Control method of an energy storage system and energy storage system

Abstract

This invention discloses a control method and an energy storage system for an energy storage system, belonging to the field of energy storage technology. The control method includes: receiving an automatic power generation control frequency regulation command from the power grid; acquiring first operating state information of the power grid, second operating state information of the flywheel energy storage module, third operating state information of the electrochemical energy storage module, environmental information, and electricity market operating information; based on the acquired information, performing multi-timescale state prediction to obtain future operating state information of the energy storage system at different time scales; and based on the first, second, and third operating state information, environmental information, electricity market operating information, and future operating state information, obtaining power allocation commands for the flywheel energy storage module and the electrochemical energy storage module, and controlling the two types of energy storage modules to perform power regulation. The method of this invention enhances the rationality of power allocation between the flywheel energy storage module and the electrochemical energy storage module during grid-connected operation.

Description

Technical Field

[0001] This invention relates to the field of energy storage technology, and in particular to a control method and an energy storage system. Background Technology

[0002] With a high proportion of intermittent renewable energy sources such as wind and solar power continuously being integrated into the grid, the grid's operating status exhibits significant random fluctuations, with frequent changes in frequency and power. Simultaneously, in applications such as industrial and commercial peak-valley arbitrage and microgrid operation, energy storage systems need to frequently switch between various operating conditions to meet multiple functional requirements, including grid frequency regulation, load balancing, and energy management. Hybrid energy storage systems, composed of flywheel energy storage and electrochemical energy storage, are gradually becoming an important technical means to improve grid flexibility due to their complementary advantages in response speed and energy support capabilities.

[0003] However, existing hybrid energy storage systems rely on direct acquisition of current operating status information for power control decisions during actual operation, resulting in a lack of foresight in control decisions. Under complex grid fluctuations and multi-source information coupling operating conditions, the energy storage system struggles to coordinate the power distribution between flywheel energy storage modules and electrochemical energy storage modules in a timely manner, thus affecting the overall coordination and stability of the system and failing to meet the demand for high-quality power regulation under complex operating conditions. Summary of the Invention

[0004] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, one objective of this invention is to propose a control method for an energy storage system. This method, based on a multi-timescale state prediction mechanism, improves the energy storage system's ability to perceive future changes in its operating state. Furthermore, based on the collaborative analysis of real-time operating state information and future operating state information, it improves the rationality of power distribution between the flywheel energy storage module and the electrochemical energy storage module, thereby enhancing the coordinated adjustment and response capabilities of the energy storage system during grid-connected operation.

[0005] The second objective of this invention is to propose an energy storage system.

[0006] To achieve the above objectives, a control method for an energy storage system according to a first aspect of the present invention is provided. The energy storage system includes a flywheel energy storage module and an electrochemical energy storage module, wherein the flywheel energy storage module and the electrochemical energy storage module operate in parallel with the power grid. The control method includes: receiving an automatic power generation control frequency regulation command from the power grid; acquiring first operating state information of the power grid, second operating state information of the flywheel energy storage module, third operating state information of the electrochemical energy storage module, environmental information, and electricity market operating information; performing multi-time-scale state prediction based on the first operating state information, the second operating state information, the third operating state information, the environmental information, and the electricity market operating information to obtain future operating state information of the energy storage system at different time scales; obtaining a power allocation command between the flywheel energy storage module and the electrochemical energy storage module based on the first operating state information, the second operating state information, the third operating state information, the environmental information, the electricity market operating information, and the future operating state information; and controlling the flywheel energy storage module and the electrochemical energy storage module to perform power regulation based on the power allocation command.

[0007] According to the control method of the energy storage system of the present invention, the method first acquires grid operation status information, flywheel energy storage module operation status information, electrochemical energy storage module operation status information, environmental information, and electricity market operation information. This allows the controlled object to simultaneously cover grid-side disturbance sources and dynamic changes within the energy storage system, thus providing a multi-dimensional state basis for subsequent decision-making. Based on this, the method further constructs future operation status information of the energy storage system at different time scales based on the aforementioned multi-source state information. This enables the system to form a unified representation of short-term fluctuations and relatively long-term change trends beyond the current operation status, thereby ensuring that control decisions are no longer limited to static information at a single moment but simultaneously cover the current operation status and future evolution trends. Then, the current and future operation status information are used together as input to generate power allocation commands for the flywheel and electrochemical energy storage modules. This ensures that the power allocation process simultaneously considers immediate operational needs and future state change trends, thus avoiding local optima or delayed decision-making problems that may result from relying solely on the current state. Ultimately, by adjusting the power of the flywheel energy storage module and the electrochemical energy storage module based on this power allocation command, the two types of energy storage modules can more rationally coordinate their power sharing relationship under the automatic power generation control frequency regulation command of the power grid. This allows for a better match between the fast response characteristics of the flywheel and the energy support characteristics of the electrochemical energy storage, thereby enhancing the coordination and rationality of power allocation between the flywheel energy storage module and the electrochemical energy storage module. Ultimately, this improves the overall collaborative regulation and response capabilities of the energy storage system during grid-connected operation.

[0008] In some embodiments, the multi-timescale state prediction includes: predicting grid frequency and power fluctuations at a first time scale based on the first operating state information; predicting load changes and renewable energy output at a second time scale based on the first operating state information, the environmental information, and the electricity market operation information; and predicting lifetime degradation trends at a third time scale based on the second operating state information and the third operating state information; wherein the first time scale is smaller than the second time scale, and the second time scale is smaller than the third time scale.

[0009] In some embodiments, obtaining power allocation instructions between the flywheel energy storage module and the electrochemical energy storage module based on the first operating state information, the second operating state information, the third operating state information, the environmental information, the electricity market operating information, and the future operating state information includes: inputting the first operating state information, the second operating state information, the third operating state information, the environmental information, the electricity market operating information, and the future operating state information into a near-end strategy optimization model; obtaining output data from the near-end strategy optimization model; and obtaining power allocation instructions between the flywheel energy storage module and the electrochemical energy storage module based on the output data.

[0010] In some embodiments, obtaining a power allocation command between the flywheel energy storage module and the electrochemical energy storage module based on the output data includes: correcting the output data based on constraints; when the corrected output data satisfies the constraints, the corrected output data is used as a power allocation command between the flywheel energy storage module and the electrochemical energy storage module; wherein the constraints include the rotational speed constraint of the flywheel energy storage module, the charge / discharge rate constraint of the electrochemical energy storage module, and the state of charge constraint of the electrochemical energy storage module.

[0011] In some embodiments, the power allocation command includes: allocating high-frequency power components to the flywheel energy storage module for power regulation, and allocating mid- and low-frequency power components to the electrochemical energy storage module for power regulation.

[0012] In some embodiments, controlling the flywheel energy storage module and the electrochemical energy storage module to perform power regulation based on the power allocation command includes: synchronously generating gate drive signals corresponding to the converter of the flywheel energy storage module, the converter of the electrochemical energy storage module, and the grid-connected inverter, respectively, based on the power allocation command; and synchronously controlling the converter of the flywheel energy storage module, the converter of the electrochemical energy storage module, and the grid-connected inverter based on the gate drive signals to achieve power conversion.

[0013] In some embodiments, the control method further includes: comparing and analyzing the first operating status information, the second operating status information, the third operating status information, the environmental information, and the electricity market operating information with historical operating status information; and making anomaly predictions for the grid operating status, the flywheel energy storage module operating status, and the electrochemical energy storage module operating status based on the comparison and analysis results.

[0014] In some embodiments, the control method further includes: acquiring the power tracking error corresponding to the power allocation command, the energy storage module operating status change information, and the energy storage system abnormal information, as execution result feedback information; and updating the multi-timescale state prediction, the near-end strategy optimization model, and the synchronization control strategy between each energy storage module and the power grid online based on the execution result feedback information.

[0015] In some embodiments, the control method further includes: performing data preprocessing on the first operating status information, the second operating status information, the third operating status information, the environmental information, and the electricity market operating information.

[0016] To achieve the above objectives, a second aspect of the present invention provides an energy storage system comprising: a grid and load module, an electrochemical energy storage module, and a flywheel energy storage module; a grid-connected converter module for bidirectional conversion between AC and DC power; a DC bus module for connecting the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected converter module, and serving as a common energy exchange node among the modules; and a control module connected to the grid and load module, the electrochemical energy storage module, the flywheel energy storage module, the grid-connected converter module, and the DC bus module, for implementing the control method of the energy storage system described in the above embodiment.

[0017] According to an embodiment of the energy storage system of the present invention, firstly, by setting up a grid and load module, an electrochemical energy storage module, and a flywheel energy storage module, the system possesses the basic conditions for unified perception and response to power fluctuations on the grid side and the internal energy state of the energy storage modules. Further, by setting up a grid-connected converter module, bidirectional conversion between AC and DC power is achieved, thereby providing an energy conversion channel for energy exchange and grid-connected regulation between the flywheel energy storage module and the electrochemical energy storage module. Simultaneously, by setting up a DC bus, the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected converter module are electrically connected, allowing each energy storage module and grid-connected device to share the same energy exchange node, thereby improving the basic consistency of power transfer and coordinated regulation among the modules. Through the coordinated connection of the control module with the grid and load module, each energy storage module, the grid-connected converter module, and the DC bus, the control module can obtain the operating status of each module and perform unified control of each execution unit, thus providing the hardware execution and coordination basis for the implementation of the control method of the above-mentioned energy storage system. Then, by adopting the control method of the energy storage system described in the above embodiment, the energy storage system's ability to perceive future changes in operating status is improved by utilizing a multi-timescale state prediction mechanism. Based on the collaborative analysis of real-time operating status information and future operating status information, the rationality of power distribution between the flywheel energy storage module and the electrochemical energy storage module is improved, thereby enhancing the energy storage system's collaborative adjustment and response capabilities during grid-connected operation.

[0018] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of a control method for an energy storage system according to an embodiment of the present invention; Figure 2 This is an overall flowchart of a control method for an energy storage system according to an embodiment of the present invention; Figure 3 This is a schematic diagram of an energy storage system according to an embodiment of the present invention.

[0020] Figure label: Energy storage system 100; Grid and load module 1; Electrochemical energy storage module 2; Flywheel energy storage module 3; Grid-connected converter module 4; DC bus module 5; Control module 6; 11 AC power grid; 12 User load / new energy power station; 13 Voltage transformer; 14 Current transformer; 21 Battery cluster; 22 Battery management system; 23 Bidirectional LLC resonant converter; 24 Electrochemical energy storage control module; 25 Battery status sensor; 31 Flywheel body; 32 Motor / generator; 33 Bidirectional three-level Buck-Boost converter; 34 Flywheel control module; 35 Flywheel status sensor; 41 LCL filter circuit; 42 Three-level NPC grid-connected inverter; 43 Inverter control module; 51 DC bus; 52 DC bus support capacitor bank; 61 AI main control chip; 62 Multi-source data acquisition module; 63 Multi-timescale state prediction module; 64 Multi-objective reinforcement learning optimization module; 65 Global synchronization control and protection module. Detailed Implementation

[0021] The embodiments of the present invention are described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. The embodiments of the present invention are described in detail below.

[0022] The following is for reference. Figure 1 A control method for an energy storage system according to an embodiment of the present invention is described.

[0023] In some embodiments, the energy storage system includes a flywheel energy storage module and an electrochemical energy storage module, which are connected to the power grid.

