotemporal feature matrix and load prediction curve are input into the constructed global federated learning model. The aggregation strategy of the global federated learning model is dynamically adjusted to calculate the weight value of each energy storage node. Let the total number of energy storage nodes in the distributed compressed air energy storage system be... m Among them, energy storage nodesj The weight value is calculated according to the following formula: ; in, t The current moment; For the current energy storage node j The weight value; This is the maximum pressure of the gas storage tank; For optimal operating temperature; Design life for the compressor; , , and This represents the scene optimization coefficient.

[0037] cloud-based Adjusting the weight values ​​of each energy storage node allows for the priority adoption of energy storage node models with better physical conditions, i.e., energy storage nodes with higher weight values, when controlling the charging and discharging and regulating the power of energy storage nodes in the energy storage system. Specifically, this means that the pressure is moderate, the temperature is close to the optimal operating temperature, the equipment is healthier, and the charging and discharging efficiency is higher, thus avoiding the decline in cluster performance due to the failure or inefficient operation of a single energy storage node.

[0038] During the cluster-based coordinated discharge process of a distributed compressed air energy storage system, the cloud calculates the gas storage capacity and power generation of each energy storage node based on real-time data obtained from WAMS. Weighted control of the power generation of each energy storage node is achieved according to its weight value, ensuring that the actual operating conditions of each node are fully considered to maximize efficiency.

[0039] Therefore, the discharge power of the distributed compressed air energy storage system is: ; in, m The total number of energy storage nodes, For energy storage nodes j The discharge power, This refers to the discharge power of a distributed compressed air energy storage system.

[0040] S4. Distribute the global control scheme to the edge devices and terminals.

[0041] The cloud will distribute the generated global control scheme to the edge devices and terminals.

[0042] In addition, real-time operational scenarios also include emergency states; when When an emergency is detected, emergency support is triggered. At the same time, the severity of the fault is classified into levels: Level 1 requires immediate support, and Level 2 requires preparation for support.

[0043] The severity of faults can be graded by comprehensively considering parameters such as operating characteristics, equipment carrying capacity, and load characteristics under different power grid structures. This grading can generally be obtained from simulation results or determined empirically. Typically, the calculated comprehensive fault assessment criterion value... Deviation from preset value The more numerous the faults, the more severe the malfunctions, and the greater the need for emergency handling of the energy storage system.

[0044] Specifically, when a real-time operational scenario is in an emergency, fault location is performed, and emergency commands are issued to the edge devices and terminals based on the fault location results. This includes the following steps: S1-1. Determine the fault type and calculate the theoretical power change of each power grid node based on the fault type.

[0045] The fault types include three-phase short circuit, two-phase short circuit, two-phase ground fault, single-phase ground fault, single-phase open circuit, and two-phase open circuit.

[0046] Theoretical power change The theoretical power change value is calculated by establishing a fault analysis mathematical model and combining parameters such as fault type and fault location, which can be used to assist in fault location.

[0047] The fault analysis mathematical model includes the power grid topology, component parameters, fault types, and fault locations. Specifically, the fault analysis mathematical model is based on Kirchhoff's laws and power flow calculation formulas, and can be constructed using simulation software, combined with the actual power grid topology and equipment parameters (such as line impedance and load characteristics). Before use, the fault analysis mathematical model can be trained using historical fault data (including historical data on fault types, fault locations, and corresponding actual power changes at each power grid node).

[0048] Once a fault is detected, the fault type, grid node location information, etc., are input into the fault analysis mathematical model, and the following steps are used to obtain the result. : 1) Steady-state calculation before the fault: Based on the grid parameters under normal conditions, solve the power flow equations to obtain the voltage, current, and power distribution of the faulty grid node before the fault (denoted as P). i,before Power grid parameters include power parameters (active power, reactive power, and power factor), voltage parameters (voltage amplitude and phase), and impedance parameters (line impedance and transformer impedance).

[0049] 2) Post-fault transient calculation: Based on the fault type and location information, modify the boundary conditions in the network equations, and resolve for the voltage, current, and power distribution (denoted as P) of the faulted power grid node after the fault. i,after ).

