Hybrid converter topological structure for supporting flywheel energy storage transient overcurrent, construction method, system, equipment and medium

Abstract

The invention relates to the technical field of power electronic converters and flywheel energy storage, and discloses a hybrid converter topological structure for supporting flywheel energy storage transient overcurrent, a construction method, a system, equipment and a medium, the topological structure comprises a main loop and an auxiliary branch, the main loop adopts a three-phase modularized multi-level topology, and the auxiliary branch adopts a three-phase modularized multi-level topology. And a half-bridge sub-module and a bridge arm inductor are configured, so that low-loss operation in a full-rotating-speed range is realized. And the design of the auxiliary redundant branch is used for realizing accurate switching of fault working conditions, so that transient impact current can be effectively shared to protect a main device, and additional loss during normal operation can be avoided. The control strategy established based on the main loop and the auxiliary branch can trigger a pre-alarm mechanism through the preset threshold value, and then redundant branch preheating is started in advance and the sampling frequency is improved, so that the problem of traditional diagnosis lag is effectively solved, and device loss and power grid disturbance caused by frequent switching are reduced.

Description

Technical Field

[0001] This invention relates to the field of power electronic converters and flywheel energy storage technology, and in particular to a hybrid converter topology, construction method, system, equipment and medium for supporting transient overcurrent of flywheel energy storage. Background Technology

[0002] In existing technologies, flywheel energy storage technology is increasingly used in scenarios such as grid frequency regulation, grid renewable energy consumption, and user-side backup power supply in the grid. Moreover, flywheel energy storage technology has a fast response speed and a long service life. Therefore, flywheel energy storage systems are gradually becoming an indispensable device in the grid. Among them, the converter of the flywheel energy storage system is the most important link to realize the interaction between the flywheel energy storage system and the grid.

[0003] During the interaction between the flywheel energy storage system and the power grid, the system needs to achieve low-loss steady-state operation across a wide flywheel speed range. This ensures a constant power output and stable voltage / current interaction with the grid. Furthermore, the flywheel energy storage system must provide transient overcurrent protection in fault scenarios such as deep voltage dips or sudden load changes to prevent damage to the power modules and ensure system uptime.

[0004] Furthermore, existing flywheel energy storage converter technology still has many problems. Currently, due to insufficient topology adaptability, flywheel energy storage converter technology cannot simultaneously achieve low-loss steady-state operation and transient overcurrent protection. Therefore, while ensuring constant power output over a wide speed range of the flywheel, it is impossible to reasonably and without introducing additional losses to cope with overcurrent impacts caused by fault scenarios such as deep voltage drops in the grid or sudden load changes.

[0005] Furthermore, the control strategy of flywheel energy storage converter technology has poor nonlinear adaptability, severely lacking intelligent control strategies that can adapt to wide speed fluctuations of the flywheel. Consequently, it cannot effectively solve the problem of existing fixed-parameter regulators dealing with nonlinear characteristics. Additionally, the fault diagnosis mechanism in flywheel energy storage converter technology is lagging and has low reliability. It lacks fault diagnosis logic that can accurately provide early warnings or maintain a low false positive / false negative rate. This leads to frequent activation of auxiliary branches, which not only increases device losses but may also cause secondary disturbances to the power grid. Summary of the Invention

[0006] In view of the aforementioned existing problems, the present invention is proposed.

[0007] Therefore, this invention provides a hybrid converter topology, construction method, system, equipment, and medium that supports transient overcurrent in flywheel energy storage, which can solve problems such as insufficient adaptability of existing flywheel energy storage converter topologies, poor nonlinear adaptability of control strategies, and lagging fault diagnosis mechanisms.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a hybrid converter topology that supports transient overcurrent in flywheel energy storage, comprising: Includes the main circuit and auxiliary redundant branches; The main circuit is composed of a three-phase modular multilevel converter. Each phase includes an upper bridge arm and a lower bridge arm. Each bridge arm is composed of multiple half-bridge sub-modules connected in series with the bridge arm inductor. The half-bridge sub-module includes two insulated gate bipolar transistors, freewheeling diodes connected in antiparallel to the transistors, and a DC capacitor for energy buffering and voltage support. The auxiliary redundant branch is connected in parallel with the three-phase AC output terminal of the main circuit. Each phase auxiliary branch includes, in sequence, a light-controlled thyristor with strong anti-electromagnetic interference capability, a current-limiting resistor, and an insulated gate bipolar transistor power module with high current capacity. The main circuit is configured to support low-loss steady-state operation across the entire speed range of the flywheel energy storage system and maintain stable power interaction with the grid. The auxiliary redundant branch is configured to be controlled to be activated under transient fault conditions such as deep voltage drops or load changes in the grid, to share the inrush current and prevent damage to the main circuit power devices, and to remain off during normal system operation to avoid introducing additional losses.

[0009] As a preferred embodiment of the hybrid converter topology for supporting flywheel energy storage transient overcurrent as described in this invention, wherein the DC capacitor in the half-bridge submodule is selected as a high-frequency, low-loss type capacitor. The number of half-bridge sub-modules in each bridge arm is determined based on the DC side voltage level of the system and the rated withstand voltage of a single capacitor, ensuring that the operating voltage of the sub-module capacitor does not exceed its rated value. The multi-level output characteristics formed by connecting multiple sub-modules in series improve the quality of AC side voltage waveform and adjustment accuracy.

[0010] As a preferred embodiment of the hybrid converter topology supporting transient overcurrent of flywheel energy storage described in this invention, the inductance of the bridge arm inductor is matched and selected according to the power level, allowable circulating current level and dynamic response performance requirements of the flywheel energy storage system, so that it can suppress interphase circulating current while providing buffering capability for power fluctuations caused by changes in flywheel speed.

[0011] As a preferred embodiment of the hybrid converter topology for supporting flywheel energy storage transient overcurrent as described in this invention, wherein: the current-limiting resistor in the auxiliary redundant branch is a metal oxide film resistor, and its resistance value is based on the maximum voltage difference between the DC side and AC side of the main circuit under the most severe grid fault conditions. The short-time overcurrent withstand capability of the power module is combined with the engineering design, and redundancy margin is introduced to compensate for the parameter drift of the resistor under high temperature operating conditions.