[0024] In some embodiments, the flywheel energy storage module can be an energy storage module that stores and releases energy based on mechanical kinetic energy. It achieves bidirectional conversion between electrical energy and kinetic energy through the energy conversion between a motor / generator and a high-speed rotating flywheel. During charging, electrical energy drives the flywheel to rotate at high speed, converting electrical energy into rotational kinetic energy for storage. During discharging, the motor generates electricity in reverse, converting mechanical kinetic energy into electrical energy for output to the grid or load side. Since the flywheel energy storage module mainly relies on mechanical inertia characteristics in the energy conversion process, it has advantages such as fast response speed (up to millisecond-level response), long cycle life (up to millions of cycles), and high power density, making it suitable for scenarios with rapid grid frequency fluctuations. However, its energy density is relatively low, making it difficult to support long-term continuous power supply. Therefore, it is more suitable for scenarios involving high-frequency power fluctuation smoothing and rapid power support.

[0025] In some embodiments, the electrochemical energy storage module can be an energy storage module that realizes the storage and release of electrical energy based on an electrochemical reaction process. It achieves the reversible conversion between electrical energy and chemical energy through the redox reaction between the internal electrode material and the electrolyte. Its working principle is that during the charging process, electrical energy is converted into chemical energy and stored through an electrochemical reaction, and during the discharging process, electrical energy is released to the DC or AC side through a reverse electrochemical reaction. Electrochemical energy storage modules have advantages such as high energy density and suitability for medium- and long-term energy support, but they are prone to accelerated cycle life decay under high-frequency charge and discharge conditions. Therefore, they are more suitable for medium- and low-frequency energy regulation and continuous power support scenarios.

[0026] In some embodiments, the electrochemical energy storage module may include, but is not limited to, different types of electrochemical energy storage modules such as lithium-ion battery energy storage modules, sodium-ion battery energy storage modules, or lead-carbon battery energy storage modules.

[0027] In some embodiments, the flywheel energy storage module and the electrochemical energy storage module operate in parallel with the grid. This means that the two types of energy storage modules interact with the grid through a grid-connected inverter, enabling the energy storage system to achieve bidirectional charging and discharging regulation according to the grid's automatic power generation control frequency regulation commands. This allows them to participate in the grid's frequency support, power balance, and ancillary service functions. Through grid-connected operation, the flywheel energy storage module can handle high-frequency, rapid power fluctuation regulation of the grid, while the electrochemical energy storage module can handle medium- and low-frequency energy compensation and continuous power support. The two complement each other through coordinated operation, thus balancing rapid dynamic response and long-term energy support capabilities, improving the overall adaptability and regulation capability of the energy storage system to grid fluctuations.

[0028] Figure 1 This is a flowchart of a control method for an energy storage system according to an embodiment of the present invention, such as... Figure 1 As shown, the control method of the energy storage system in this embodiment of the invention includes at least the following steps: S1 receives the automatic power generation control frequency regulation command from the power grid.

[0029] In some embodiments, the Automatic Generation Control (AGC) frequency regulation command for the power grid can be a regulation command generated in real time by the power grid dispatch center based on the power grid frequency deviation and power balance requirements. This command instructs the grid-connected energy storage system to participate in the dynamic regulation of the power grid's active power to maintain grid frequency stability and power balance. When load fluctuations or fluctuations in renewable energy output cause the frequency to deviate from its rated value, the AGC frequency regulation command is used to adjust and allocate the charging and discharging power of various grid-connected resources (including flywheel energy storage modules and electrochemical energy storage modules) in real time, thereby achieving rapid recovery of the power grid frequency and suppression of power deviation, improving the reliability of power grid operation.

[0030] In some embodiments, the AGC frequency regulation command of the power grid can be received by the control module in the energy storage system, and the AGC frequency regulation command can be used as one of the input bases for power control decisions, so as to achieve unified and coordinated control of the flywheel energy storage module and the electrochemical energy storage module.

[0031] S2 acquires the first operating status information of the power grid, the second operating status information of the flywheel energy storage module, the third operating status information of the electrochemical energy storage module, environmental information, and electricity market operation information.

[0032] In some embodiments, the first operating state information of the power grid may include, but is not limited to, voltage information, current information, frequency information, phase information, etc. on the power grid side. This information is used to characterize the current power quality state, power balance state, and frequency stability state of the power grid, thereby reflecting the overall operating condition of the power grid at the current moment.

[0033] In some embodiments, the first operating state information of the power grid can be obtained by the voltage transformer (PT) and current transformer (CT) on the grid side, and then output to the control module of the energy storage system for subsequent control decisions.

[0034] In some embodiments, the second operating status information of the flywheel energy storage module may include, but is not limited to: flywheel speed information, vibration parameters, winding temperature information, bearing temperature information, current information, etc. This information is used to characterize the mechanical operating status, electrical operating status and power response capability of the flywheel energy storage module, thereby reflecting the current energy storage level and dynamic adjustment capability of the flywheel energy storage module.

[0035] In some embodiments, the second operating status information of the flywheel energy storage module can be acquired by a flywheel status sensor installed in the flywheel energy storage module, and the acquired second operating status information can be output to the control module of the energy storage system.

[0036] In some embodiments, the third operating state information of the electrochemical energy storage module may include, but is not limited to, the voltage information, current information, temperature information, internal resistance information, state of charge (SOC) information, and state of health (SOH) information of the battery cells. This information is used to characterize the charge and discharge state, energy storage level, and degradation degree of the electrochemical energy storage module, thereby reflecting the operating performance and lifespan of the electrochemical energy storage module.

[0037] In some embodiments, the third operating state information of the electrochemical energy storage module can be collected and processed by the battery state sensor and the battery management system (BMS) installed in the electrochemical energy storage module, and then output to the control module of the energy storage system for subsequent control decisions.

[0038] In some embodiments, environmental information may include, but is not limited to, external operating environment parameters such as ambient temperature and humidity; electricity market operation information may include, but is not limited to, time-of-use electricity price information, electricity market dispatch plan information, and electricity market transaction-related operation data, which are used to characterize external economic dispatch constraints and operation strategy constraints, thereby affecting the operation decisions of the energy storage system.

[0039] In some embodiments, environmental information and electricity market operation information can be provided by external information sources and transmitted to the control module through a communication interface.

[0040] In some embodiments, the first operating status information, the second operating status information, the third operating status information, environmental information, and electricity market operating information are all uniformly acquired and aggregated by the multi-source data acquisition module in the control module of the energy storage system to form the basic input data required for subsequent control decisions, thereby ensuring the consistency and integrity of various types of information in the time dimension.

[0041] S3, based on the first operating status information, the second operating status information, the third operating status information, environmental information, and electricity market operating information, performs multi-time-scale state prediction to obtain the future operating status information of the energy storage system at different time scales.

[0042] In some embodiments, multi-timescale state prediction can refer to hierarchical prediction of the future operating state of an energy storage system within different time ranges based on first operating state information, second operating state information, third operating state information, environmental information, and electricity market operating information. This hierarchical prediction characterizes the changing trends of the power grid operating state, load changing trends, and the evolutionary trends of the energy storage module operating state, thereby forming a multi-level description of the future operating state of the energy storage system. The purpose of multi-timescale state prediction is to enable the energy storage system not only to perceive its current operating state but also to obtain future operating trend information at different time scales in advance, thus providing a more comprehensive state basis for subsequent power allocation decisions.

[0043] In some embodiments, the necessity of multi-timescale state prediction lies in the fact that state predictions at different time scales are used to characterize the evolution characteristics of grid operation disturbances and system operating state changes within different time ranges, and to match these with the operating characteristics of flywheel energy storage modules and electrochemical energy storage modules, thereby achieving hierarchical coordinated control of power allocation. Since flywheel energy storage modules and electrochemical energy storage modules differ in response speed, power regulation capability, and energy support capability—with flywheel energy storage modules suitable for rapidly changing power regulation scenarios and electrochemical energy storage modules suitable for relatively stable energy regulation scenarios—obtaining future operating state information at different time scales allows the power allocation process to simultaneously consider the impact of changes at different time scales on the energy storage system. This improves the matching and coordination of power allocation between flywheel and electrochemical energy storage modules, and enhances the rationality and adaptability of the overall power regulation of the energy storage system.

[0044] S4 obtains power allocation instructions between the flywheel energy storage module and the electrochemical energy storage module based on the first operating status information, the second operating status information, the third operating status information, environmental information, electricity market operating information, and future operating status information.

[0045] In some embodiments, the power allocation command of the flywheel energy storage module and the electrochemical energy storage module is used to determine the power ratio or power output target that each of the two types of energy storage modules should undertake in the process of participating in grid frequency regulation and power balance under the constraints of current operating state and future operating state information, so as to reasonably allocate the power regulation demand of the overall energy storage system between the flywheel energy storage module and the electrochemical energy storage module.

[0046] In some embodiments, through power distribution commands, the flywheel energy storage module and the electrochemical energy storage module can each undertake corresponding power regulation tasks according to their respective dynamic response characteristics and energy support capabilities. The flywheel energy storage module is used to undertake the regulation needs of rapidly changing power components, while the electrochemical energy storage module is used to undertake the needs of relatively gradual power changes and energy compensation. This enables the energy storage system as a whole to achieve a more coordinated power sharing relationship during grid-connected operation.

[0047] S5, based on power distribution commands, controls the flywheel energy storage module and the electrochemical energy storage module to perform power regulation.

[0048] In some embodiments, the flywheel energy storage module and the electrochemical energy storage module are controlled to regulate power based on power distribution commands. This allows the flywheel energy storage module and the electrochemical energy storage module to output or absorb corresponding power according to the power distribution results, thereby achieving rapid response and accurate tracking of the grid's automatic power generation control frequency regulation commands.

[0049] In some embodiments, by coordinating the power regulation of the flywheel energy storage module and the electrochemical energy storage module based on power distribution commands, the flywheel energy storage module can fully utilize its rapid response characteristics to cope with instantaneous power fluctuations, while the electrochemical energy storage module provides continuous energy support, thereby improving the stability and coordination of the overall power regulation of the energy storage system.

[0050] According to the control method of the energy storage system of the present invention, the method first acquires grid operation status information, flywheel energy storage module operation status information, electrochemical energy storage module operation status information, environmental information, and electricity market operation information. This allows the controlled object to simultaneously cover grid-side disturbance sources and dynamic changes within the energy storage system, thus providing a multi-dimensional state basis for subsequent decision-making. Based on this, the method further constructs future operation status information of the energy storage system at different time scales based on the aforementioned multi-source state information. This enables the system to form a unified representation of short-term fluctuations and relatively long-term change trends beyond the current operation status, thereby ensuring that control decisions are no longer limited to static information at a single moment but simultaneously cover the current operation status and future evolution trends. Then, the current and future operation status information are used together as input to generate power allocation commands for the flywheel and electrochemical energy storage modules. This ensures that the power allocation process simultaneously considers immediate operational needs and future state change trends, thus avoiding local optima or delayed decision-making problems that may result from relying solely on the current state. Ultimately, by adjusting the power of the flywheel energy storage module and the electrochemical energy storage module based on this power allocation command, the two types of energy storage modules can more rationally coordinate their power sharing relationship under the automatic power generation control frequency regulation command of the power grid. This allows for a better match between the fast response characteristics of the flywheel and the energy support characteristics of the electrochemical energy storage, thereby enhancing the coordination and rationality of power allocation between the flywheel energy storage module and the electrochemical energy storage module. Ultimately, this improves the overall collaborative regulation and response capabilities of the energy storage system during grid-connected operation.