[0050] 3) Calculate the theoretical power change based on the power distribution before and after the fault in the power grid node: =P i,after -P i,before。

[0051] S1-2. Obtain the actual power change at each power grid node. and rated power In this embodiment, the actual power change is... It refers to the actual power change of a power grid node measured within a time window of milliseconds between the time of the fault and the time of the fault. It is used to judge and clarify the power surge caused by the fault and is an evaluation value for fault detection.

[0052] S1-3. Based on the theoretical power change of each power grid node. Actual power change and rated power Calculate the fault location accuracy of each power grid node. .

[0053] Fault location accuracy Calculate using the following formula: ; in, This is for adjusting the coefficient.

[0054] The above formula is used to calculate the fault location accuracy. It can integrate information from the entire network, thereby enabling the measurement of the collaborative error of multiple power grid nodes and avoiding misjudgment of the overall positioning effect caused by the local error of a single power grid node.

[0055] S1-4. Fault location is performed based on the fault location accuracy of each power grid node.

[0056] Through fault location accuracy Compare measured power changes Compared with the calculated value based on the fault model The differences in these parameters help assess the accuracy of fault location. The smaller the value, the greater the change in measured power. Compared with the calculated value based on the fault model The smaller the difference, the more accurate the fault location.

[0057] S1-5. Issue emergency commands to the edge and terminal based on the fault location results.

[0058] The cloud platform prioritizes controlling the energy storage nodes (i.e., distributed compressed air energy storage devices) closest to the fault point in the terminal to rapidly release energy. These nodes provide short-circuit current support and temporary emergency power supply. Controlling the rapid energy release of these nodes can be achieved either by the cloud issuing emergency commands to the terminal for direct regulation, or by the cloud indirectly regulating the nodes by issuing emergency commands to the edge.

[0059] Simultaneously, the cloud will also adjust the output power of other energy storage nodes to maintain the voltage and power balance of the power grid. The adjustment of the output power of other energy storage nodes can be achieved by the cloud issuing emergency commands to the terminal to directly control the energy storage nodes; or by the cloud issuing emergency commands to the edge terminal to indirectly control the energy storage nodes.

[0060] Therefore, the distributed compressed air energy storage system cluster collaborative control method adopted in this embodiment can achieve efficient dynamic collaborative optimization of the energy storage node cluster, improving the overall performance and operating efficiency of the system. The distributed architecture of the global federated learning model makes the system easy to expand; the physical state-aware dynamic weighted federated learning algorithm incorporates the real-time physical parameters (pressure, temperature, health, etc.) of each energy storage system device into the model for aggregated weight calculation, replacing the traditional weighting method based solely on model parameters or data volume, effectively improving the practicality and reliability of cluster collaboration.

[0061] Example 2 A method for coordinated control of a distributed compressed air energy storage system cluster includes the following steps: The edge device receives the global control scheme from the cloud and updates the local model parameters according to the global control scheme. The global control scheme is generated as follows: the cloud aggregates the energy storage nodes based on the spatiotemporal feature matrix and load prediction curve of each energy storage node uploaded by the terminal, obtains the weight of each energy storage node, and generates the global control scheme according to the weight of each energy storage node. Obtain real-time status information of each energy storage node in the terminal; Based on the real-time status information of each energy storage node in the terminal, the load forecast curve is obtained through the updated local model and then uploaded to the cloud. Based on the overall control scheme and the real-time status information of each energy storage node in the terminal, control commands are issued to the terminal.

[0062] The local model is an LSTM time series prediction model. The real-time status information of each energy storage node includes load electricity price, grid electricity price, energy storage system status parameters, and ambient temperature; the energy storage system status parameters include compressed gas pressure and compressed gas temperature.

[0063] Training methods for local models include: Acquire the data training set; the data training set includes the historical state information of each energy storage node of the terminal and the actual future load value corresponding to the historical state information of each energy storage node; Using the local model to be trained, the predicted load value for future moments is obtained based on the historical state information of each energy storage node; The objective function is established to minimize the mean squared error between the predicted and actual future load values, and is used to optimize the deviation between the model prediction and the actual value. The objective function is as follows: MIN[ ]; Where M is the sample size. The predicted load value for future time periods. This represents the actual load value at a future time. Solve the objective function, adjust the local model parameters, and obtain the trained local model.