[0012] This invention also provides a method for constructing a hybrid converter topology that supports transient overcurrent in flywheel energy storage, comprising: A three-phase modular multilevel converter is constructed, with each phase consisting of an upper bridge arm and a lower bridge arm. Each bridge arm contains multiple half-bridge sub-modules and one bridge arm inductor. The number of sub-modules is determined according to the system DC voltage level and the rated withstand voltage of the sub-module capacitors. Low-loss DC capacitors are selected, and the bridge arm inductors are configured according to the power level and dynamic performance requirements to enable the main circuit to support low-loss steady-state operation of the flywheel across the entire speed range and stable power interaction with the grid. Auxiliary branches are connected in parallel at the three-phase output terminals of the main circuit. Each auxiliary branch consists of a light-controlled thyristor, a current-limiting resistor, and a high-current insulated-gate bipolar transistor power module connected in series. The light-controlled thyristor is triggered by an optical fiber to improve anti-interference capability, and the current-limiting resistor is used to suppress the current surge when the branch is connected. The main circuit adopts a multi-level control architecture that combines outer loop voltage / power control and inner loop current control, and introduces a parameter adaptive mechanism to dynamically adjust the current regulator parameters according to flywheel speed deviation, speed change rate and current tracking error. The auxiliary branch remains off during normal operation, and its power module internal freewheeling diode forms a natural bypass path. The system synchronously monitors the branch leakage current and device junction temperature for health status assessment. Based on real-time monitoring of bridge arm current, grid voltage and flywheel speed change rate, the operating trend is analyzed by combining time series prediction model; When the conditions for overcurrent warning, deep voltage drop in the grid, or sudden change in speed are met, the auxiliary branch is activated, and the power output is maintained through the coordinated distribution of current in the main and auxiliary circuits. After the fault ends, the system is restored based on the continuous stable state of multiple parameters, and then smoothly switches back to the independent operation mode of the main circuit.

[0013] As a preferred embodiment of the construction method of the hybrid converter topology supporting flywheel energy storage transient overcurrent described in this invention, the parameter adaptive mechanism adopts a deep reinforcement learning model. The model takes the deviation between the actual speed and the reference speed of the flywheel, the speed change rate, and the tracking error between the actual value and the reference value of the inner loop current as input states, and outputs the proportional gain correction amount and the resonant gain correction amount of the current regulator. The deep reinforcement learning model is trained offline and then embedded in the main controller. During online operation, it updates the input state and outputs parameter correction instructions at fixed time intervals, enabling the current regulator to adapt to the dynamic response requirements under conditions of flywheel speed step change, gradual operation, or steady-state fluctuation.

[0014] As a preferred embodiment of the method for constructing a hybrid converter topology that supports transient overcurrent of flywheel energy storage as described in this invention, the timing prediction model adopts a long short-term memory neural network, the input of which is the flywheel speed, DC side voltage and output current data in multiple consecutive sampling periods, and the output is the predicted value of flywheel output power at a specified future time. The system weights and fuses the predicted power value with the power command issued by the power grid based on the current speed change rate range to generate an active current reference value, and adjusts the switching sequence of the main circuit submodules in advance accordingly. The long short-term memory neural network is also used for fault prediction. Its input is continuous time-series data of submodule capacitor voltage, insulated gate bipolar transistor junction temperature and bridge arm current, and its output is the probability of the system failing in future operating cycles. When the probability exceeds a preset threshold, the system triggers a warning, starts redundant branch preheating, and increases the sampling frequency of key parameters.

[0015] Secondly, the present invention provides a system for constructing a hybrid converter topology that supports transient overcurrent in flywheel energy storage, comprising: The model building module is used to build a three-phase modular multilevel converter. Each phase consists of an upper bridge arm and a lower bridge arm. Each bridge arm contains multiple half-bridge sub-modules and a bridge arm inductor. Auxiliary branches are connected in parallel at the three-phase output terminals of the main circuit. Each phase auxiliary branch is composed of a light-controlled thyristor, a current-limiting resistor and a high-current insulated gate bipolar transistor power module connected in series. The main circuit configuration module enables the main circuit to adopt a multi-level control architecture that combines outer loop voltage / power control with inner loop current control, and introduces a parameter adaptive mechanism to dynamically adjust the current regulator parameters based on flywheel speed deviation, speed change rate and current tracking error. The auxiliary branch configuration module is used to ensure that the auxiliary branch remains off during normal operation. Its power module internal freewheeling diode forms a natural bypass path. The system synchronously monitors the branch leakage current and device junction temperature for health status assessment. The analysis module is used for real-time monitoring of the change rate of bridge arm current, grid voltage and flywheel speed, and to analyze the operating trend in combination with the time series prediction model.

[0016] Thirdly, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0017] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0018] Compared with existing technologies, the beneficial effects of this invention are that it proposes a hybrid converter topology and construction method to support flywheel energy storage transient overcurrent. The topology includes a main circuit and auxiliary branches. The main circuit adopts a three-phase modular multi-level topology and is configured with half-bridge sub-modules and bridge arm inductors to achieve low-loss operation across the entire speed range. The auxiliary redundant branch design is used to achieve precise switching under fault conditions, which can effectively share the transient inrush current to protect the main components and avoid additional losses during normal operation. The control strategy established based on the main circuit and auxiliary branches can trigger a pre-alarm mechanism through a preset threshold, thereby starting the redundant branches in advance for preheating and increasing the sampling frequency, thus effectively solving the problem of traditional diagnostic lag and reducing device losses and grid disturbances caused by frequent switching. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 The present invention provides a method flowchart for constructing a hybrid converter topology that supports transient overcurrent in flywheel energy storage, as an embodiment of the present invention.

[0021] Figure 2 This is an internal structural diagram of an electronic device that supports a hybrid converter topology for flywheel energy storage transient overcurrent and a construction method, as provided in an embodiment of the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0023] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a hybrid converter topology and construction method for supporting flywheel energy storage transient overcurrent, including: This invention provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to realize the hybrid converter topology and construction method that supports flywheel energy storage transient overcurrent in conjunction with multiple embodiments. Figure 1 A flowchart illustrating a hybrid converter topology supporting transient overcurrent of flywheel energy storage and its construction method is shown. The hybrid converter topology supporting transient overcurrent of flywheel energy storage includes a main circuit and an auxiliary redundant branch. In an optional implementation, the main circuit is composed of a three-phase modular multilevel converter, i.e., a three-phase MMC structure. Each phase includes an upper bridge arm and a lower bridge arm. Each bridge arm is composed of multiple half-bridge sub-modules connected in series with the bridge arm inductor. The half-bridge sub-module includes two insulated gate bipolar transistors, i.e., high-voltage IGBTs, freewheeling diodes connected in antiparallel to the transistors, i.e., IGBT built-in freewheeling diodes, and a DC capacitor for energy buffering and voltage support, i.e., a high-frequency low-loss electrolytic capacitor or a film capacitor. In an optional implementation, the auxiliary redundant branch is connected in parallel with the three-phase AC output terminal of the main circuit. Each phase of the auxiliary branch includes, in sequence, a light-controlled thyristor with strong anti-electromagnetic interference capability, namely a light-controlled SCR, a trigger signal transmitted through optical fiber, a current-limiting resistor, namely a metal oxide film resistor, and a high-current capacity insulated gate bipolar transistor power module, namely a high-current IGBT module, which supports high-frequency PWM regulation and integrates overcurrent protection function. The main circuit is configured to support low-loss steady-state operation across the entire speed range of the flywheel energy storage system and maintain stable power interaction with the grid. The auxiliary redundant branch is configured to be controlled to be activated during transient fault conditions such as deep voltage drops in the grid or sudden load changes, to share the inrush current and prevent damage to the power devices in the main circuit, and to remain off during normal system operation to avoid introducing additional losses.