[0051] In some embodiments, multi-timescale state prediction includes: predicting grid frequency and power fluctuations at a first time scale based on first operating state information; predicting load changes and renewable energy output at a second time scale based on the first operating state information, environmental information, and electricity market operation information; and predicting lifetime degradation trends at a third time scale based on the second and third operating state information. The first time scale is shorter than the second time scale, and the second time scale is shorter than the third time scale.

[0052] In some embodiments, the first time scale can be a millisecond-level time scale, used to characterize the high-frequency dynamic changes of the power grid in a short period of time, such as transient characteristics such as power grid frequency disturbances and rapid power fluctuations, thereby reflecting the instantaneous operating state change trend of the power grid under rapid disturbance conditions.

[0053] In some embodiments, the prediction of grid frequency and power fluctuations at a first time scale based on the first operating state information is because the first operating state information can directly reflect the instantaneous operating characteristics of the grid side. By performing high-frequency time scale analysis on this type of information, the rapid changing trends of grid frequency and power can be captured more accurately, thereby providing a basis for power regulation of the energy storage system under rapid response conditions.

[0054] In some embodiments, the purpose of predicting grid frequency and power fluctuations is to obtain the dynamic change trend of the grid in a short time scale in advance, so that fast-response energy storage modules such as flywheel energy storage modules can participate in power support and frequency regulation in a timely manner, avoiding the response lag problem caused by relying solely on the current state for control decisions, thereby improving the energy storage system's response and adaptability to transient disturbances in the grid.

[0055] In some embodiments, power grid frequency and power fluctuation prediction can be achieved using a prediction model combining TCN and Transformer. TCN (Temporal Convolutional Network) is used to extract local temporal features of the first operating state information, while Transformer (attention mechanism network) is used to model global temporal dependencies, thereby predicting the trend of power grid frequency and power changes within a short future time window. For example, preferably, power grid frequency and power fluctuations within the next 100 ms can be predicted, and the prediction results are output with a time resolution of 1 ms, thus obtaining the high-frequency dynamic change trend of the power grid at the first time scale. However, it should be noted that the present invention is not limited to the specific values ​​mentioned above and is not subject to any limitations.

[0056] In some embodiments, the second time scale can be a minute-level time scale, used to characterize the load changes and dynamic changes in the output of new energy sources during the short-to-medium-term operation of the power grid, thereby reflecting the power balance trend changes of the power grid over a relatively long time window.

[0057] In some embodiments, load changes and renewable energy output forecasts at a second time scale are made based on first operating status information, environmental information, and electricity market operating information. This is because the first operating status information reflects the current operating status of the power grid, the environmental information reflects changes in external conditions affecting renewable energy output, and the electricity market operating information reflects changes in electricity demand-side prices and dispatch constraints. By integrating and analyzing the above multi-source information, comprehensive forecasts can be made of load changes and renewable energy output changes in the power grid within a short to medium time range, thereby improving the ability to make forward-looking judgments on changes in power grid energy supply and demand.

[0058] In some embodiments, load change and new energy output forecasting can refer to forecasting changes in the power load demand of the power grid and changes in the power output of new energy sources such as wind power and photovoltaic power in the short to medium time range. The purpose is to obtain the changing trend of the power supply and demand structure of the power grid in advance, thereby providing a basis for energy scheduling and power allocation of the energy storage system in the short to medium time scale and improving the system's adaptability to changes in power balance.

[0059] In some embodiments, the load change and renewable energy output prediction at the second time scale can be implemented using an attention-based LSTM model. Specifically, multi-dimensional time series input data is constructed based on first operating state information, environmental information, and electricity market operating information. An LSTM model is used to model the time dependency between load and renewable energy output, and an attention mechanism is introduced to enhance key time features, thereby enabling the prediction of renewable energy output and load changes within the next few tens of minutes. For example, preferably, the renewable energy output and load changes within the next 60 minutes can be predicted, and the prediction results are output with a 1-minute time resolution to obtain the short-to-medium-term operating trend of the power grid at the second time scale. It should be noted that the present invention is not limited to the specific values ​​mentioned above and is not restricted thereto.

[0060] In some embodiments, the third time scale can be a long-period time scale, used to characterize the state evolution characteristics of the flywheel energy storage module and the electrochemical energy storage module over a long operating period, thereby reflecting the health status change trend of the flywheel energy storage module and the electrochemical energy storage module during long-term operation, and providing a basis for long-term operation evaluation and power allocation strategy optimization of the energy storage system.

[0061] In some embodiments, lifetime degradation trend prediction based on a third time scale is performed using second and third operating state information because the second operating state information reflects the real-time operating state of the flywheel energy storage module, such as changes in rotational speed, temperature, and power, while the third operating state information reflects the real-time operating state of the electrochemical energy storage module, such as state of charge, internal resistance, and temperature. By performing long-term correlation analysis on the aforementioned operating state information, trend predictions of performance changes in the energy storage module during long-term operation can be made, thereby improving the forward-looking understanding of the lifetime evolution patterns of the energy storage module.

[0062] In some embodiments, the lifespan degradation trend can refer to the mechanical lifespan change trend of the flywheel energy storage module and the state of health (SOH) degradation trend of the electrochemical energy storage module. The purpose is to obtain the performance degradation change pattern of the energy storage module in advance during long-term operation, thereby providing a basis for adjusting the power distribution strategy of the energy storage system during long-term operation, so as to improve the long-term stability and reliability of the overall system operation.

[0063] In some embodiments, the prediction of lifetime degradation trends at the third time scale can be implemented using a GNN (Graph Neural Network) model. Specifically, a state association graph structure of energy storage modules is constructed based on the second and third operating state information. The GNN model is used to model the state association relationships and long-term evolution patterns between different energy storage modules, thereby enabling the prediction of the lifetime degradation trends of flywheel energy storage modules and electrochemical energy storage modules over the next few dozen days. For example, preferably, the lifetime degradation trends of flywheel energy storage modules and electrochemical energy storage modules can be predicted over the next 30 days, and the prediction results are output with a time resolution of 1 day to obtain the long-term state change trend of the energy storage system at the third time scale. It should be noted that the present invention is not limited to the specific values ​​mentioned above and is not restricted herein.

[0064] In general, existing technologies always statically allocate power demand to flywheel energy storage modules and electrochemical energy storage modules according to a pre-set fixed power frequency division rule, without fully considering the characteristics of grid power fluctuations and the dynamic changes in the operating status of each energy storage module. This can easily lead to a mismatch between power allocation and actual response capability, resulting in some energy storage modules bearing a non-optimal power component.

[0065] This invention, based on the different physical response characteristics of flywheel energy storage modules and electrochemical energy storage modules, adapts a millisecond-level high-frequency prediction mechanism to the fast response characteristics of flywheel energy storage modules, enabling flywheel energy storage modules to better handle high-frequency power regulation requirements. Simultaneously, it adapts a minute-level short-to-medium-term prediction mechanism to the long-term energy support characteristics of electrochemical energy storage modules, enabling electrochemical energy storage modules to better handle medium- and low-frequency energy regulation requirements. This achieves adaptive power component allocation among different energy storage modules. It avoids the coarse matching problem caused by the fixed frequency division method in existing technologies, thereby improving the precision of power allocation and the overall coordination of system operation.

[0066] In some embodiments, based on first operating state information, second operating state information, third operating state information, environmental information, electricity market operating information, and future operating state information, power allocation instructions for the flywheel energy storage module and the electrochemical energy storage module are obtained, including: inputting the first operating state information, second operating state information, third operating state information, environmental information, electricity market operating information, and future operating state information into a near-end strategy optimization model, obtaining output data of the near-end strategy optimization model, and obtaining power allocation instructions for the flywheel energy storage module and the electrochemical energy storage module based on the output data.

[0067] In some embodiments, the Proximal Policy Optimization (PPO) model can be a reinforcement learning-based policy optimization model used to output a power allocation strategy in a continuous action space based on the current system state. Under the premise of satisfying the system operation constraints, the power allocation strategy of the flywheel energy storage module and the electrochemical energy storage module gradually approaches the optimal control strategy, thereby optimizing the operating performance of the energy storage system.

[0068] In some embodiments, the near-end strategy optimization model may include a strategy network and a value network, wherein the strategy network corresponds to the Actor network and is used to output the power allocation action between the flywheel energy storage module and the electrochemical energy storage module based on the state vector, and the value network corresponds to the Critic network and is used to evaluate the value of the strategy execution effect under the current state, thereby providing an evaluation basis for strategy updates.

[0069] In some embodiments, after inputting various operating status information into the near-end strategy optimization model, the control module first fuses the grid operating status information, flywheel energy storage module operating status information, electrochemical energy storage module operating status information, environmental information, and electricity market operating information, and combines this with future operating status information predicted at multiple time scales to form a state vector characterizing the current operating status and future evolution trend of the energy storage system. Then, the Actor network outputs power allocation coefficients between the flywheel energy storage module and the electrochemical energy storage module based on the state vectors. These power allocation coefficients represent the target power allocation relationship between the flywheel energy storage module and the electrochemical energy storage module. The Critic network evaluates the value of the current strategy based on the state vectors, thereby assisting in strategy optimization and gradually bringing the power allocation strategy closer to the optimal control strategy.

[0070] In some embodiments, the "output data of the near-end strategy optimization model" can be the power allocation coefficients output by the Actor network under the current state vector input conditions. Based on the power allocation coefficients and the real-time operating states of the flywheel energy storage module and the electrochemical energy storage module, the control module generates power allocation commands corresponding to the flywheel energy storage module and the electrochemical energy storage module.

[0071] In some embodiments, obtaining power allocation instructions between the flywheel energy storage module and the electrochemical energy storage module based on output data includes: correcting the output data based on constraints; when the corrected output data satisfies the constraints, the corrected output data is used as the power allocation instructions between the flywheel energy storage module and the electrochemical energy storage module. The constraints include the rotational speed constraint of the flywheel energy storage module, the charge / discharge rate constraint of the electrochemical energy storage module, and the state of charge constraint of the electrochemical energy storage module.

[0072] In some embodiments, the output data is corrected based on constraints because the power allocation coefficients output by the near-end strategy optimization model represent the theoretically optimal control result calculated based on the current state space, which may not directly meet the actual operational constraints of the flywheel energy storage module and the electrochemical energy storage module. Therefore, before generating power allocation commands, the control module performs constraint verification on the power allocation coefficients and makes necessary corrections based on the constraints to ensure that the generated power allocation commands meet the safe operation requirements of each energy storage module and improve the executability of the control strategy.

[0073] In some embodiments, the specific process of correcting the output data based on constraints can be as follows: The Actor network in the near-end strategy optimization model outputs the power allocation coefficient between the flywheel energy storage module and the electrochemical energy storage module based on the state vector. The control module performs constraint verification on the power allocation coefficient according to the real-time operating status of the flywheel energy storage module and the electrochemical energy storage module. When the power allocation coefficient does not meet the constraints, the control module limits the amplitude or redistributes the power allocation coefficient according to the speed constraints of the flywheel energy storage module, the charge / discharge rate constraints of the electrochemical energy storage module, and the state of charge constraints of the electrochemical energy storage module. When the corrected power allocation coefficient meets the constraints, the control module generates power allocation instructions between the flywheel energy storage module and the electrochemical energy storage module based on the corrected power allocation coefficient, thereby improving the safety and reliability of the energy storage system while ensuring the optimization of the control strategy.