[0064] By minimizing the mean square error, the LSTM time series forecasting model can better capture the correlation between time series data of parameters such as load electricity price, grid electricity price, ambient temperature and energy storage system status and charging and discharging power, thereby improving the forecast accuracy.

[0065] Furthermore, a collaborative optimization model is built at the edge, as follows: ; Where N is the number of energy storage nodes in the energy storage system; These are the charging power and discharging power of energy storage node j, respectively; This represents the net electricity cost incurred by energy storage node j due to charging and discharging operations.

[0066] The constraints are as follows: ; ; ; in, This is the maximum charging power of the energy storage system. These are the minimum and maximum gas storage capacities of energy storage node j, respectively. These are the overall charging efficiency and discharging efficiency of the energy storage system, respectively. λ These are Lagrange multipliers used to balance system-level constraints; For time intervals.

[0067] Under the above constraints, the collaborative optimization model is solved, and the predicted future load values ​​obtained by the LSTM time series prediction model are further optimized. This allows the edge end to optimize and constrain the prediction results from an economic perspective during the process of obtaining the load prediction curve, which helps the cloud generate the most economically optimal global control scheme.

[0068] In addition, when the cloud determines that the real-time operating scenario is in an emergency, the edge device also receives emergency instructions from the cloud and sends emergency control instructions to the terminal according to the emergency instructions.

[0069] When the edge device receives a global control scheme issued by the cloud under normal conditions or an emergency command issued under emergency conditions, it will dynamically adjust the charging and discharging power of each energy storage node in the region based on the actual status of the terminal, while reserving 10% of redundant capacity to cope with sudden loads, thereby realizing the control of multiple energy storage nodes in the region.

[0070] Example 3 A method for coordinated control of a distributed compressed air energy storage system cluster includes the following steps: The system collects the spatiotemporal characteristic matrix and real-time status information of each energy storage node. The spatiotemporal characteristic matrix for each node includes information such as gas storage tank pressure, temperature, remaining compressor lifespan, state of health (SOH), current charge / discharge efficiency, grid frequency deviation, photovoltaic output, and load demand. The real-time status information for each node includes load electricity price, grid electricity price, energy storage system status parameters, and ambient temperature. The energy storage system status parameters include compressed gas pressure and compressed gas temperature.

[0071] The spatiotemporal feature matrix is ​​uploaded to the cloud, and the real-time status information of the energy storage nodes is uploaded to the edge. Receive and execute global control schemes issued from the cloud; Receive and execute control commands issued from the edge device.

[0072] In addition, when the cloud determines that the real-time operating scenario is in an emergency, the terminal can also receive and execute emergency instructions issued by the cloud; it can also receive and execute emergency control instructions issued by the edge terminal.

[0073] Furthermore, when the cloud determines that the real-time operating scenario is in an emergency, the terminal can respond autonomously and quickly. The energy storage nodes closest to the fault point can also respond quickly, releasing electrical energy to ensure power supply and sending fault signals to the edge.

[0074] Each energy storage node in the terminal is equipped with a local fault detection and emergency control module. This module can monitor local grid parameters such as voltage, current, and frequency in real time. When a grid fault is detected, such as a sudden voltage drop, a rapid increase in current, or frequency fluctuations, it can respond autonomously and quickly without waiting for instructions from the cloud or edge, immediately switching to a preset emergency power supply mode to provide emergency support to the grid, effectively shortening response time and improving support efficiency in the event of a grid fault.

[0075] At the same time, when the terminal detects that the pressure in the gas storage tank exceeds the threshold ( > When a grid fault occurs, the system triggers a local emergency shutdown and sends a fault signal to the edge to ensure the safety of the energy storage system itself. During grid faults, the distributed compressed air energy storage system adopts a phased power output strategy based on the severity of the fault and the grid's needs, with each system cooperating to rationally allocate power output tasks.