[0024] In this embodiment of the invention, the DC capacitor in the half-bridge submodule is selected as a high-frequency, low-loss capacitor, namely a high-frequency, low-loss electrolytic capacitor or a film capacitor. The number of half-bridge sub-modules in each bridge arm is determined based on the DC side voltage level of the system and the rated withstand voltage capability of a single capacitor. That is, it is calculated based on the DC side voltage of the main MMC and the rated voltage of the sub-module capacitor and rounded up to ensure that the working voltage of the sub-module capacitor does not exceed its rated value. The multi-level output characteristics formed by connecting multiple sub-modules in series improve the quality of AC side voltage waveform and adjustment accuracy.

[0025] In one alternative implementation, the inductance of the bridge arm inductor is selected to match the power level, allowable circulating current level, and dynamic response performance requirements of the flywheel energy storage system. The selection principle is to use 5–8mH for low-power systems and 8–10mH for high-power systems, so that it can suppress interphase circulating current while providing buffering capacity for power fluctuations caused by changes in flywheel speed.

[0026] In this embodiment of the invention, the current-limiting resistor in the auxiliary redundant branch is a metal oxide film resistor, and its resistance value is based on the maximum voltage difference between the DC side and AC side of the main circuit under the most severe power grid fault conditions, that is, 1.7–1.8 times the difference between the rated voltage of the power grid and the voltage drop value of the power grid. The power module's short-time overcurrent withstand capability (i.e., 2.5 times the rated current based on the IGBT's short-time withstand capability) is incorporated into the engineering design, while redundancy margin (i.e., 10% redundancy is reserved) is introduced to compensate for the parameter drift of the resistor under high-temperature operating conditions.

[0027] It should be noted that the hybrid converter topology and construction method supporting flywheel energy storage transient overcurrent specifically include: S101 is used to build a three-phase modular multilevel converter. Each phase consists of an upper bridge arm and a lower bridge arm. Each bridge arm contains multiple half-bridge sub-modules and a bridge arm inductor. The number of sub-modules is determined according to the system DC voltage level and the rated withstand voltage of the sub-module capacitors. Low-loss DC capacitors are selected, and the bridge arm inductors are configured according to the power level and dynamic performance requirements to enable the main circuit to support low-loss steady-state operation of the flywheel across the entire speed range and stable power interaction with the grid. S102, auxiliary branches are connected in parallel at the three-phase output terminals of the main circuit. Each auxiliary branch is composed of a light-controlled thyristor, a current-limiting resistor and a high-current insulated gate bipolar transistor power module connected in series. The light-controlled thyristor is triggered by an optical fiber to improve anti-interference capability, and the current-limiting resistor is used to suppress the current surge when the branch is connected. S103 adopts a multi-level control architecture that combines outer loop voltage / power control with inner loop current control in the main circuit. That is, the outer loop voltage loop and power loop, the inner loop current loop, and the nearest level approximation method (NLM) three-level control are introduced. A parameter adaptive mechanism is introduced (that is, the PR regulator parameters are optimized based on deep Q network (DQN) to dynamically adjust the current regulator parameters according to the flywheel speed deviation, speed change rate, and current tracking error. S104, the auxiliary branch is kept off during normal operation. Its power module internal freewheeling diode forms a natural bypass path. The system synchronously monitors the branch leakage current (i.e., collected through series sampling resistor) and device junction temperature. That is, the freewheeling diode junction temperature is monitored through the module's built-in temperature sensor for health status assessment. S105 analyzes the operating trend based on real-time monitoring of bridge arm current, grid voltage and flywheel speed change rate (i.e., speed is collected and calculated by photoelectric encoder) and combined with time series prediction model (i.e. LSTM model). S106 When the judgment conditions of overcurrent warning (i.e., the bridge arm current exceeds 1.2 times the rated current for two consecutive sampling cycles), grid voltage deep drop (i.e., the AC side voltage is lower than the set threshold) or speed change (i.e., the speed change rate exceeds the threshold set based on mechanical inertia) are met, the auxiliary branch is triggered to start and the power output is maintained through the coordinated distribution of the main and auxiliary circuit currents. S107 After the fault ends, the system recovers based on the continuous stable state of multiple parameters (i.e., the recovery condition is met for 10 consecutive 10kHz sampling cycles), and smoothly switches back to the independent operation mode of the main circuit.

[0028] In this embodiment of the invention, the parameter adaptive mechanism adopts a deep reinforcement learning model (i.e., deep Q network DQN). This model takes the deviation between the actual speed and the reference speed of the flywheel, the speed change rate, and the tracking error between the actual value and the reference value of the inner loop current as input states, and outputs the proportional gain correction amount and the resonant gain correction amount of the current regulator, i.e. the Kp and Kr correction amounts of the PR regulator. The deep reinforcement learning model is trained offline and then embedded in the main controller (i.e., trained and deployed using speed fluctuation data of 1000–6000 rpm). During online operation, the input state is updated at fixed time intervals and parameter correction instructions are output, enabling the current regulator to adapt to the dynamic response requirements under flywheel speed step changes, gradual operation, or steady-state fluctuation conditions.

[0029] In this embodiment of the invention, the time-series prediction model adopts a long short-term memory neural network (i.e., a two-layer LSTM structure, with 64 neurons in the first layer and 32 neurons in the second layer). Its input is the flywheel speed, DC side voltage and output current data within multiple consecutive sampling periods, and its output is the predicted value of the flywheel output power at a specified future time. The system weights and fuses the predicted power value with the power command issued by the grid based on the current speed change rate range to generate an active current reference value, and adjusts the switching sequence of the main circuit submodules in advance accordingly. In an optional implementation, the long short-term memory neural network is also used for fault prediction. Its input is continuous time-series data of submodule capacitor voltage, insulated gate bipolar transistor junction temperature and bridge arm current, and its output is the probability of the system failing in future operating cycles. When the probability exceeds a preset threshold, the system triggers a warning, starts redundant branch preheating, and increases the sampling frequency of key parameters.