[0074] In some embodiments, the constraints include the rotational speed constraint of the flywheel energy storage module, the charge / discharge rate constraint of the electrochemical energy storage module, and the state of charge (SOC) constraint of the electrochemical energy storage module. The specific reasons for these constraints are as follows: the rotational speed constraint of the flywheel energy storage module is used to limit the flywheel speed to within the operating range allowed by its safe mechanical structure, to avoid excessive mechanical stress overload due to excessively high speed or insufficient energy support capacity due to excessively low speed; the charge / discharge rate constraint of the electrochemical energy storage module is used to limit the range of charge / discharge current of the battery per unit time, to avoid accelerated battery aging or thermal safety risks due to high-rate charge / discharge; the SOC constraint of the electrochemical energy storage module is used to limit the battery SOC to a reasonable range, to avoid adverse effects on battery life and safety due to overcharging or over-discharging. For example, the rotational speed constraint of the flywheel energy storage module is 30%–80% of the rated speed, the charge / discharge rate constraint of the electrochemical energy storage module is ≤0.5C, and the SOC constraint of the electrochemical energy storage module is 20%–80%. However, it should be noted that the present invention is not limited to the above specific values ​​and is not limited herein.

[0075] In some embodiments, the corrected output data not only ensures the safe operation of the flywheel energy storage module and the electrochemical energy storage module, but also further maintains the optimized characteristics of the power distribution strategy under the premise of meeting the constraints. This enables the flywheel energy storage module to undertake the task of high-frequency power component regulation, and the electrochemical energy storage module to undertake the task of medium and low-frequency energy regulation, thereby achieving the stability and coordination of the overall operation of the energy storage system.

[0076] In some embodiments, the power allocation command includes: allocating high-frequency power components to the flywheel energy storage module for power regulation, and allocating mid- and low-frequency power components to the electrochemical energy storage module for power regulation.

[0077] In some embodiments, high-frequency power components refer to power components in grid power demand that change rapidly, have short durations, and fluctuate at high frequencies. These correspond to transient power changes caused by rapid grid frequency fluctuations, instantaneous load changes, or rapid disturbances in renewable energy output. Because these power components are characterized by rapid changes and short response times, energy storage modules need to be able to adjust charging and discharging power within a very short time to quickly respond to changes in grid power demand. Conversely, low- and medium-frequency power components refer to power components in grid power demand that change relatively slowly, have long durations, and fluctuate at low frequencies. These correspond to power changes caused by load changes, slow changes in renewable energy output, or long-term energy balance requirements. Because these power components have longer durations, the focus is more on continuous energy support capabilities, while the requirement for instantaneous response speed is relatively lower.

[0078] In some embodiments, the allocation of high-frequency power components to flywheel energy storage modules for power regulation and the allocation of mid-to-low-frequency power components to electrochemical energy storage modules for power regulation are determined based on the different physical and operational characteristics of the two types of energy storage modules. Specifically, flywheel energy storage modules have characteristics such as fast response speed, high power density, and long cycle life, enabling them to quickly complete charge-discharge switching and making them suitable for undertaking the rapid power regulation task corresponding to high-frequency power components; while electrochemical energy storage modules have higher energy density and continuous energy output capability, making them more suitable for undertaking the continuous energy regulation task corresponding to mid-to-low-frequency power components. By matching high-frequency power components with flywheel energy storage modules and mid-to-low-frequency power components with electrochemical energy storage modules, the rapid response advantage of flywheel energy storage modules and the long-term energy support advantage of electrochemical energy storage modules can be fully utilized. This achieves precise matching between different power components and the operating characteristics of energy storage modules, improves the collaborative operation capability between flywheel energy storage modules and electrochemical energy storage modules, and further enhances the rationality and operational efficiency of the overall power regulation of the energy storage system.

[0079] In some embodiments, high-frequency power components and mid-to-low-frequency power components can be obtained by frequency component analysis of the target power signal corresponding to the grid power demand signal or the automatic generation control frequency modulation command. For example, digital filtering, frequency domain decomposition, wavelet decomposition, or other signal decomposition methods can be used to decompose the target power signal into power components in different frequency ranges. Power components with a frequency variation higher than a preset frequency threshold are designated as high-frequency power components, and power components with a frequency variation lower than the preset frequency threshold are designated as mid-to-low-frequency power components. For example, preferably, the preset frequency threshold can be set to 0.1Hz, meaning that power components with a frequency variation greater than 0.1Hz are allocated to the flywheel energy storage module for power regulation, and power components with a frequency variation less than or equal to 0.1Hz are allocated to the electrochemical energy storage module for power regulation. Of course, the preset frequency threshold can be adjusted according to the grid operating conditions, energy storage system capacity configuration, or control requirements, and is not limited to the example values ​​described above.

[0080] In some embodiments, power regulation of the flywheel energy storage module and the electrochemical energy storage module is controlled based on power distribution commands, including: synchronously generating gate drive signals corresponding to the converters of the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected inverter, respectively, based on the power distribution commands. Based on the gate drive signals, synchronous control is performed on the converters of the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected inverter to achieve power conversion.

[0081] In some embodiments, the gate drive signal may include a pulse width modulation (PWM) drive pulse, a space vector pulse width modulation (SVPWM) drive pulse, or other drive signals used to control the on and off of power semiconductor devices, so as to control each converter and inverter to complete the bidirectional power conversion between AC power and DC power, and realize the energy exchange between the flywheel energy storage module, the electrochemical energy storage module and the grid.

[0082] In some embodiments, by synchronously generating gate drive signals for the converters of the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected inverter, and by synchronously controlling the aforementioned devices based on these gate drive signals, microsecond-level timing synchronization of the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected inverter can be achieved. This effectively reduces synchronization errors caused by differences in control timing among multiple power conversion devices, and improves the consistency and coordination of the power exchange process between the flywheel energy storage module, the electrochemical energy storage module, and the grid. Simultaneously, each power conversion device can collaboratively complete charging and discharging control and grid-connected power regulation according to a unified control rhythm, improving the overall response speed, power control accuracy, and power point tracking performance of the energy storage system, enabling power allocation strategies to be applied more promptly and accurately to the actual power regulation process.

[0083] In some embodiments, the control method for the energy storage system further includes: comparing and analyzing first operating status information, second operating status information, third operating status information, environmental information, and electricity market operating information with historical operating status information. Based on the comparison and analysis results, anomaly predictions are made regarding the operating status of the power grid, the operating status of the flywheel energy storage module, and the operating status of the electrochemical energy storage module.

[0084] Specifically, the control module continuously collects first operating status information, second operating status information, third operating status information, environmental information, and electricity market operating information, and performs unified time-series alignment and feature extraction with historical operating status information to obtain the correspondence between current operating status feature data and historical operating status feature data.

[0085] Furthermore, based on historical operating status information, historical normal operation models and abnormal evolution models are constructed for both the grid side and the energy storage side. These models characterize the feature distribution under different operating states. Then, the current operating state feature data is input into the historical state model for similarity matching and deviation calculation. When the distribution of the current operating state in the feature space gradually approaches the historical abnormal evolution trajectory, or when the similarity to historical abnormal states exceeds a preset threshold, an abnormal evolution trend is determined to exist. Based on this, the abnormal prediction result is output within a time window with a preset prediction lead time, thus providing a basis for subsequent global synchronization control and protection strategy adjustments. This enables early identification and preventative control of grid faults and potential abnormalities in flywheel energy storage modules and electrochemical energy storage modules.

[0086] Preferably, the preset prediction time advance can be 200ms, but it should be noted that the present invention is not limited to the above specific value and is not limited here.

[0087] Therefore, by correlating and analyzing the current operating status information with the historical operating status information, the control module not only makes judgments based on the real-time status, but also introduces historical operating status information as a reference, thereby enhancing the ability to perceive the evolution trend of the operating status on the grid side and the energy storage side. This enables the system to output early warning information before the abnormal state has fully occurred, and adjust the control strategy of the energy storage system accordingly, thereby achieving preventive control of potential anomalies and improving the safety of system operation.

[0088] In some embodiments, the control method for the energy storage system further includes: acquiring the power tracking error corresponding to the power allocation command, information on changes in the operating status of the energy storage modules, and abnormal information of the energy storage system, as feedback information for execution results. Based on the feedback information for execution results, the multi-timescale state prediction, the near-end strategy optimization model, and the synchronization control strategy between each energy storage module and the power grid are updated online.

[0089] In some embodiments, power point tracking error (PPI) can refer to the deviation between the target power corresponding to the power allocation command and the actual output power of the flywheel energy storage module and the electrochemical energy storage module, used to characterize the tracking accuracy of the power allocation command during actual execution. By obtaining PPI, the control accuracy of the current power control strategy can be evaluated, and it can serve as an important basis for subsequent control strategy optimization and model updates, thereby improving the dynamic adjustment capability of the energy storage system during grid-connected operation.

[0090] In some embodiments, the energy storage module operating status change information can refer to the changes in key operating parameters of the flywheel energy storage module and the electrochemical energy storage module during power regulation. Examples include changes in the rotational speed, temperature, and vibration of the flywheel energy storage module, and changes in the state of charge, voltage, current, temperature, and health status of the electrochemical energy storage module. This information is used to characterize the dynamic response characteristics and health status evolution trend of the energy storage module during power regulation. The purpose of obtaining energy storage module operating status change information is to reflect the degree of influence of the power allocation strategy on the internal state of the energy storage system, thereby assessing the impact of the control strategy on the safety and lifespan characteristics of the energy storage module.

[0091] In some embodiments, the abnormal information of the energy storage system can refer to comprehensive characterization data of abnormal disturbances on the grid side, abnormal operation information of the flywheel energy storage module, and abnormal operation information of the electrochemical energy storage module. For example, it includes information such as abnormal fluctuations in grid frequency, voltage exceeding limits, abnormal flywheel speed, abnormal battery temperature, or exceeding state of charge limits, used to characterize whether abnormal operating conditions exist during system operation. The purpose of acquiring abnormal information is to provide a safety constraint basis for the control strategy, enabling the system to promptly identify potential fault risks and trigger control strategy adjustments or protection mechanisms.

[0092] In some embodiments, the online updating of multi-timescale state prediction, near-end strategy optimization model, and synchronization control strategy between each energy storage module and the power grid based on execution result feedback information can refer to the control module constructing feedback learning samples based on power tracking error, energy storage module operating state change information, and anomaly information, and using the feedback learning samples to jointly optimize and update the parameters of the multi-timescale state prediction model, reinforcement learning strategy network parameters, and synchronization control strategy parameters, thereby forming a closed-loop adaptive control mechanism.

[0093] Specifically, the control module first calculates the reward value for the current control cycle based on the power tracking error. This reward value characterizes the comprehensive performance of the multi-objective optimization results, including power tracking accuracy, energy storage lifetime impact, system operating efficiency, and economic indicators. Then, a multi-objective optimization function is constructed based on the reward value. This function includes objectives for minimizing power tracking error, minimizing electrochemical energy storage lifetime loss, maximizing system operating efficiency, and optimizing economic performance over the entire lifecycle. The weights of each optimization objective in the multi-objective optimization function are adaptively adjusted according to grid operating conditions, the energy storage module's state of equilibrium (SOH), and time-of-use pricing dynamics, enabling the system to dynamically balance the relationship between tracking performance, lifetime loss, and economic efficiency under different operating conditions.