[0076] Example 4 A distributed compressed air energy storage system cluster collaborative control method includes: The cloud determines the real-time operating scenario, which includes normal and emergency states. When the real-time operation scenario is in a normal state, the cloud receives the spatiotemporal feature matrix of each energy storage node uploaded by the terminal and the load prediction curve uploaded by the edge terminal; where the normal state is the state of stable grid operation; based on the spatiotemporal feature matrix and the load prediction curve, the energy storage nodes are aggregated using the constructed global federated learning model to obtain the weight of each energy storage node, and a global control scheme is generated according to the weight of each energy storage node. The global control scheme is distributed to the edge and terminal.

[0077] The edge device receives the global control scheme from the cloud and updates the local model parameters according to the global control scheme; The edge device acquires the real-time status information of each energy storage node of the terminal; based on the real-time status information of each energy storage node of the terminal, it obtains the load prediction curve through the updated local model and uploads the load prediction curve to the cloud. Based on the overall control scheme and the real-time status information of each energy storage node in the terminal, control commands are issued to the terminal.

[0078] The terminal receives and executes the global control scheme issued by the cloud; it also receives and executes the control instructions issued by the edge terminal.

[0079] When a real-time operational scenario is in an emergency, the cloud performs fault location and issues emergency commands to the edge and terminal based on the fault location results. Specifically, this includes: Determine the fault type and calculate the theoretical power change of each energy storage node based on the fault type; Obtain the actual power change and rated power of each energy storage node; The fault location accuracy of each energy storage node is calculated based on the theoretical power change, actual power change, and rated power of each energy storage node. Fault location is determined based on the fault location accuracy of each energy storage node. Emergency commands are issued to the edge devices and terminals based on the fault location results.

[0080] The steps for determining the real-time running scenario in the cloud include: Obtain the frequency change, voltage change, and actual power change of each energy storage node; The comprehensive fault assessment criterion value is calculated based on the frequency change, voltage change, and actual power change of each energy storage node; The comprehensive fault assessment criterion value is compared with the preset value; if the comprehensive fault assessment criterion value is not greater than the preset value, it is a normal state; if the comprehensive fault assessment criterion value is greater than the preset value, it is an emergency state.

[0081] The edge device receives emergency instructions from the cloud and sends emergency control instructions to the terminal based on the emergency instructions.

[0082] When the edge device receives a global control scheme issued by the cloud under normal conditions or an emergency command issued under emergency conditions, it will dynamically adjust the charging and discharging power of each energy storage node in the region based on the actual status of the terminal, while reserving 10% of redundant capacity to cope with sudden loads, thereby realizing the control of multiple energy storage nodes in the region.

[0083] The terminal can receive and execute emergency commands sent from the cloud; it can also receive and execute emergency control commands sent from the edge.

[0084] Furthermore, each energy storage node in the terminal is equipped with a local fault detection and emergency control module. This module can monitor local grid parameters such as voltage, current, and frequency in real time. When a grid fault is detected, such as a sudden voltage drop, a rapid increase in current, or frequency fluctuations, it can autonomously and quickly respond without waiting for instructions from the cloud or edge. It immediately switches to a preset emergency power supply mode, with the energy storage node closest to the fault releasing power and sending a fault signal to the edge. This provides emergency support to the grid, effectively shortening response time and improving support efficiency in the event of grid faults.

[0085] At the same time, when the terminal detects that the pressure in the gas storage tank exceeds the threshold ( > When a grid fault occurs, the system triggers a local emergency shutdown and sends a fault signal to the edge to ensure the safety of the energy storage system itself. During grid faults, the distributed compressed air energy storage system adopts a phased power output strategy based on the severity of the fault and the grid's needs, with each system cooperating to rationally allocate power output tasks.

