Control method and device for LCL type grid-connected inverter
By optimizing the control strategy of the LCL grid-connected inverter using a deep reinforcement learning model, the problems of control accuracy and stability were solved, and effective tracking and regulation of high-frequency current harmonics were achieved, thereby improving the control performance of the system.
Patent Information
- Application Number
- CN202411059076.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-02
AI Technical Summary
Existing control methods for LCL-type grid-connected inverters suffer from poor control accuracy and stability, making it difficult to effectively track and regulate high-frequency current harmonics.
A deep reinforcement learning model is adopted to analyze the state of the LCL grid-connected inverter through a deep Q-network, and to make decisions on the three-phase switching action. The control strategy is optimized by iterative training using empirical elements in the experience playback pool.
It improves the control accuracy and stability of LCL-type grid-connected inverters, suppresses grid-side current resonance, and reduces total harmonic distortion.
Smart Images

Figure CN119010616A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of power electronic control, and in particular to a method and device for controlling an LCL type grid-connected inverter. Background Art
[0002] In the context of "dual carbon", the development of inverter technology has attracted much attention. As the core of energy conversion and control in the power generation system, its performance will directly affect the quality of the power system. In order to suppress the output harmonics of the inverter, it is necessary to install a filter between the inverter and the power system. LCL filters are widely used to suppress the problem of poor power quality of grid-connected inverters due to their small size and high-frequency attenuation capabilities.
[0003] At present, LCL grid-connected inverters usually adopt a dual closed-loop linear control strategy, but this method has the problem of limited current control bandwidth, so it is difficult to achieve tracking and regulation of higher frequency current harmonics. To this end, relevant technicians have proposed nonlinear control strategies with faster dynamic response speed and stronger tracking performance, such as repetitive control and sliding mode control.
[0004] Among them, the repetitive control is to add the error to the output signal cycle by cycle. Although it has good steady-state performance, the embedded delay link makes the output current lag behind the reference value by one power frequency cycle. In addition, it is also proposed to add a proportional integral link in the repetitive control to effectively improve the dynamic performance of the repetitive control, but the complex compensation function makes it difficult to implement in engineering applications; and the sliding mode control is based on the error switching function. Although it has a large current control bandwidth, it causes additional output current jitter due to problems such as time delay and space lag, thereby affecting the control accuracy and stability of the system.
[0005] Therefore, there is an urgent need for a LCL-type grid-connected inverter control method that can meet the requirements of power system stability and robustness. Summary of the invention
[0006] The embodiments of the present application provide a method and device for controlling an LCL type grid-connected inverter, so as to at least solve the technical problem that when the related nonlinear control technology is used to control the LCL type grid-connected inverter, the control accuracy and control stability are poor.
[0007] According to one aspect of an embodiment of the present application, a method for controlling an LCL-type grid-connected inverter is provided, comprising: obtaining a first state of a first LCL-type grid-connected inverter at a first moment, wherein the first state at least includes: a two-phase current and a two-phase voltage between the first LCL-type grid-connected inverter and a power grid, and a two-phase current and a two-phase voltage inside the first LCL-type grid-connected inverter; analyzing the first state using a deep reinforcement learning model, and deciding a target three-phase switching action of the first LCL-type grid-connected inverter, wherein the deep reinforcement learning model is obtained by iteratively training a deep Q network using multiple experience elements in an experience replay pool, and the experience elements include a second state of a second LCL-type grid-connected inverter at a second moment, a three-phase switching action selected according to a greedy strategy, a reward obtained when executing the three-phase switching action, and a new second state, and the reward is an error between a two-phase current between the second LCL-type grid-connected inverter and the power grid and a preset power grid reference current when the second LCL-type grid-connected inverter executes the selected three-phase switching action; and controlling the first LCL-type grid-connected inverter according to the target three-phase switching action obtained by the decision.