[0030] Example 2: Based on the above examples, a specific implementation of a hybrid converter topology and construction method supporting transient overcurrent in flywheel energy storage can be designed as follows: Construct a flywheel energy storage hybrid converter main circuit MMC topology to ensure low-loss steady-state operation of the flywheel energy storage system.

[0031] Furthermore, a three-phase MMC structure is adopted, consisting of three phase units A, B, and C, with each phase divided into an upper bridge arm and a lower bridge arm. This provides a stable grid interface for constant power output across a wide speed range of the flywheel, enabling stable interaction of grid-side voltage and current.

[0032] Furthermore, single-arm bridges are... It consists of a half-bridge submodule and a bridge arm inductor.

[0033] Furthermore, the half-bridge sub-module includes two high-voltage IGBTs, each with a freewheeling diode to avoid losses and potential failure points introduced by additional components. One DC capacitor is a high-frequency, low-loss electrolytic capacitor or film capacitor to reduce charging and discharging losses. The sub-module's output voltage can be switched to 0 or the capacitor's rated voltage via the IGBT's on / off state. The half-bridge structure reduces the conduction and switching losses of the switching devices, achieving low-loss steady-state operation. Further calculations of the number of submodules are performed using the following formula: .in, The DC-side voltage of the main MMC is set according to the power level and voltage adaptation requirements of the flywheel energy storage system. This refers to the rated voltage of the submodule capacitors. To avoid insulation breakdown or shortened lifespan due to capacitor overvoltage, the calculation uses "rounding up" to ensure that the voltage of each submodule capacitor does not exceed the rated value. At the same time, the level superposition effect of multiple submodules connected in series improves the regulation accuracy and harmonic characteristics of the main MMC output voltage.

[0034] In one optional implementation, the bridge arm inductance is calculated. The bridge arm inductance is selected based on the system's maximum allowable circulating current and the power rating of the flywheel energy storage system to balance circulating current suppression with dynamic response speed. For low-power systems (1-3MW), a 5-8mH inductor can be used, while for high-power systems (3-10MW), an 8-10mH inductor is required to buffer power fluctuations caused by changes in flywheel speed, providing sufficient settling time for current closed-loop control and ensuring constant power output across a wide flywheel speed range.

[0035] Furthermore, an auxiliary redundant branch topology is constructed for the flywheel energy storage hybrid converter to ensure the replenishment of transient overcurrent capacity of the flywheel energy storage system.

[0036] Furthermore, the auxiliary branch is connected in parallel with the three-phase output of the main MMC, forming a parallel topology of main circuit + auxiliary branch, enabling independent regulation of the current in each phase. Each branch contains three core components: a high-speed switch, a power module, and a current-limiting resistor.

[0037] Furthermore, the high-speed switch employs a light-controlled thyristor (SCR) with strong electromagnetic interference resistance, transmitting trigger signals via optical fiber, unaffected by the strong electromagnetic environment of the flywheel system. Considering that most current flywheel energy storage systems are medium-power systems, the power module uses a high-current IGBT module with modular packaging, offering high control flexibility, supporting high-frequency PWM regulation, and integrating overcurrent protection. The capacity design incorporates calculations of the flywheel system's fault current peak, considering a voltage drop fault rate of 1.5. Module 2 The module is adaptable to extreme faults. The current-limiting resistor is a metal oxide film resistor, connected in series between the SCR and the power module to suppress the inrush current when the branch is switched on.

[0038] Furthermore, the auxiliary branch current-limiting resistor Resistance value calculation. A 10% redundancy is reserved to address resistance drift caused by temperature rise, based on Ohm's law and engineering redundancy design. The derivation is as follows This enables accurate selection and reliable operation of resistors.

[0039] Furthermore, the maximum inrush current is calculated. The maximum inrush current is taken as 2.5 based on the short-time withstand capability of the power module's IGBT. It can balance shock suppression and device redundancy.

[0040] Furthermore, calculate the voltage difference. . Press "Main MMC DC Side Voltage During Fault" With output voltage The calculation of the "maximum difference" takes into account the worst fault conditions of deep voltage drop in the power grid, ensuring that the resistance value can cover all fault scenarios. The value should be 1.7-1.8 times the rated voltage of the power grid. The value is taken as the voltage drop value of the power grid under the worst operating conditions.

[0041] Furthermore, a control strategy for the flywheel energy storage hybrid converter was developed. Under normal operating conditions, a deep Q-network (DQN) was used to optimize the main MMC control parameters, and LSTM was used to realize dynamic speed-current mapping. This adapts to the nonlinear characteristics of the flywheel's wide speed fluctuation and solves the nonlinear adjustment problem under the wide range of flywheel speed fluctuations.

[0042] In one optional implementation, the optically controlled thyristor (SCR) of the auxiliary branch is maintained in a zero-bias off state. The drive unit monitors the voltage across the SCR in real time. If the voltage drops abnormally, it is determined to be an SCR leakage current fault, and an alarm is immediately triggered. The power module IGBTs are bypassed through internally integrated freewheeling diodes. Specifically, both the upper and lower bridge arm IGBTs in the module are in the off state, and the current flows naturally through the freewheeling diodes to form a path, eliminating the need for additional mechanical bypass switches. At the same time, the junction temperature of the freewheeling diodes is monitored by a temperature sensor built into the module to avoid overheating of the devices due to long-term bypass. The static loss is the product of the forward voltage drop of the freewheeling diode and the branch leakage current, which is controlled to be ≤0.001% of the rated power. The leakage current is collected by a series sampling resistor to achieve online leakage current monitoring. If the leakage current is >1mA for three consecutive sampling cycles, it is determined to be a freewheeling diode fault, triggering the switching of redundant branches to achieve "no-downtime pre-maintenance" and improve the reliability of steady-state operation.

[0043] Furthermore, the main MMC control adopts a three-level control system consisting of outer loop voltage control, inner loop current control, and nearest level approximation (NLM) method. The DQN algorithm is used to adaptively optimize the PI / PR regulator parameters, thereby improving the regulation accuracy and dynamic response under nonlinear operating conditions.

[0044] Furthermore, the outer loop control is divided into a voltage loop and a power loop, with an inner loop current reference value generated by a PI regulator. The voltage loop uses the rated DC-side voltage of the main MMC. With the goal in mind, compare with reality Output reactive current reference Suppress DC voltage fluctuations; the power loop uses the power command of the flywheel energy storage system. To achieve the target, compare with the actual output power. Output active current reference To ensure power tracking error.