[0086] Example 5 A distributed compressed air energy storage system cluster collaborative control device includes: In the cloud, the system is used to determine the real-time operating scenario. When the real-time operating scenario is in a normal state, it receives the spatiotemporal feature matrix of each energy storage node uploaded by the terminal and the load prediction curve uploaded by the edge terminal. Based on the spatiotemporal feature matrix and the load prediction curve, it aggregates the energy storage nodes using the constructed global federated learning model to obtain the weight of each energy storage node. It then generates a global control scheme based on the weight of each energy storage node and distributes the global control scheme to the edge terminal and the terminal. At the edge, the system receives the global control scheme from the cloud and updates the local model parameters according to the global control scheme. The global control scheme is generated by the cloud based on the spatiotemporal feature matrix and load prediction curve of each energy storage node uploaded by the terminal, and aggregates the energy storage nodes using a pre-constructed global federated learning model to obtain the weight of each energy storage node. Obtain real-time status information of each energy storage node in the terminal; based on the real-time status information of each energy storage node in the terminal, obtain the load forecast curve through the updated local model, and upload the load forecast curve to the cloud; based on the global control scheme and the real-time status information of each energy storage node in the terminal, issue control commands to the terminal. The terminal is used to collect the spatiotemporal feature matrix and real-time status information of each energy storage node. The spatiotemporal feature matrix includes: gas storage tank pressure, temperature, compressor remaining life, current charge / discharge efficiency, photovoltaic output, and load demand of each energy storage node. The real-time status information includes photovoltaic output, load demand, gas storage tank pressure, grid frequency deviation, and equipment health status of each energy storage node. The terminal uploads the spatiotemporal feature matrix to the cloud and uploads the real-time status information of the energy storage nodes to the edge. It receives and executes the global control scheme issued by the cloud and receives and executes the control commands issued by the edge.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A cluster collaborative control method for a distributed compressed air energy storage system, characterized in that, Includes the following steps: Determine the real-time operating scenario; When the real-time operating scenario is in a normal state, the receiving terminal uploads the spatiotemporal feature matrix of each energy storage node and the load prediction curve uploaded by the edge terminal; wherein, the normal state is the state of stable operation of the power grid; Based on the spatiotemporal feature matrix and load forecast curve, the energy storage nodes are aggregated using the constructed global federated learning model to obtain the weight of each energy storage node, and a global control scheme is generated based on the weight of each energy storage node. The global control scheme is distributed to the edge and terminal.

2. The distributed compressed air energy storage system cluster collaborative control method according to claim 1, characterized in that, The real-time operation scenario also includes an emergency state; wherein, the emergency state is the state when a power grid failure occurs; Determine the fault type and calculate the theoretical power change of each power grid node based on the fault type; Obtain the actual power change and rated power of each power grid node; The fault location accuracy of each power grid node is calculated based on the theoretical power change, actual power change, and rated power of each power grid node. Fault location is performed based on the fault location accuracy of each power grid node to obtain the fault location result; Emergency commands are issued to the edge devices and terminals based on the fault location results.

3. The distributed compressed air energy storage system cluster collaborative control method according to claim 2, characterized in that, The step of calculating the theoretical power change of each power grid node based on the fault type includes: Input the fault type and the location information of each power grid node into the constructed fault analysis mathematical model; Based on the grid parameters under normal conditions, the power flow equations are solved to obtain the power distribution before the grid fault. Based on the fault type and location information, modify the boundary conditions in the network equations and re-solve the power distribution after the faulty power grid node fails. The theoretical power change is calculated based on the power distribution before and after the fault in the power grid node.

4. The distributed compressed air energy storage system cluster collaborative control method according to claim 1, characterized in that, The step of determining the real-time operating scenario includes: Obtain the frequency change, voltage change, and actual power change of each power grid node; The comprehensive fault assessment criterion value is calculated based on the frequency change, voltage change, and actual power change of each power grid node; The comprehensive fault assessment criterion value is compared with the preset value; if the comprehensive assessment criterion value is not greater than the preset value, it is a normal state; if the comprehensive assessment criterion value is greater than the preset value, it is an emergency state.

5. A cluster collaborative control method for a distributed compressed air energy storage system, characterized in that, Includes the following steps: The system receives a global control scheme from the cloud and updates the local model parameters according to the global control scheme. The global control scheme is generated as follows: the cloud aggregates the energy storage nodes based on the spatiotemporal feature matrix and load prediction curve of each energy storage node uploaded by the terminal, obtains the weight of each energy storage node, and generates a global control scheme based on the weight of each energy storage node. Obtain real-time status information of each energy storage node in the terminal; Based on the real-time status information of each energy storage node in the terminal, the load forecast curve is obtained through the updated local model and then uploaded to the cloud. Based on the overall control scheme and the real-time status information of each energy storage node in the terminal, control commands are issued to the terminal.