[0008] Optionally, obtaining a target state of the first LCL type grid-connected inverter at a first moment includes: obtaining the three-phase current and three-phase voltage between the first LCL type grid-connected inverter and the grid and the three-phase current and three-phase voltage inside the first LCL type grid-connected inverter at the first moment; using the three-phase voltage between the first LCL type grid-connected inverter and the grid as an input signal of a phase-locked loop to obtain a phase; based on the phase, using Park transformation to convert the three-phase current and three-phase voltage between the first LCL type grid-connected inverter and the grid and the three-phase current and three-phase voltage inside the first LCL type grid-connected inverter from a three-phase stationary coordinate system to a dq coordinate system that rotates synchronously with the voltage, to obtain a two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the grid and a two-phase current and two-phase voltage inside the first LCL type grid-connected inverter.
[0009] Optionally, the training process of the deep reinforcement learning model includes: obtaining a training sample set and a sample label set, wherein the training sample set includes: a plurality of training samples consisting of a second state of the second LCL type grid-connected inverter at a second moment, and the sample label set includes: sample labels of a plurality of training samples, and the sample labels are used to characterize the three-phase switching action of the second LCL type grid-connected inverter that minimizes the difference between the two-phase current between the second LCL type grid-connected inverter and the grid and a preset grid reference current; constructing an initial model based on a deep Q network; and iteratively training the initial model using the training sample set and the sample label set to obtain a deep reinforcement learning model.
[0010] Optionally, the process of determining the grid reference current includes: for each second LCL type grid-connected inverter, obtaining the active power instruction and reactive power instruction of the second LCL type grid-connected inverter at the second moment; determining the grid reference current based on the active power instruction and the reactive power instruction and the two-phase voltage between the second LCL type grid-connected inverter and the grid at the second moment.
[0011] Optionally, the initial model is iteratively trained using the training sample set and the sample label set to obtain a deep reinforcement learning model, including: defining a state space based on the training sample set, wherein each state in the state space represents a second state of the second LCL type grid-connected inverter at a second moment; defining an action space based on the sample label set, wherein each action in the action space represents a three-phase switching action of the second LCL type grid-connected inverter; initializing a deep Q network, wherein the deep Q network includes an online network and a target network; using the online network to select a second state from the state space, and according to the second state, using a greedy algorithm to select a second state with a large Q value. A three-phase switching action at a preset threshold; determining a reward and a new second state when the second LCL type grid-connected inverter performs the selected three-phase switching action; storing experience elements consisting of the second state of the second LCL type grid-connected inverter at the second moment, the three-phase switching action performed by the second LCL type grid-connected inverter, the reward when the second LCL type grid-connected inverter performs the selected three-phase switching action, and the new second state into an experience replay pool; using a target network to determine a target Q value for each experience element in the experience replay pool, and adjusting the model parameters of the initial model based on the Q value of each experience element and the target Q value to obtain a deep reinforcement learning model.
[0012] Optionally, the target network is used to determine the target Q value of each experience element in the experience replay pool, and the model parameters of the initial model are adjusted based on the Q value of each experience element and the target Q value, including: looping the following steps in each iteration until a preset termination condition is met: randomly extracting experience elements from the experience replay pool according to the experience replay mechanism, and using the target network to determine the target Q value of each extracted experience element; determining the loss function based on analyzing the Q value of each experience element and the corresponding target Q value using the online network; minimizing the loss function using the gradient descent method to update the network parameters of the online network; and copying the network parameters of the online network to the target network after at least one iterative training process.
[0013] Optionally, the first LCL type grid-connected inverter is controlled according to the target three-phase switching action obtained by the decision, including: when any phase switch within the target three-phase switching action is closed, controlling the corresponding phase switch in the first LCL type grid-connected inverter to be in a high level state; when any phase switch within the target three-phase switching action is open, controlling the corresponding phase switch in the first LCL type grid-connected inverter to be in a low level state.