[0045] In one optional implementation, the DQN algorithm is used to design and optimize the parameters of the inner-loop PR regulator to address the problem that a fixed-parameter PR regulator struggles to quickly track the current reference value when the flywheel speed changes abruptly. The proportional gain of the PR regulator is adjusted in real time using DQN. ) and resonance coefficient ( The input state variables of DQN include three timing characteristics: flywheel speed deviation. ( This refers to the actual rotational speed. (Reference speed), speed change rate Inner loop current tracking error ( This represents the actual active / reactive current. (This is the reference current output from the outer loop), and all state quantities are normalized and mapped to the [0,1] interval. The action quantity output by DQN is the parameter correction quantity for the PR regulator. ( (change) ( (Change amount), ensuring parameter adjustment remains within a safe range. With the goal of minimizing current tracking error and shortening settling time, the reward function is designed as follows: .in (Weight 0.7) (Weight 0.3) is the adjustment coefficient. The time for current tracking to steady state is defined (rewards are deducted if it exceeds 5ms). Rewards are used to guide the DQN to learn the optimal parameter adjustment strategy. The DQN model is trained using flywheel speed fluctuation data ranging from 1000 to 6000 rpm. After training iterations, the model is embedded in the main controller. During online operation, the state variables are updated every 50μs, and parameter corrections are output in real time, enabling the PR regulator to adapt to changes in speed.

[0046] Furthermore, NLM switching and dead-time optimization calculates the total number of submodules that need to be deployed based on the three-phase modulated wave voltage. Prioritize switching on submodules with small voltage deviations to reduce capacitor voltage fluctuations; dead-time compensation avoids voltage distortion caused by dead time, eliminating the need for additional LC filtering.

[0047] Furthermore, the flywheel energy storage hybrid converter control uses timing prediction to anticipate the impact of flywheel speed changes on output power, providing an active current reference value for the main MMC power loop. Improve the speed-power response speed.

[0048] Furthermore, to ensure that the LSTM model can adapt to all flywheel operating scenarios, it collects data on multiple types of operating conditions, including step speed conditions to adapt to the nonlinear characteristics when the speed changes abruptly; gradual speed change conditions to cover stable speed change scenarios; steady-state fluctuation conditions to improve the prediction accuracy of the model during steady-state operation; and fault transition conditions to ensure that the model can still output prediction results stably under fault scenarios.

[0049] In one optional implementation, outlier handling of the collected raw data employs the "3σ criterion," first calculating each parameter (n, ..., ...)... , mean Compared with the standard deviation σ, those exceeding [ -3σ, Outliers within the range of +3σ are then replaced with linear interpolation to complete the removed data. For an anomaly at the k-th sampling point, the linear interpolation result of the (k-1)th and (k+1)th points is used to replace it, ensuring the continuity of the time-series data. A time-series window is constructed. The input feature (X) takes three parameters from 10 consecutive sampling periods, with each input sample being a 10×3 matrix. The output label (Y) is the actual flywheel output power value at the 20th sampling period after the input window ends. As the prediction target of LSTM ( (labels), among which ( (This refers to the power factor collected in real-time from the grid side). The input features are mapped to the [0,1] interval using "Min-Max normalization," as shown in the formula. ,in , These are the minimum and maximum values ​​of each parameter in the entire dataset, respectively. The normalization parameters need to be stored in the controller for inverse normalization during model inference.

[0050] Furthermore, an LSTM model is constructed. The input layer is defined as "(None, 10, 3)", where "None" supports arbitrary batch size, facilitating flexible adjustment during training and inference; "10" corresponds to the time steps of the past 10 sampling periods; and "3" represents the number of features (n, ..., 3) at each time step. , The input layer requires no activation function; it directly feeds the normalized temporal features into the hidden layer. The hidden layer contains two LSTM layers: the first LSTM layer has 64 neurons, a number validated through grid search to achieve an optimal balance between prediction accuracy and computational cost; to ensure the next LSTM layer receives complete temporal features, "return_sequences" is set to True; the activation function is tanh. The second LSTM layer has 32 neurons to reduce computational cost and meet real-time inference requirements; since the next layer is fully connected, it only needs to output the feature vector of the last time step, so "return_sequences" is set to False. The output layer is a fully connected layer with only one neuron, used to output the power prediction value for the next 2ms. The activation function chosen is ReLU, which avoids negative output power while maintaining high computational efficiency; a power upper limit constraint is also added. , (Rated power of the flywheel) to prevent predicted values ​​from exceeding the system's load-bearing capacity and ensure safety in engineering applications.

[0051] In one optional implementation, the loss function used during the model compilation stage is the mean squared error (MSE), as shown in the formula: This function is sensitive to power prediction errors and can effectively guide model learning to reduce prediction bias. The Adam optimizer is selected, with an initial learning rate of 0.001 and a "learning rate decay" strategy employed. This ensures rapid convergence in the early stages of training while allowing for fine-tuning of parameters later to improve accuracy. The mean absolute percentage error (MAPE) is used as the evaluation metric, and the formula is... This metric directly reflects the percentage of prediction error, making it easy to intuitively evaluate model performance.

[0052] Furthermore, the trained LSTM model undergoes lightweighting, employing post-training quantization. Based on the test set data, quantization parameters are statistically analyzed to convert the 32-bit floating-point model to a 16-bit fixed-point model, avoiding precision loss that could impact performance. Model pruning is then performed, removing redundant neurons and connections. For the second LSTM layer, 28 neurons are retained. For fully connected layers, connections with absolute weight values ​​less than 0.01 are pruned to ensure timely participation of prediction results. The calculation.

[0053] Furthermore, the coordinated control of the LSTM and power loop is achieved through coordinated control with the main MMC power loop, thereby predicting... With the power grid Reasonable integration, generation Furthermore, the submodule switching strategy was optimized to ultimately shorten the power tracking lag time. To balance "predictive foresight" and "command accuracy," a dynamic weighted fusion strategy was adopted to generate the final power reference value. Define the "speed change rate threshold". ,when (Steady-state / gradually changing operating conditions) Prioritize trusting the information issued by the power grid. To ensure the stability of steady-state operation; when (Under sudden operating conditions) Prioritize LSTM's early predictions to fully leverage its advantages in time series forecasting. Meanwhile... Must meet To avoid exceeding the flywheel power adjustment range and ensure system safety.

[0054] Furthermore, the power loop is based on Generate in advance The calculation formula is: ,in This is the real-time voltage on the grid side. The target power factor is calculated, and the result is then subjected to amplitude limiting. , (Rated current), to prevent overcurrent risk. Pre-generated... Input NLM modulation module, optimize submodule switching strategy: based on The modulated wave voltage that the main MMC needs to output is calculated in reverse. Then calculate the total number of sub-modules that need to be invested. (N is the total number of submodules in each bridge arm); due to By generating the list of sub-modules in advance, the NLM module can determine the sub-module switching list one sampling period in advance, giving priority to sub-modules with small voltage deviations, reducing capacitor voltage fluctuations, and ensuring that the main MMC can quickly respond to power change requirements, fully meeting the power grid's requirements for power tracking accuracy and response speed.