6. The distributed compressed air energy storage system cluster collaborative control method according to claim 5, characterized in that, The real-time status information of each energy storage node includes load electricity price, grid electricity price, energy storage system status parameters, and ambient temperature; the energy storage system status parameters include compressed gas pressure and compressed gas temperature. The training methods for the local model include: Obtain a data training set; the data training set includes the historical state information of each energy storage node of the terminal and the actual future load value corresponding to the historical state information of each energy storage node; Using the local model to be trained, the predicted load value for future moments is obtained based on the historical state information of each energy storage node; The objective function is established to minimize the mean squared error between the predicted future load value and the actual future load value; the objective function is solved, the local model parameters are adjusted, and the trained local model is obtained.

7. The distributed compressed air energy storage system cluster collaborative control method according to claim 5, characterized in that, It also includes the following steps: Receive emergency instructions from the cloud; Emergency control instructions are issued to the terminal according to the emergency instructions.

8. A cluster collaborative control method for a distributed compressed air energy storage system, characterized in that, Includes the following steps: The spatiotemporal feature matrix and real-time status information of each energy storage node are collected; wherein, the spatiotemporal feature matrix includes: gas storage tank pressure, temperature, compressor remaining life, current charge and discharge efficiency, photovoltaic output and load demand of each energy storage node; the real-time status information includes load electricity price, grid electricity price, energy storage system status parameters and ambient temperature; The spatiotemporal feature matrix is ​​uploaded to the cloud, and the real-time status information of the energy storage nodes is uploaded to the edge. Receive and execute global control schemes issued from the cloud; Receive and execute control commands issued from the edge device.

9. The distributed compressed air energy storage system cluster collaborative control method according to claim 8, characterized in that, It also includes the following steps: The real-time operating scenario is determined based on the spatiotemporal feature matrix and real-time status information. The real-time operating scenario includes normal state and emergency state. When in a state of emergency The energy storage nodes closest to the fault point release the stored electrical energy first and send a fault signal to the edge.

10. A distributed compressed air energy storage system cluster collaborative control system, characterized in that, include: In the cloud, the system is used to determine the real-time operating scenario. When the real-time operating scenario is in a normal state, it receives the spatiotemporal feature matrix of each energy storage node uploaded by the receiving terminal and the load prediction curve uploaded by the edge terminal. The normal state is the state of stable operation of the power grid. Based on the spatiotemporal feature matrix and the load prediction curve, the system uses the constructed global federated learning model to aggregate the energy storage nodes, obtain the weight of each energy storage node, and generate a global control scheme based on the weight of each energy storage node. The global control scheme is distributed to the edge devices and terminals; At the edge, the system receives the global control scheme from the cloud and updates the local model parameters according to the global control scheme. The global control scheme is generated by the cloud based on the spatiotemporal feature matrix and load prediction curve of each energy storage node uploaded by the terminal, and aggregates the energy storage nodes using a pre-constructed global federated learning model to obtain the weight of each energy storage node. Obtain real-time status information of each energy storage node in the terminal; based on the real-time status information of each energy storage node in the terminal, obtain the load forecast curve through the updated local model, and upload the load forecast curve to the cloud; based on the global control scheme and the real-time status information of each energy storage node in the terminal, issue control commands to the terminal. The terminal is used to collect the spatiotemporal feature matrix and real-time status information of each energy storage node. The spatiotemporal feature matrix includes: gas tank pressure, temperature, compressor remaining life, current charge and discharge efficiency, photovoltaic output, and load demand of each energy storage node. The real-time status information includes load electricity price, grid electricity price, energy storage system status parameters, and ambient temperature. The terminal uploads the spatiotemporal feature matrix to the cloud and uploads the real-time status information of the energy storage nodes to the edge. It also receives and executes the global control scheme issued by the cloud. Receive and execute control commands issued from the edge device.

Images