[0094] Furthermore, the current state vector, the power allocation action output by the proximal policy optimization model, the reward value, and the next state vector are collectively constructed as reinforcement learning training samples and stored in the experience sample library. These samples are used to update the Actor network and Critic network in the proximal policy optimization model online. The Actor network continuously adjusts the power allocation strategy based on the state value evaluation results and advantage function output by the Critic network, so that the power allocation command gradually approaches the optimal control strategy.

[0095] Meanwhile, the multi-timescale state prediction model is corrected and its parameters are updated by using the feedback information of the execution results, so that the prediction results can be closer to the actual operating state evolution trend, and the synchronization control strategy is dynamically adjusted to correct the timing deviation and control deviation in the gate drive signal generation process, thereby improving the overall synchronization control accuracy and stability of the system.

[0096] In summary, through the aforementioned closed-loop update mechanism based on feedback information, the energy storage system forms a self-learning optimization closed-loop structure of "state perception - predictive analysis - strategy decision-making - execution control - result feedback - model update". This enables the multi-timescale state prediction capability, power allocation strategy optimization capability, and synchronization control capability to be continuously iterated and optimized with the operating data, thereby achieving adaptive optimal control and long-term continuous performance improvement of the energy storage system under different grid operating conditions.

[0097] In some embodiments, to ensure stable learning and gradual convergence of the network, corresponding algorithm parameters can be set during model training or online updates. For example, preferably, the learning rate can be set to 0.0003 to control the magnitude of each parameter update for the Actor and Critic networks, making the policy update process smoother and avoiding training oscillations caused by excessive update magnitude; the discount factor can be set to 0.99 to calculate the degree of influence of future rewards on the current decision, enabling the near-end policy optimization model to optimize power tracking performance while taking into account the long-term operating status, lifespan loss, and overall system lifecycle economy of the energy storage module; the batch size can be set to 64, meaning that after accumulating 64 training samples consisting of states, actions, rewards, and the state at the next time step, the parameters of the Actor and Critic networks are updated once, thus balancing model training efficiency, policy convergence stability, and online learning performance. It should be noted that the above parameters are only examples, and in practical applications, they can be adjusted according to the scale of the energy storage system, operating conditions, and model training requirements. This invention is not limited to the specific values ​​mentioned above and is not restricted in this regard.

[0098] In some embodiments, during the operation of the energy storage system, when the control module detects an abnormal grid operating status or an operational abnormality in the flywheel energy storage module and the electrochemical energy storage module based on the abnormality prediction result or execution result feedback information, the fault handling and recovery control process is triggered, and the power distribution command is dynamically adjusted.

[0099] Specifically, when a grid fault such as a voltage dip is detected, the control module controls the flywheel energy storage module to rapidly release stored energy based on adjusted power distribution commands, providing rapid power and inertia support. Simultaneously, it controls the electrochemical energy storage module to maintain DC bus voltage stability, thereby achieving fault ride-through capability for the energy storage system. When an abnormal operating state is detected in the electrochemical energy storage module, such as excessively high battery cell temperature, the control module reduces the charge / discharge rate and power load ratio of the electrochemical energy storage module, and correspondingly increases the power support ratio of the flywheel energy storage module, reducing thermal stress and lifespan loss. Once the grid operating status and the operating status of each energy storage module return to normal, the control module automatically switches back to normal frequency regulation control mode, thereby restoring the energy storage system to a stable grid-connected frequency regulation operating state, achieving closed-loop control for fault handling and operation recovery.

[0100] In some embodiments, the control method for the energy storage system further includes: performing data preprocessing on the first operating state information, the second operating state information, the third operating state information, environmental information, and electricity market operating information, thereby improving the reliability and accuracy of subsequent multi-timescale state prediction and power allocation decisions.

[0101] In some embodiments, data preprocessing may include uniformly cleaning and aligning various types of raw operational data with the time dimension, and normalizing data of different dimensions to eliminate scale differences between different data sources.

[0102] In some embodiments, data preprocessing may further include data cleaning based on statistical rules, such as using the 3σ criterion to remove outliers in the collected data to eliminate abnormal data points caused by sensor errors or communication fluctuations. Interpolation methods are also used to complete missing data, such as linear interpolation or sliding window interpolation, to maintain the continuity of the data sequence and thus avoid interference from missing data on the prediction model input.

[0103] In some embodiments, data preprocessing may further include performing consistency checks on the current data based on historical operating status information and marking potentially abnormal data points, thereby achieving a preliminary health assessment of the system's operating status and enabling subsequent multi-timescale state prediction models to be trained and inferred on a higher quality data basis.

[0104] Therefore, the above data preprocessing process makes the data input to the subsequent multi-timescale state prediction and near-end strategy optimization models more reliable and consistent, thereby improving the stability and accuracy of energy storage system control.

[0105] Figure 2 This is an overall flowchart of a control method for an energy storage system according to an embodiment of the present invention, as follows: Figure 2 As shown, the overall flow of the control method for the energy storage system in this embodiment of the invention includes at least the following steps: S10, the energy storage system performs the system initialization process before being put into operation.

[0106] Specifically, after the energy storage system is powered on, the control module performs unified initialization control on the flywheel energy storage module, electrochemical energy storage module, grid-connected inverter, and DC bus. The control module also loads the pre-trained near-end strategy optimization model and multi-timescale state prediction model, and starts the data acquisition and preprocessing process to obtain the initial operating status information of the system and complete the initial health status assessment.

[0107] Furthermore, the system is pre-charged via the DC bus to stabilize the DC bus voltage to a preset rated voltage. The grid-connected inverter completes synchronization and grid connection preparation with the grid. After completing self-test, the flywheel energy storage module increases to a preset speed and enters standby mode. The electrochemical energy storage module completes battery status detection and initialization calibration through the battery management system and enters standby operation mode. The preset rated voltage is preferably 1500V, and the preset speed is preferably 50% of the rated speed. However, it should be noted that the invention is not limited to these specific values ​​and is not restricted here.

[0108] Therefore, through the above initialization process, the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected inverter are all in a stable and controllable standby state, and the control module completes model loading and data initialization, thus laying the operational foundation for receiving automatic power generation control frequency regulation commands from the grid and executing power distribution control.

[0109] S11, the multi-source data acquisition module in the control module receives the AGC frequency regulation command from the power grid, and simultaneously acquires the first operating status information of the power grid, the second operating status information of the flywheel energy storage module, the third operating status information of the electrochemical energy storage module, environmental information, and electricity market operation information.

[0110] Specifically, after receiving the AGC frequency regulation command from the power grid, the multi-source data acquisition module in the control module simultaneously acquires the following operating status information at a preset sampling frequency: first operating status information such as voltage, current, frequency, and phase information from the power grid side; second operating status information such as flywheel speed, vibration parameters, winding temperature, bearing temperature, and current information; third operating status information such as voltage, current, temperature, internal resistance, state of charge, and health status information of individual battery cells; external operating environment parameters such as ambient temperature and humidity; time-of-use electricity price information; power market dispatch plan information; and electricity market transaction-related operating data. The preset sampling frequency is preferably 1kHz, and the data transmission delay is preferably less than or equal to 1ms. However, it should be noted that this invention is not limited to the above specific values ​​and is not restricted here.

[0111] S12, perform data preprocessing on the first operating status information, the second operating status information, the third operating status information, environmental information, and electricity market operating information.

[0112] The data preprocessing includes: cleaning and normalizing the collected raw operating status information; removing outliers using the 3σ criterion and completing missing data through linear interpolation; and conducting an initial assessment of the system's health status based on historical operating status information and marking potential anomalies.

[0113] S13. Based on the first operating status information, the second operating status information, the third operating status information, the environmental information, and the electricity market operating information, perform multi-time-scale state prediction to obtain the future operating status information of the energy storage system at different time scales.

[0114] Specifically, multi-timescale state prediction includes: millisecond-level prediction of grid frequency and power fluctuations based on first operating state information; minute-level prediction of load changes and renewable energy output based on first operating state information, environmental information, and electricity market operation information; and long-cycle timescale prediction of lifetime degradation trends based on second and third operating state information.

[0115] S14, input the first operating status information, the second operating status information, the third operating status information, environmental information, electricity market operating information, and future operating status information into the near-end strategy optimization model to obtain the output data of the near-end strategy optimization model, and obtain the power allocation command between the flywheel energy storage module and the electrochemical energy storage module based on the output data.

[0116] Specifically, after obtaining future operating status information, the system combines the current first operating status information of the power grid, the second operating status information of the flywheel energy storage module, the third operating status information of the electrochemical energy storage module, as well as environmental information and electricity market operating information, to identify the current operating scenario and determine whether the energy storage system has switched from one operating scenario to another, such as switching from the power grid AGC frequency regulation scenario to the new energy consumption scenario, from the new energy consumption scenario to the industrial and commercial peak-valley arbitrage scenario, or switching to the microgrid isolated grid operation scenario, etc.

[0117] Furthermore, when a change in operating scenario is detected, the system adjusts the weights of each optimization objective in the multi-objective optimization function according to the control objectives corresponding to the new operating scenario. These optimization objectives include power point tracking error, energy storage module lifespan loss, energy storage system operating efficiency, and overall lifecycle economics. Simultaneously, the system updates the corresponding constraints based on the new operating scenario, including the allowable speed range of the flywheel energy storage module, the state of charge range of the electrochemical energy storage module, and charge / discharge rate limits, ensuring the control strategy meets the operational requirements of the current operating scenario. Then, the system uses a ramp-transition method to smoothly switch control parameters, achieving a seamless switching of operating modes to avoid power surges, bus voltage fluctuations, or control shocks during scenario switching.

[0118] Furthermore, after completing the scene switch, the system reconstructs the state vector of the near-end policy optimization model based on the updated optimization target weights, constraints, current real-time running status information, and future running status information, and then enters the subsequent reinforcement learning decision-making process; if no change in the running scene is detected, the current optimization target weights and constraints remain unchanged, and the current state vector is directly used to enter the reinforcement learning decision-making process.

[0119] Further, the system enters the reinforcement learning decision-making stage. The system combines real-time operating status information collected at the current moment with future operating status information predicted at multiple time scales to form the state vector of the near-end strategy optimization model. The state vector is a data set that comprehensively characterizes the current operating status and future development trend of the energy storage system. It includes the first operating status information of the power grid, the second operating status information of the flywheel energy storage module, the third operating status information of the electrochemical energy storage module, environmental information, electricity market operating information, and future operating status information. The near-end strategy optimization model receives the state vector as input. The Actor network outputs a power allocation coefficient between the flywheel energy storage module and the electrochemical energy storage module based on the state vector. This power allocation coefficient represents the target power allocation relationship between the flywheel energy storage module and the electrochemical energy storage module. The Critic network estimates the current state value based on the state vector and calculates the advantage function by combining the reward value after strategy execution and the state vector at the next moment. This evaluates the value of the strategy output by the Actor network, providing a value assessment basis for subsequent online updates of the Actor network and the Critic network, thus enabling the near-end strategy optimization model to continuously learn the optimal power allocation strategy under different operating conditions.