[0055] Furthermore, we construct LSTM-assisted fault diagnosis logic to reduce false positives and false negatives, enabling fault diagnosis to shift from "passive response" to "proactive prevention".

[0056] In one optional implementation, to prevent the main MMC submodule IGBT from experiencing a sudden temperature rise in junction temperature due to short-term overcurrent, which could lead to device damage, an overcurrent warning threshold is set. Based on the IGBT's short-term withstand curve, the overcurrent warning threshold is determined to be 1.2. This approach allows sufficient adjustment time for subsequent fault response while avoiding frequent warnings due to excessively low thresholds. Hall effect current sensors, installed in each arm of the main MMC, collect real-time data on the upper / lower arm currents of phases A, B, and C, maintaining synchronization with the sampling frequency of the subsequent LSTM module to reduce data synchronization errors. The maximum value of the three-phase arm current is taken and 1.2... If the threshold is exceeded for two consecutive sampling periods, a fault warning is triggered.

[0057] Furthermore, the abnormal operating condition determination targets two typical scenarios affecting system stability—grid voltage dips and flywheel speed fluctuations—and determines the determination threshold by combining grid operation specifications and flywheel mechanical characteristics. The determination of the two abnormal operating conditions and the current over-limit determination form an "OR logic," meaning that when either abnormal condition is met, the system enters a "fault candidate state."

[0058] Furthermore, when the grid voltage drops, the monitoring point is selected as the AC output voltage of the main MMC. When the voltage value is greater than or equal to the threshold, it is judged as an "abnormal deep voltage drop". This threshold is based on the grid fault level classification standard. This leads to a serious imbalance between the flywheel output power and the grid demand, which can easily cause overcurrent in the main circuit.

[0059] Furthermore, the flywheel speed fluctuation is monitored using the speed change rate as the parameter. The rotational speed is calculated from the data collected by the photoelectric encoder installed on the flywheel motor shaft. When the value is greater than or equal to the threshold, it is judged as "abnormal speed change" - this value is determined based on the mechanical inertial characteristics of the flywheel rotor: speed changes above the threshold will cause the rotor kinetic energy to be rapidly converted into electrical energy, causing the DC side voltage of the main MMC to rise sharply, which in turn will cause the bridge arm current to exceed the limit.

[0060] Furthermore, an LSTM fault prediction module is constructed to identify potential faults in advance by analyzing the temporal variation trends of key parameters, thereby reducing the false negative rate.

[0061] Furthermore, select the submodule capacitor voltage. IGBT junction temperature With bridge arm inductor current Three types of core parameters. Submodule capacitor voltage. This directly reflects the health of the capacitor. Capacitor aging will cause its capacitance value to decrease and its charging and discharging speed to slow down, which manifests as... Increased fluctuation amplitude is detected through the voltage sampling circuit built into the submodule. IGBT junction temperature. It is a core indicator reflecting the performance degradation of the device. The on-state voltage drop of an IGBT increases with increasing junction temperature. The temperature continuously exceeded the threshold, causing a sharp increase in device losses, which was monitored by the module's built-in K-type thermocouple. (Bridge arm inductor current) This indirectly reflects the insulation status of the inductor. A short circuit between inductor turns will cause the inductance value to decrease and the harmonic components of the current to increase. This can be collected by the Hall current sensor of the main MMC bridge arm without additional hardware costs.

[0062] Furthermore, a fault feature library is constructed covering two scenarios: "progressive failures" and "sudden failure precursors." Progressive failures are identified through the accumulation of time-series feature examples from long-term operational data. Sudden failure precursors are identified through data collection via fault simulation experiments. The final fault feature library contains both fault and normal samples, with the sample time span covering the full speed range of the flywheel and typical failure scenarios, ensuring the model's generalization ability.

[0063] In one optional implementation, the LSTM fault prediction model outputs the probability of future faults through deep temporal feature extraction. The input layer design uses three types of time-series data from the past 20 sampling periods as input, forming an input matrix with dimensions of "time step = 20, feature number = 3". A time window is selected to fully capture precursor features; all input data is normalized to the [0,1] interval using Min-Max normalization to eliminate the impact of dimensional differences on model training. The hidden layer uses a two-layer LSTM structure. The first layer has 48 neurons with the return_sequences parameter set to True, outputting complete temporal features to the second layer; the second layer has 24 neurons with the return_sequences parameter set to False, outputting only the feature vector of the last time step. Compared to a single-layer LSTM, this two-layer structure can uncover deeper temporal correlations. The output layer is a fully connected layer with one neuron, using the Sigmoid activation function, outputting a fault probability of 0-1. During the model training phase, the loss function is binary cross-entropy, the optimizer is Adam, and an early stopping strategy is enabled to avoid overfitting.

[0064] Furthermore, the early warning triggering logic is designed in accordance with actual engineering practices. When the failure probability of the LSTM output is ≥0.8, a "fault early warning alarm" is triggered. The system does not immediately activate the auxiliary branch, but starts the warm-up process of the redundant submodule / branch, increases the sampling frequency of key parameters, and enhances the monitoring sensitivity; if the subsequent threshold determination triggers a fault, the redundant unit can be switched directly.

[0065] Furthermore, after fault diagnosis and confirmation, the auxiliary branch is quickly put into control, and after the fault is cleared, it resumes normal operation.

[0066] Furthermore, after fault diagnosis and confirmation, the core control system, as the command initiator, receives the judgment signal from the fault detection module and triggers a hardware interrupt to quickly generate a digital command for "auxiliary branch activation". The command is sent to the auxiliary branch drive unit through multimode optical fiber. A "CRC check" mechanism is added to the command transmission to avoid incorrect branch activation caused by transmission errors, thereby further improving reliability.

[0067] Furthermore, a collaborative distribution logic of "main MMC constant current + auxiliary branch compensation" is adopted, which is implemented through dual closed-loop control. The main MMC current closed loop collects the current of each bridge arm in real time through bridge arm current sensors, and adjusts the switching frequency of the submodules via a PI controller to ensure stable output current and prevent the main MMC from triggering overcurrent protection due to current surges. The auxiliary branch current calculation calculates the total output current requirement. Inversely derived from grid voltage and flywheel power ( The current that the auxiliary branch needs to handle Current balance control involves the FPGA comparing the currents of the three-phase auxiliary branches in real time. If the current deviation of a certain phase exceeds 10%, the gate drive signal of the corresponding phase power module is finely adjusted to ensure that the imbalance of the three-phase current does not exceed the threshold, thus avoiding single-phase overload.

[0068] Furthermore, duration control is implemented based on the power module's thermal losses and junction temperature limitations to design the conduction time. The duration differences for different fault types stem from the fault energy requirements. In the grid voltage dip fault mode, the flywheel needs to maintain power output during a voltage dip; the fault duration of 15ms is typically determined by the grid recovery speed, covering grid dip scenarios. In the load surge fault mode, the overcurrent caused by load surges results in a fault duration of 10ms, reducing power module losses. Junction-case thermal resistance (°C) Conduction loss during fault Within the response time, the junction temperature rise is below the rated junction temperature of 175℃, ensuring module safety.

[0069] Furthermore, the main and auxiliary circuits work together to maintain voltage stability. The main MMC submodules are sorted and voltage-equalized using a method that samples the capacitor voltage of all submodules every 20μs; the submodules are sorted from highest to lowest voltage; and the number of submodules to be deployed in the upper bridge arm is determined accordingly. Before selection A high-voltage submodule is selected for the upper bridge arm, and a low-voltage submodule is selected for the lower bridge arm in the same way. Using this algorithm, the submodule voltage deviation can be controlled within a threshold, thereby ensuring the DC-side voltage of the main MMC remains within acceptable limits. The fluctuations meet the DC voltage stability requirements of the power grid. Voltage surges in the auxiliary branch are suppressed, and the main MMC output voltage remains stable during faults. The current-limiting resistor in series with the auxiliary branch may experience sudden changes due to current variations. Voltage drop To counteract voltage surges. Parallel connection of main and auxiliary circuits is prone to circulating current due to voltage differences; a current-limiting resistor is needed. Circulation Suppress and prevent circulating current from diverting main power, ensuring constant flywheel power output.

[0070] Furthermore, a 10kHz high-frequency sampling method is used, and recovery is only determined after 10 consecutive sampling cycles that meet the conditions. These 10 consecutive sampling cycles can filter out instantaneous fluctuations in grid voltage and flywheel speed. A current threshold of 0.1 below the main MMC overcurrent threshold is set. In the buffer zone, secondary overcurrent protection is avoided from being triggered immediately. The main power circuit is disconnected via "gate signal switching + freewheeling diode conduction," and the main MMC smoothly transitions from a fault state to normal operation via "soft-start voltage reduction," avoiding disturbances to the power grid caused by sudden current changes. The voltage reduction rate is limited by the power grid voltage fluctuation. The reference value of the main MMC's PI current controller decreases linearly. A submodule voltage equalization algorithm maintains DC-side voltage stability, preventing capacitor charging and discharging imbalances caused by current adjustments. The gate drive signal of the power module (IGCT / IGBT) quickly switches from "on voltage" to "off voltage." After the gate is turned off, the module's main circuit current freewheels through the internal freewheeling diode, and the current decays rapidly, ensuring no residual current in the branch. The branch current sensor monitors the current for three consecutive sampling cycles in real time to determine if the power module is completely bypassed, and the main power circuit is disconnected. During the reset process, the power grid voltage, flywheel speed, and main MMC current are monitored in real time. If any parameter is abnormal (e.g., the power grid voltage drops again), the reset is immediately paused and the auxiliary branch is re-triggered to ensure system fault tolerance.

[0071] Example 3, referring to Figure 2 This embodiment also provides a system for constructing a hybrid converter topology that supports transient overcurrent in flywheel energy storage, comprising: The model building module is used to build a three-phase modular multilevel converter. Each phase consists of an upper bridge arm and a lower bridge arm. Each bridge arm contains multiple half-bridge sub-modules and a bridge arm inductor. Auxiliary branches are connected in parallel at the three-phase output terminals of the main circuit. Each phase auxiliary branch is composed of a light-controlled thyristor, a current-limiting resistor and a high-current insulated gate bipolar transistor power module connected in series. The main circuit configuration module enables the main circuit to adopt a multi-level control architecture that combines outer loop voltage / power control with inner loop current control, and introduces a parameter adaptive mechanism to dynamically adjust the current regulator parameters based on flywheel speed deviation, speed change rate and current tracking error. The auxiliary branch configuration module is used to ensure that the auxiliary branch remains off during normal operation. Its power module internal freewheeling diode forms a natural bypass path. The system synchronously monitors the branch leakage current and device junction temperature for health status assessment. The analysis module is used for real-time monitoring of the change rate of bridge arm current, grid voltage and flywheel speed, and to analyze the operating trend in combination with the time series prediction model.

[0072] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0073] This embodiment also provides an electronic device, which can be a terminal, and its internal structure diagram can be as follows. Figure 2 As shown, the electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for constructing a hybrid converter topology that supports transient overcurrent in flywheel energy storage. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the device's casing, or an external keyboard, touchpad, or mouse.

[0074] This embodiment also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it performs the following steps: A three-phase modular multilevel converter is constructed, with each phase consisting of an upper bridge arm and a lower bridge arm. Each bridge arm contains multiple half-bridge sub-modules and a bridge arm inductor. Auxiliary branches are connected in parallel at the three-phase output terminals of the main circuit. Each auxiliary branch consists of a light-controlled thyristor, a current-limiting resistor, and a high-current insulated-gate bipolar transistor power module connected in series. The main circuit adopts a multi-level control architecture that combines outer loop voltage / power control and inner loop current control, and introduces a parameter adaptive mechanism to dynamically adjust the current regulator parameters according to flywheel speed deviation, speed change rate and current tracking error. The auxiliary branch remains off during normal operation, and its power module internal freewheeling diode forms a natural bypass path. The system synchronously monitors the branch leakage current and device junction temperature for health status assessment. Based on real-time monitoring of bridge arm current, grid voltage and flywheel speed change rate, the operating trend is analyzed by combining time series prediction model; When the conditions for overcurrent warning, deep voltage drop in the grid, or sudden change in speed are met, the auxiliary branch is activated, and the power output is maintained through the coordinated distribution of current in the main and auxiliary circuits. After the fault ends, the system is restored based on the continuous stable state of multiple parameters, and then smoothly switches back to the independent operation mode of the main circuit.