[0120] Furthermore, since the power allocation coefficients output by the Actor network represent the theoretically optimal control results in the current state space, the system further enters the constraint verification and correction stage. The control module performs constraint verification on the power allocation coefficients based on the real-time operating status of the flywheel energy storage module and the electrochemical energy storage module, including rated power, flywheel speed, state of charge (SOC), state of health (SOH), charge / discharge rate, and operating temperature. When the power allocation coefficients do not meet the preset constraints, the control module applies amplitude limits, proportional redistribution, or boundary mapping to the power allocation coefficients based on the flywheel energy storage module's speed constraints, the electrochemical energy storage module's charge / discharge rate constraints, and the electrochemical energy storage module's state of charge constraints, ensuring that the corrected power allocation coefficients meet the preset constraints. Then, the control module generates power allocation commands corresponding to the flywheel energy storage module and the electrochemical energy storage module based on the corrected power allocation coefficients, ensuring that the generated commands meet the safe operation requirements of the flywheel energy storage module, the electrochemical energy storage module, and the overall energy storage system. At the same time, the control module can also automatically perform adaptive power frequency division according to the power change frequency, giving priority to allocating high-frequency power components to the flywheel energy storage module for adjustment, and giving priority to allocating medium and low-frequency power components to the electrochemical energy storage module for adjustment, thereby giving full play to the complementary advantages of the flywheel's fast response speed and the battery's large energy capacity.

[0121] S15, based on the power distribution command, synchronously generates gate drive signals corresponding to the converters of the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected inverter, respectively. Based on the gate drive signals, synchronous control is performed on the converters of the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected inverter to achieve power conversion.

[0122] Specifically, the system enters the global synchronization control phase based on the power allocation command. The global synchronization control module uniformly generates the gate drive signals corresponding to the converters of the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected inverter, and performs unified timing coordination on each gate drive signal to achieve microsecond-level timing synchronization control among multiple converters, avoiding power fluctuations and control shocks caused by control asynchrony. Simultaneously, the system combines current full-scale operating status information and historical operating status models to predict anomalies in the grid operating status, flywheel energy storage module operating status, and electrochemical energy storage module operating status, anticipating potential grid faults or energy storage system anomalies in advance, and adjusting the system operating status accordingly to achieve pre-fault control. Preferably, the timing synchronization error among multiple converters can be less than or equal to 1 μs (microseconds), but it should be noted that the invention is not limited to the above specific values ​​and is not restricted herein.

[0123] Furthermore, each converter executes the corresponding power conversion process according to the synchronously generated gate drive signal, controls the flywheel energy storage module and electrochemical energy storage module to complete the charging and discharging action, and achieves the target power output.

[0124] S16: Obtain the power tracking error corresponding to the power allocation command, the energy storage module operating status change information, and the energy storage system abnormal information as execution result feedback information. Based on the execution result feedback information, update the multi-timescale state prediction, the near-end strategy optimization model, and the synchronization control strategy between each energy storage module and the power grid online.

[0125] Specifically, the control module first calculates the reward value for the current control cycle based on the power tracking error. This reward value characterizes the comprehensive performance of multiple optimization objectives, including power tracking accuracy, lifespan loss of the dual energy storage system, system operating efficiency, and overall lifespan economics. Then, a multi-objective optimization function is constructed based on the reward value. This function includes objectives for minimizing power tracking error, minimizing electrochemical energy storage lifespan loss, maximizing system operating efficiency, and optimizing overall lifespan economics. The weights of each optimization objective in the multi-objective optimization function are adaptively adjusted according to grid conditions, the SOH status of the energy storage module, and time-of-use pricing dynamics, enabling the system to dynamically balance the relationship between tracking performance, lifespan loss, and economics under different operating conditions.

[0126] Furthermore, the current state vector, the power allocation coefficients output by the proximal policy optimization model, the reward value, and the state vector at the next moment are collectively constructed as reinforcement learning training samples and stored in the experience sample library. These samples are used to update the Actor network and Critic network in the proximal policy optimization model online. The Actor network continuously adjusts the power allocation strategy based on the state value evaluation results and the advantage function output by the Critic network, so that the power allocation command gradually approaches the optimal control strategy.

[0127] Meanwhile, the multi-timescale state prediction model is corrected and its parameters are updated by using the feedback information of the execution results, so that the prediction results can be closer to the actual operating state evolution trend, and the synchronization control strategy is dynamically adjusted to correct the timing deviation and control deviation in the gate drive signal generation process, thereby improving the overall synchronization control accuracy and stability of the system.

[0128] S17, repeat steps S10-S16, continuously responding to the AGC frequency regulation command of the power grid.

[0129] In summary, this invention achieves online self-evolution of control strategies through a closed-loop architecture encompassing "multi-source data acquisition → multi-timescale state prediction → multi-target reinforcement learning power allocation → global synchronous control → execution result feedback." It fundamentally solves the core defects of existing technologies, such as open-loop control lag, fixed frequency division coarseness, decoupling of real-time control and lifespan management, and passive threshold protection lag. It represents an intelligent hybrid energy storage control technology that upgrades from "passive reaction" to "active prediction."

[0130] In some embodiments, existing multi-objective optimization technologies primarily focus on the dual objectives of "frequency regulation accuracy + battery life" or "investment cost + operating cost," failing to deeply couple and optimize the four core objectives: power point tracking accuracy, lifespan loss of dual energy storage systems, system operating efficiency, and overall lifespan economics. Furthermore, the optimization objective weights are fixed, unable to dynamically adapt to changes in grid conditions, energy storage health, and market electricity prices, easily leading to single-objective optimization and global performance degradation. Moreover, the prevalence of offline optimization methods such as particle swarm optimization and genetic algorithms cannot meet the dynamic requirements of real-time control.

[0131] The core advantages of this invention lie in: constructing a multi-objective optimization function with four core objectives—power point tracking accuracy, lifespan loss of dual energy storage, system operating efficiency, and lifespan economy—breaking through the objective dimension barriers of existing technologies; dynamically and adaptively adjusting the weights of each optimization objective based on grid operating conditions, energy storage SOH status, and time-of-use pricing to achieve global optimization under all operating conditions, rather than local optimization under fixed operating conditions; extending multi-objective optimization from offline capacity configuration to online real-time power allocation, realizing multi-objective collaboration across all levels from the planning layer to the operation layer to the control layer, and solving the core problem of disconnect between offline optimization and real-time control in existing technologies.

[0132] In some embodiments, existing energy storage control methods are based solely on single-dimensional load forecasting or renewable energy output forecasting, or perform PID (Proportion-Integration-Differentiation) parameter optimization only for a single energy storage unit. Their control processes often employ an open-loop structure of "prediction first, control later," with the prediction model and control decision model operating independently, failing to form a closed-loop collaborative optimization mechanism. Furthermore, the prediction algorithm is not deeply integrated with the physical characteristics of the flywheel energy storage module and the electrochemical energy storage module, failing to fully leverage the millisecond-level response hardware advantage of the flywheel. This results in lag in control response, frequent high-frequency charge-discharge shocks to the battery, and the inability to fully realize the complementary advantages of hybrid energy storage.

[0133] The core advantage of this invention lies in its construction of a full-link closed-loop control system encompassing multi-timescale state prediction, power allocation, and underlying converter collaborative control. Based on execution result feedback, it updates the multi-timescale state prediction, near-end strategy optimization model, and synchronization control strategies between each energy storage module and the grid online, completely resolving the lag problem of existing open-loop control technologies. Furthermore, based on the different physical response characteristics of flywheel and electrochemical energy storage modules, it adapts the millisecond-level high-frequency prediction mechanism to the fast response characteristics of the flywheel energy storage module, enabling it to better handle high-frequency power regulation demands. Simultaneously, it adapts the minute-level short-to-medium-term prediction mechanism to the long-term energy support characteristics of the electrochemical energy storage module, enabling it to better handle medium-to-low-frequency energy regulation demands, thereby achieving adaptive power component allocation among different energy storage modules. This avoids the coarse matching problem caused by the fixed frequency division method in existing technologies, thus improving the precision of power allocation and the overall coordination of system operation.

[0134] In some embodiments, the existing technology mainly designs control strategies for single application scenarios such as grid frequency regulation and new energy consumption. Scenario switching requires re-adjusting the model and parameters, resulting in poor engineering adaptability. At the same time, the threshold-triggered passive protection architecture is adopted, with the upper-layer scheduling, middle-layer power allocation, and lower-layer control layers completely decoupled, resulting in delayed fault response, high malfunction rate, and insufficient system reliability.

[0135] The core advantage of this invention lies in its three-layer fully closed-loop collaborative control architecture: upper-layer global optimization scheduling, mid-layer adaptive power allocation, and lower-layer collaborative converter control. This architecture enables real-time data interaction and bidirectional command feedback between the three layers, resolving the response delay problem caused by hierarchical decoupling in existing technologies. Furthermore, this invention enables seamless switching across four core scenarios: grid AGC frequency regulation, wind and solar renewable energy integration, industrial and commercial peak-valley arbitrage, and microgrid island operation, without requiring model retraining, significantly reducing engineering implementation and commissioning costs. In addition, a predictive global protection architecture replaces the threshold-triggered passive protection of existing technologies. Through comprehensive analysis of the grid operating status, flywheel energy storage module operating status, and electrochemical energy storage module operating status, it achieves early identification and prediction of potential grid faults and energy storage unit anomalies. It adjusts power allocation commands and system operating strategies before a fault occurs, thereby achieving full-process protection and control including pre-fault prevention, stable operation during a fault, and rapid recovery after a fault.

[0136] In some embodiments, the present invention can be directly verified through MATLAB / Simulink simulation and prototype experiments, achieving a comprehensive performance improvement compared to existing technologies. The control method for the energy storage system achieves coordinated optimization control of the grid operation status, flywheel energy storage module operation status, and electrochemical energy storage module operation status by uniformly modeling and analyzing the automatic generation control frequency regulation scenario on the grid side and the new energy grid-connected operation scenario. This results in an overall improvement in power point tracking performance, system stability, and power quality.

[0137] Specifically, based on the differentiated dynamic response characteristics of flywheel energy storage modules and electrochemical energy storage modules, the energy storage system performs high-precision tracking of AGC frequency regulation commands, which significantly reduces the frequency regulation response time of the energy storage system and enhances its frequency support capability during the primary frequency regulation process of the power grid, thereby strengthening the inertia support and frequency stability of high-proportion new energy power systems.

[0138] Furthermore, in grid-connected scenarios where the output of new energy sources fluctuates significantly, such as wind power or photovoltaic power, the energy storage system can quickly smooth out power fluctuations based on the synergistic power regulation capabilities of the flywheel energy storage module and the electrochemical energy storage module. This reduces the amplitude of DC bus voltage fluctuations and improves the stability and anti-disturbance capability of new energy grid connection.

[0139] Furthermore, by using control methods for energy storage systems, the total harmonic distortion rate of grid-connected current can be reduced, ensuring that the grid-connected power quality meets the relevant grid connection standards. This effectively avoids harmonic amplification problems and improves the quality and safety of power interaction between the energy storage system and the grid.

[0140] In some embodiments, the energy storage system optimizes the collaborative power distribution strategy between the flywheel energy storage module and the electrochemical energy storage module, thereby preventing the electrochemical energy storage module from undertaking high-frequency low-current charging and discharging and deep charging and discharging conditions during operation. This reduces its equivalent cycle count and slows down the capacity decay process, thereby extending the cycle life of the electrochemical energy storage module and reducing safety risks such as thermal runaway, thus extending its retirement cycle under the same frequency regulation conditions.

[0141] In some embodiments, the flywheel energy storage module is subjected to operating range constraint control based on power distribution commands, so that the flywheel energy storage module always operates within the target range of rated speed, thereby reducing the time of inefficient operation, reducing bearing wear and wind resistance loss, and improving the mechanical service life of the flywheel energy storage module so that it can match the overall design life of the power station.

[0142] In some embodiments, based on the coordinated operation of the flywheel energy storage module and the electrochemical energy storage module, the energy storage system improves the overall energy efficiency and economy of the system through the power distribution commands of the flywheel energy storage module and the electrochemical energy storage module and the underlying converter synchronous control strategy.

[0143] Specifically, by optimizing power allocation based on multi-timescale state prediction results and combining the control strategy output by the near-end strategy optimization model, adaptive sharing of power between the flywheel energy storage module and the electrochemical energy storage module under different power components is achieved. This reduces converter switching losses, improves the charge-discharge cycle efficiency of the energy storage system, and enhances the overall energy efficiency of the system. Simultaneously, in grid AGC frequency regulation scenarios, the energy storage system can achieve high-precision power tracking based on power allocation commands, improving overall frequency regulation performance and increasing the frequency regulation ancillary service revenue per unit installed capacity. Furthermore, by synergistically optimizing multi-timescale state prediction, the near-end strategy optimization model, and the synchronous control strategy, the energy storage system can reduce operating losses and equipment degradation rates throughout its entire lifecycle. This extends the cycle life of the electrochemical energy storage module and the mechanical life of the flywheel energy storage module, reduces the levelized cost of electricity (LCOE) over its entire lifecycle, and improves the long-term economic efficiency and engineering scalability of the system.

[0144] In some embodiments, the energy storage system adopts a modular topology structure with a DC bus, enabling flywheel energy storage modules and electrochemical energy storage modules to be connected to the energy storage system in a standardized module form. This allows for flexible adaptation to different application scenarios such as new energy power plants, thermal power frequency regulation systems, industrial and commercial load sides, and grid-side energy storage, and completes engineering-level expansion deployment without changing the core control methods and system architecture.

[0145] In some embodiments, a multi-scenario adaptive model constructed based on first operating status information, second operating status information, third operating status information, environmental information, and the electricity market operating information enables the energy storage system to achieve adaptive parameter adjustment and rapid matching of control strategies during on-site deployment, thereby reducing the amount of manual on-site debugging and shortening the system debugging cycle.

[0146] Furthermore, by using the feedback information from the execution results to update the multi-timescale state prediction model, the near-end strategy optimization model, and the synchronous control strategy online, the energy storage system is equipped with continuous optimization capabilities during operation, enabling rapid location of operational anomalies and intelligent generation of operation and maintenance strategies. This reduces operation and maintenance complexity and labor costs, and improves the maintainability and adaptability of system engineering applications.

[0147] In some embodiments, the present invention builds a simulation model of a hybrid energy storage system consisting of a 1MW flywheel and a 2MW / 2MWh battery based on MATLAB / Simulink, and compares and verifies the control method of the energy storage system with the existing fixed frequency division control strategy. The simulation scenario is a 24-hour operating condition, and actual power grid AGC frequency regulation command data is input to evaluate the control performance and operation effect of the system.

[0148] Table 1. Simulation results comparing the performance of the control method of the present invention with existing fixed frequency division control strategies.

[0149] In summary, based on Table 1 above, the present invention demonstrates significant improvements in power point tracking accuracy, dynamic response speed, electrochemical energy storage module lifespan loss control, system energy efficiency, and overall frequency regulation performance, verifying the superior technical effect of the synergistic effect of multi-timescale state prediction, near-end strategy optimization model, and synchronous control strategy.

[0150] In some embodiments, during a primary frequency regulation experiment in the power grid, the energy storage system, under the coordinated power regulation of the flywheel energy storage module and the electrochemical energy storage module, can complete a rapid response within ≤5ms and provide inertial support for the rated power when the grid frequency drops by 0.2Hz. Simultaneously, it keeps the DC bus voltage fluctuation within ±0.8% and achieves a total harmonic distortion (THD) of 1.8% for the grid-connected current, meeting the requirements of national standard GB / T 19964-2012. Furthermore, under continuous operation for 72 hours, the charge / discharge rate of the electrochemical energy storage module is constrained and optimized through control methods to maintain it below 0.3C, thereby avoiding high-frequency, low-current charging and discharging conditions.

[0151] Furthermore, through evaluation and modeling analysis of the long-term lifespan impact of flywheel energy storage modules and electrochemical energy storage modules, the results show that the cycle life of electrochemical energy storage modules is extended from approximately 6 years in the existing technology to approximately 10 years, and the number of battery replacements during the entire lifespan is reduced from 3 times to 2 times. Simultaneously, based on the synergistic optimization of power distribution commands and multi-timescale state prediction, the revenue from frequency regulation ancillary services is increased by approximately 85%, and the system investment payback period is shortened from approximately 6.5 years to approximately 3.2 years.

[0152] Furthermore, by analyzing the system's full lifecycle operating costs, the levelized cost of electricity (LCOE) of the energy storage system was reduced by approximately 28.3%, thus verifying the high reliability, high economic efficiency, and scalable value of the flywheel energy storage module and electrochemical energy storage module collaborative control method in engineering applications.

[0153] The following is for reference. Figure 3 An energy storage system according to an embodiment of the present invention is described.

[0154] Figure 3 This is a schematic diagram of an energy storage system according to an embodiment of the present invention, such as... Figure 3 As shown, the energy storage system 100 includes: a grid and load module 1, an electrochemical energy storage module 2, a flywheel energy storage module 3, a grid-connected converter module 4, a DC bus module 5, and a control module 6.

[0155] In some embodiments, the power grid and load module 1 may include an AC power grid 11, a user load / new energy power station 12, a voltage transformer 13, and a current transformer 14.

[0156] In some embodiments, the AC power grid 11 can be used to provide AC power input; the user load / new energy power station 12 can be used to characterize the grid-side load consumption or new energy power output.

[0157] In some embodiments, the AC power grid 11 and the user load / new energy power station 12 are connected in parallel to the AC input terminal of the grid-connected converter module, thereby forming the AC side common access point of the energy storage system 100. Voltage transformers 13 are connected in parallel on the AC bus to collect voltage signals from the grid side in real time, and current transformers 14 are connected in series on the AC bus to collect current signals from the grid side in real time.

[0158] Therefore, through the coordinated measurement of voltage transformer 13 and current transformer 14, the first operating status information of the power grid can be further obtained, including voltage information, current information, frequency information, phase information, etc. on the power grid side, and the first operating status information of the power grid can be transmitted to the control module 6.

[0159] In some embodiments, the electrochemical energy storage module 2 may include a battery cluster 21, a battery management system 22, a bidirectional LLC resonant converter 23, an electrochemical energy storage control module 24, and a battery state sensor 25.

[0160] In some embodiments, the battery cluster 21 is used to provide an electrochemical energy storage carrier to realize the storage and release of low-frequency and long-term energy in the system; the battery management system 22 is used to monitor parameters such as voltage, current, temperature, and internal resistance of individual battery cells in real time, and calculate the state of charge (SOC) and state of health (SOH) of the battery, thereby ensuring the safety and lifespan of the battery operation; the bidirectional LLC resonant converter 23 is used to realize bidirectional power conversion between the battery side and the DC bus 51, and reduces switching losses and improves long-term charging and discharging efficiency through soft switching operation; the electrochemical energy storage control module 24 is used to receive the power distribution command output by the control module 6 and generate the corresponding gate drive signal to control the switching action of the bidirectional LLC resonant converter 23; the battery status sensor 25 is used to collect the voltage, current, and temperature status information of the battery during operation in real time, and feed it back to the battery management system 22 and the control module 6.

[0161] In some embodiments, the battery cluster 21 is connected to the low-voltage DC side of the bidirectional LLC resonant converter 23; the high-voltage DC side of the bidirectional LLC resonant converter 23 is connected to the DC bus 51; the battery status sensor 25 is installed on each cell of the battery cluster 21 and the bus; the output of the battery status sensor 25 is connected to the battery management system 22; the output of the battery management system 22 is connected to the control module 6 to provide the control module 6 with the third operating status information of the electrochemical energy storage module 2; the input of the electrochemical energy storage control module 24 is connected to the control module 6, and the output of the electrochemical energy storage control module 24 is connected to the gate drive terminal of the bidirectional LLC resonant converter 23, thereby realizing precise control of the charging and discharging process of the electrochemical energy storage module 2.

[0162] In some embodiments, the flywheel energy storage module 3 may include a flywheel body 31, a motor / generator 32, a bidirectional three-level Buck-Boost converter 33, a flywheel control module 34, and a flywheel status sensor 35.

[0163] In some embodiments, the flywheel body 31 and the motor / generator 32 can adopt a coaxial rigid connection structure to achieve efficient coupling and conversion between mechanical energy and electrical energy. The motor / generator 32 is used to convert electrical energy into mechanical energy and store it in the flywheel during the charging process of the flywheel energy storage module 3, and to convert the mechanical energy of the flywheel into electrical energy for output during the discharging process.

[0164] In some embodiments, the AC terminal of the bidirectional three-level Buck-Boost converter 33 is connected to the AC terminal of the motor / generator 32, and the DC terminal of the bidirectional three-level Buck-Boost converter 33 is connected to the DC bus 51. This enables bidirectional energy conversion between the flywheel energy storage module 3 and the DC bus 51, and adapts to the wide speed range operating conditions of the flywheel, thereby improving energy conversion efficiency and dynamic response capability.

[0165] In some embodiments, the input terminal of the flywheel control module 34 is connected to the control module 6, and the output terminal of the flywheel control module 34 is connected to the gate drive terminal of the bidirectional three-level Buck-Boost converter 33. It is used to generate a corresponding gate drive signal according to the power distribution command, so as to control the conduction and cutoff of the power semiconductor device and realize the precise adjustment of the charging and discharging power of the flywheel energy storage module 3.

[0166] In some embodiments, the flywheel status sensor 35 is disposed on the flywheel body 31 and the motor / generator 32 to collect the second operating status information of the flywheel energy storage module 3 in real time, such as flywheel speed information, vibration parameters, winding temperature information, bearing temperature information, current information, etc., and outputs the collected second operating status information to the control module 6 for subsequent power distribution decisions.

[0167] In some embodiments, the grid-connected converter module 4 is used to realize bidirectional conversion between AC power and DC power. The grid-connected converter module 4 includes: an LCL filter circuit 41, a three-level NPC grid-connected inverter 42, and an inverter control module 43.

[0168] In some embodiments, the LCL filter circuit 41 is disposed between the grid and load module 1 and the three-level NPC grid-connected inverter 42, with its AC side connected to the AC side of the grid and load module 1 and its DC side connected to the AC terminal of the three-level NPC grid-connected inverter 42.