[0075] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0076] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0077] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A hybrid converter topology to support flywheel energy storage transient overcurrent, characterized by, The main circuit and the auxiliary redundant branch are included. The main circuit is composed of a three-phase modular multilevel converter, each phase including an upper bridge arm and a lower bridge arm, each bridge arm being connected in series with a plurality of half-bridge sub-modules and a bridge arm inductor; the half-bridge sub-module includes two insulated gate bipolar transistors, two freewheeling diodes respectively connected in antiparallel to the transistors, and a DC capacitor for energy buffering and voltage support. The auxiliary redundant branch is connected in parallel with the three-phase AC output end of the main circuit, and each phase auxiliary branch includes, in sequence, a light-controlled thyristor with strong anti-electromagnetic interference capability, a current-limiting resistor, and a high-current insulated gate bipolar transistor power module. The main circuit is configured to support low-loss steady-state operation within the full-speed operating range of the flywheel energy storage system and maintain stable power interaction with the power grid. The auxiliary redundant branch is configured to be controlled to be put into operation under transient fault conditions such as deep voltage sag of the power grid or sudden change of the load, to share the impact current and prevent damage to the power devices of the main circuit, and to remain in an off state during normal operation of the system to avoid introducing additional losses.

2. A hybrid converter topology for supporting flywheel energy storage transient overcurrent as claimed in claim 1, wherein, The DC capacitor in the half-bridge sub-module is a high-frequency low-loss capacitor. The number of half-bridge sub-modules in each bridge arm is determined according to the DC voltage level of the system and the rated voltage withstand capability of a single capacitor, to ensure that the operating voltage of the sub-module capacitor does not exceed its rated value. The multi-level output characteristics formed by the series connection of multiple sub-modules improve the quality and regulation accuracy of the AC side voltage waveform.

3. A hybrid converter topology for supporting flywheel energy storage transient overcurrent as claimed in claim 2, wherein, The inductance of the bridge arm inductor is matched and selected according to the power level, the allowed circulating current level, and the dynamic response performance requirements of the flywheel energy storage system, so that it can suppress inter-phase circulating current while providing buffering capability for power fluctuations caused by changes in flywheel speed.

4. A hybrid converter topology for supporting flywheel energy storage transient overcurrent as claimed in claim 3, wherein, The current-limiting resistor in the auxiliary redundant branch is a metal oxide film resistor, and its resistance value is determined according to the maximum voltage difference between the DC side and the AC side of the main circuit under the most severe power grid fault condition. The short-time overcurrent tolerance capability of the power module is considered in the engineering design, and a redundancy margin is introduced to compensate for the parameter drift of the resistor under high-temperature operating conditions.

5. A method for constructing a hybrid converter topology for supporting transient overcurrent of a flywheel energy storage, according to any one of claims 1-4, characterized in that, a three-phase modular multilevel converter is built, each phase being composed of an upper bridge arm and a lower bridge arm, each bridge arm including a plurality of half-bridge sub-modules and a bridge arm inductor; an auxiliary branch is connected in parallel at the three-phase output end of the main circuit, each phase auxiliary branch being composed of a light-controlled thyristor, a current-limiting resistor, and a high-current insulated gate bipolar transistor power module connected in series; the main circuit adopts a multi-level control architecture combining outer loop voltage / power control and inner loop current control, and introduces a parameter self-adaptive mechanism to dynamically adjust the current regulator parameters according to the flywheel speed deviation, the speed change rate, and the current tracking error; the auxiliary branch remains off during normal operation, and the internal freewheeling diodes of the power module form a natural bypass path, and the system synchronously monitors the branch leakage current and the device junction temperature for health status evaluation; based on real-time monitoring of the bridge arm current, the grid voltage, and the flywheel speed change rate, the operation trend is analyzed in combination with a time series prediction model; When the overcurrent early warning, power grid voltage deep drop or sudden change of speed of the determination conditions are met, the auxiliary branch is triggered to put into operation, and the power output is maintained through the main auxiliary loop current cooperative distribution; After the fault is over, the system recovery is judged according to the multi-parameter continuous stable state, and the main loop independent operation mode is switched back smoothly.

6. The method of constructing a hybrid converter topology for supporting flywheel energy storage transient overcurrent according to claim 5, wherein, The parameter adaptive mechanism adopts a deep reinforcement learning model, which takes the deviation of the actual speed of the flywheel from the reference speed, the speed change rate and the tracking error between the actual value and the reference value of the inner loop current as the input state, and outputs the proportional gain correction and the resonance gain correction of the current regulator; The deep reinforcement learning model is solidified in the main controller after offline training, and the input state is updated and the parameter correction instruction is output at fixed time intervals during online operation, so that the current regulator can adapt to the dynamic response demand under the flywheel speed step change, gradual change operation or steady state fluctuation working condition.

7. The method of constructing a hybrid converter topology for supporting flywheel energy storage transient overcurrent according to claim 6, wherein, The time sequence prediction model adopts a long short-term memory neural network, which takes the flywheel speed, DC side voltage and output current data in a plurality of continuous sampling periods as input, and outputs the flywheel output power prediction value at a specified time in the future; The system weights and fuses the power prediction value and the power instruction issued by the power grid according to the interval where the current speed change rate is located, generates an active current reference value, and adjusts the switching sequence of the main loop submodule in advance according to the active current reference value; The long short-term memory neural network is also used for fault prediction, which takes the continuous time sequence data of the submodule capacitor voltage, the insulated gate bipolar transistor junction temperature and the bridge arm current as input, and outputs the probability of system failure in the future operation period; When the probability exceeds a preset threshold, the system triggers a pre-warning, starts the preheating of the redundant branch and increases the sampling frequency of the key parameters.

8. A system for constructing a hybrid converter topology supporting flywheel energy storage transient overcurrent, applying the method of any one of claims 5 to 7, characterized in that, It comprises: A model building module is used to build a three-phase modular multilevel converter, each phase being composed of an upper bridge arm and a lower bridge arm, each bridge arm comprising a plurality of half-bridge type submodules and a bridge arm inductor; an auxiliary branch is connected in parallel at the three-phase output end of the main loop, and each auxiliary branch is composed of a light-controlled thyristor, a current-limiting resistor and a high-current insulated gate bipolar transistor power module connected in series; A main loop configuration module is used to make the main loop adopt a multi-level control architecture combining outer loop voltage / power control and inner loop current control, and introduce a parameter adaptive mechanism to dynamically adjust the current regulator parameters according to the flywheel speed deviation, speed change rate and current tracking error; An auxiliary branch configuration module is used to ensure that the auxiliary branch remains off during normal operation, the internal freewheeling diode of the power module forms a natural bypass path, and the system synchronously monitors the branch leakage current and the device junction temperature for health status evaluation; An analysis module is used to analyze the running trend based on real-time monitoring of the bridge arm current, the power grid voltage and the flywheel speed change rate in combination with the time sequence prediction model. 9.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to realize the steps of the construction method of the hybrid converter topology supporting the transient overcurrent of the flywheel energy storage as claimed in any one of claims 5-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the construction method of the hybrid converter topology supporting the transient overcurrent of the flywheel energy storage as claimed in any one of claims 5-7.

Images