[0169] In some embodiments, the LCL filter circuit 41 is disposed between the grid and load module 1 and the three-level NPC grid-connected inverter 42. Its AC terminal is connected to the AC terminal of the grid and load module 1, and its DC terminal is connected to the AC terminal of the three-level NPC grid-connected inverter 42. It is used to filter out high-frequency switching harmonics generated during grid connection, so as to reduce the harmonic content of grid-connected current, improve grid-connected power quality, and reduce power disturbance to the grid and load module 1.

[0170] In some embodiments, the DC terminal of the three-level NPC grid-connected inverter 42 is connected to the DC bus 51 to achieve bidirectional conversion between DC and AC power. Specifically, during charging, the AC power from the grid side is rectified into DC power and delivered to the DC bus 51; during discharging, the DC power from the DC bus 51 is inverted into AC power and fed back to the grid and load module 1. Simultaneously, the three-level NPC topology reduces voltage stress and switching losses in switching devices, improving the reliability and efficiency of medium- and high-voltage grid-connected operation.

[0171] In some embodiments, the input terminal of the inverter control module 43 is connected to the control module 6, and the output terminal is connected to the gate drive terminal of the three-level NPC grid-connected inverter 42. It is used to generate corresponding gate drive signals according to the power distribution instructions output by the control module 6, and to control the conduction and cutoff of each power semiconductor device in the three-level NPC grid-connected inverter 42, thereby realizing precise regulation of grid-connected power, voltage and current.

[0172] In some embodiments, the DC bus module 5 includes a DC bus 51 and a DC bus support capacitor group 52, which are used to connect the flywheel energy storage module 3, the electrochemical energy storage module 2 and the grid-connected converter module 4, and serve as a common energy exchange node between the modules.

[0173] In some embodiments, the DC bus 51 is simultaneously connected to the DC terminal of the three-level NPC grid-connected inverter 42, the DC terminal of the bidirectional three-level Buck-Boost converter 33 in the flywheel energy storage module 3, and the DC terminal of the bidirectional LLC resonant converter 23 in the electrochemical energy storage module 2, thereby realizing bidirectional power flow and energy exchange between the grid side and the flywheel energy storage module 3 and the electrochemical energy storage module 2.

[0174] In some embodiments, the DC bus support capacitor bank 52 can be connected in parallel across the two ends of the DC bus 51 to support and stabilize the DC bus voltage, thereby absorbing transient power fluctuations generated by each converter during power switching, reducing DC bus voltage ripple and voltage spikes, and thus improving the overall stability and anti-disturbance capability of the energy storage system 100.

[0175] In some embodiments, the control module 6 includes: an AI main control chip 61, a multi-source data acquisition module 62, a multi-timescale state prediction module 63, a multi-objective reinforcement learning optimization module 64, and a global synchronization control and protection module 65.

[0176] In some embodiments, the AI ​​main control chip 61 can be an NVIDIA Jetson Orin NX, which provides high computing power support to perform multi-source data fusion processing, multi-timescale state prediction, and multi-objective reinforcement learning computing tasks, thereby enabling real-time intelligent decision-making for the complex energy storage system 100.

[0177] In some embodiments, the multi-source data acquisition module 62 is used to acquire the first operating status information, the second operating status information, and the third operating status information output by the power grid and load module 1, the electrochemical energy storage module 2, and the flywheel energy storage module 3, as well as to receive environmental information and electricity market operation information.

[0178] In some embodiments, the multi-timescale state prediction module 63 is used to perform multi-timescale state prediction based on first operating state information, second operating state information, third operating state information, environmental information and electricity market operating information, to obtain future operating state information of the energy storage system 100 at different time scales, thereby obtaining multi-timescale state prediction results.

[0179] In some embodiments, the multi-objective reinforcement learning optimization module 64 includes a proximal policy optimization model, which includes an Actor network and a Critic network. The Actor network is used to output the power allocation coefficient between the flywheel energy storage module 3 and the electrochemical energy storage module 2 based on the state vector. The Critic network is used to evaluate the value of the current state and calculate the policy optimization direction based on the reward function. The reward function is used to comprehensively characterize the power tracking error, the operating efficiency of the energy storage system, the life loss of the energy storage module, and the economic indicators of the system.

[0180] In some embodiments, the global synchronization control and protection module 65 generates corresponding gate drive signals according to power allocation instructions, and synchronously outputs the gate drive signals to the electrochemical energy storage control module 24, the flywheel control module 34, and the inverter control module 43 according to a unified time base, so as to achieve microsecond-level synchronous control of each converter. Simultaneously, when an abnormal operating state is detected in the grid and load module 1, the electrochemical energy storage module 2, or the flywheel energy storage module 3, a protection control strategy is triggered to quickly adjust or disconnect the protection of the energy storage system's operating state.

[0181] In some embodiments, the control module 6 is connected to the grid and load module 1, the electrochemical energy storage module 2, the flywheel energy storage module 3, the grid-connected converter module 4, and the DC bus module 5, respectively, to implement the control method of the energy storage system described in the above embodiments.

[0182] According to an embodiment of the present invention, the energy storage system 100 firstly establishes a grid and load module 1, an electrochemical energy storage module 2, and a flywheel energy storage module 3, thereby providing the system with the basic conditions for unified perception and response to power fluctuations on the grid side and the internal energy state of the energy storage modules. Further, by setting up a grid-connected converter module 4, bidirectional conversion between AC and DC power is achieved, providing an energy conversion channel for energy exchange and grid-connected regulation between the flywheel energy storage module 3 and the electrochemical energy storage module 2. Simultaneously, by setting up a DC bus 51, the flywheel energy storage module 3, the electrochemical energy storage module 2, and the grid-connected converter module 4 are electrically connected, allowing each energy storage module and grid-connected device to share the same energy exchange node, thereby improving the basic consistency of power transfer and coordinated regulation among the modules. Through the coordinated connection of the control module 6 with the grid and load module 1, each energy storage module, the grid-connected converter module 4, and the DC bus 51, the control module 6 can acquire the operating status of each module and perform unified control of each execution unit, thus providing the hardware execution and coordination basis for the implementation of the control method of the energy storage system described in the above embodiment. Then, by adopting the control method of the energy storage system described in the above embodiment, the energy storage system 100's ability to perceive future changes in operating status is improved by utilizing a multi-timescale state prediction mechanism. Based on the collaborative analysis of real-time operating status information and future operating status information, the rationality of power distribution between the flywheel energy storage module 3 and the electrochemical energy storage module 2 is improved, thereby enhancing the collaborative adjustment and response capabilities of the energy storage system 100 during grid-connected operation.

[0183] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example.

[0184] Although embodiments of the invention have been shown and described, those skilled in the art will understand 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 claims and their equivalents.

Claims

1. A control method of an energy storage system, characterized by, The energy storage system includes a flywheel energy storage module and an electrochemical energy storage module. Both the flywheel energy storage module and the electrochemical energy storage module are connected to the power grid. The control method includes: Receive the automatic power generation control frequency regulation command of the power grid; The system acquires the first operating status information of the power grid, the second operating status information of the flywheel energy storage module, the third operating status information of the electrochemical energy storage module, environmental information, and electricity market operation information. Based on the first operating status information, the second operating status information, the third operating status information, the environmental information, and the electricity market operating information, multi-time-scale state prediction is performed to obtain the future operating status information of the energy storage system at different time scales. Based on the first operating status information, the second operating status information, the third operating status information, the environmental information, the electricity market operating information, and the future operating status information, power allocation instructions for the flywheel energy storage module and the electrochemical energy storage module are obtained; Based on the power distribution command, the flywheel energy storage module and the electrochemical energy storage module are controlled to perform power regulation.

2. The control method of an energy storage system according to claim 1, wherein The multi-timescale state prediction includes: Based on the first operating status information, predict the power grid frequency and power fluctuations at the first time scale; Based on the first operating status information, the environmental information, and the electricity market operating information, load changes and new energy output are predicted at a second time scale; Based on the second and third operating status information, a lifespan decay trend prediction is performed at a third time scale. Wherein, the first time scale is smaller than the second time scale, and the second time scale is smaller than the third time scale.

3. The control method of an energy storage system according to claim 1, wherein Based on the first operating status information, the second operating status information, the third operating status information, the environmental information, the electricity market operating information, and the future operating status information, power allocation instructions between the flywheel energy storage module and the electrochemical energy storage module are obtained, including: The first operating status information, the second operating status information, the third operating status information, the environmental information, the electricity market operating information, and the future operating status information are input into the near-end strategy optimization model; Obtain the output data of the near-end strategy optimization model; Based on the output data, power allocation instructions are obtained between the flywheel energy storage module and the electrochemical energy storage module.

4. The control method of an energy storage system according to claim 3, wherein Based on the output data, power allocation instructions are obtained between the flywheel energy storage module and the electrochemical energy storage module, including: The output data is corrected based on the constraints. When the corrected output data satisfies the constraint conditions, the corrected output data serves as the power distribution command between the flywheel energy storage module and the electrochemical energy storage module. The constraints include the rotational speed constraint of the flywheel energy storage module, the charge / discharge rate constraint of the electrochemical energy storage module, and the state of charge constraint of the electrochemical energy storage module.

5. The control method for the energy storage system according to claim 4, characterized in that, The power allocation instructions include: allocating high-frequency power components to the flywheel energy storage module for power regulation, and allocating medium- and low-frequency power components to the electrochemical energy storage module for power regulation.

6. The control method of an energy storage system according to claim 1, wherein Based on the power allocation command, control the flywheel energy storage module and the electrochemical energy storage module to perform power regulation, including: Based on the power allocation command, gate drive signals corresponding to the converter of the flywheel energy storage module, the converter of the electrochemical energy storage module, and the grid-connected inverter are generated synchronously. Based on the gate drive signal, the converter of the flywheel energy storage module, the converter of the electrochemical energy storage module, and the grid-connected inverter are synchronously controlled to achieve power conversion.

7. The control method of an energy storage system according to claim 6, wherein The control method further includes: The first operating status information, the second operating status information, the third operating status information, the environmental information, and the electricity market operating information are compared and analyzed with historical operating status information. Based on the comparative analysis results, anomalies are predicted for the power grid operation status, the flywheel energy storage module operation status, and the electrochemical energy storage module operation status.

8. The control method of an energy storage system according to claim 3, wherein, The control method further includes: The power tracking error, energy storage module operating status change information, and energy storage system abnormal information corresponding to the power allocation command are obtained as execution result feedback information. Based on the feedback information of the execution results, the multi-timescale state prediction, the near-end strategy optimization model, and the synchronization control strategy between each energy storage module and the power grid are updated online.

9. The control method of an energy storage system according to claim 1, wherein, The control method further includes: Data preprocessing is performed on the first operating status information, the second operating status information, the third operating status information, the environmental information, and the electricity market operating information.

10. An energy storage system characterized by, The energy storage system includes: Grid and load modules, electrochemical energy storage modules, and flywheel energy storage modules; Grid-connected converter modules are used to achieve bidirectional conversion between AC and DC power. A DC bus module is provided to connect the flywheel energy storage module, the electrochemical energy storage module, and the grid-connected converter module, and to serve as a common energy exchange node between the modules. A control module is connected to the power grid and load module, the electrochemical energy storage module, the flywheel energy storage module, the grid-connected converter module, and the DC bus module, respectively, and is used to implement the control method of the energy storage system according to any one of claims 1-9.

Images