[0014] According to another aspect of an embodiment of the present application, there is also provided an LCL type grid-connected inverter control device, including: an acquisition module, used to acquire a first state of a first LCL type grid-connected inverter at a first moment, wherein the first state at least includes: a two-phase current and a two-phase voltage between the first LCL type grid-connected inverter and the power grid, and a two-phase current and a two-phase voltage inside the first LCL type grid-connected inverter; a decision module, used to analyze the first state using a deep reinforcement learning model, and make a decision to obtain a target three-phase switching action of the first LCL type grid-connected inverter, wherein the deep reinforcement learning model is used The deep Q network is iteratively trained using multiple experience elements in the experience replay pool, the experience elements include the second state of the second LCL type grid-connected inverter at the second moment, the three-phase switching action selected according to the greedy strategy, the reward obtained when the three-phase switching action is executed, and the new second state, and the reward is the error between the two-phase current between the second LCL type grid-connected inverter and the grid and the preset grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action; the control module is used to control the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision.
[0015] According to another aspect of an embodiment of the present application, a non-volatile storage medium is also provided, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned LCL type grid-connected inverter control method by running the computer program.
[0016] According to another aspect of an embodiment of the present application, a computer program product is further provided, the computer program product comprising a stored computer program, wherein when the computer program is executed by a processor, the above-mentioned LCL type grid-connected inverter control method is implemented.
[0017] In an embodiment of the present application, the control system analyzes the current state of the LCL type grid-connected inverter by adopting a deep Q network algorithm to obtain the switching action of each phase of the LCL type grid-connected inverter, and controls the level state of the corresponding phase switch output according to the switching action, so that the grid-side current can accurately and quickly follow the change of the command value, and suppress the grid-side current resonance to reduce the total harmonic distortion of the grid-side current in the interval, thereby solving the technical problem of poor control accuracy and control stability when the relevant nonlinear control technology controls the LCL type grid-connected inverter. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0019] Figure 1It is a hardware structure block diagram of a computer terminal for implementing an LCL type grid-connected inverter control method according to an embodiment of the present application;
[0020] Figure 2 is a flow chart of an optional LCL type grid-connected inverter control method according to an embodiment of the present application;
[0021] Figure 3 is a structural schematic diagram of an optional LCL type grid-connected inverter control device according to an embodiment of the present application;
[0022] Figure 4 It is a schematic diagram of the structure of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION
[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.
[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0025] In addition, the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. For example, an interface is set up between this system and the relevant user or organization. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain relevant information after receiving the consent information fed back by the aforementioned user or organization.
[0026] In order to better understand the embodiments of the present application, the technical terms involved in the embodiments of the present application are explained as follows:
[0027] LCL grid-connected inverter is a device that converts electricity (Direct Current, DC) into electricity (Alternating Current, AC) and connects to the power grid. It is mainly composed of LCL filter, inverter and controller. Among them, LCL filter is used to suppress the harmonic components in the inverter output current to ensure that the output current meets the requirements of the power grid, while the inverter is responsible for converting electrical energy into electrical energy and outputting it to the load through the power grid.
[0028] The DQN (Deep Q-Network) algorithm is a deep reinforcement learning algorithm. Its main idea is to use a deep neural network to approximate the Q function, namely the Q value, which represents the expected cumulative reward of the agent taking a certain action in a given state.
[0029] The Bellman Optimality Equation is a basic concept in reinforcement learning, which is used to find the optimal solution of the action-value function under a given strategy.
[0030] Example 1
[0031] According to an embodiment of the present application, a method embodiment of an LCL type grid-connected inverter control method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0032] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing the control method of an LCL type grid-connected inverter is shown. Figure 1 As shown, the computer terminal 10 (or mobile device 10) may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.
[0033] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).
[0034] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the LCL type grid-connected inverter control method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the LCL type grid-connected inverter control method of the above-mentioned application is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.
[0035] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.
[0036] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).
[0037] In the above operating environment, Figure 2is a flow chart of an optional LCL type grid-connected inverter control method according to an embodiment of the present application, such as Figure 2 As shown, the method at least includes steps S202-S206, wherein:
[0038] Step S202: acquiring a first state of a first LCL grid-connected inverter at a first moment.
[0039] The first state at least includes: two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the power grid, and two-phase current and two-phase voltage inside the first LCL type grid-connected inverter.
[0040] Step S204: Analyze the first state using a deep reinforcement learning model to determine a target three-phase switching action of the first LCL-type grid-connected inverter.
[0041] In the technical solution provided in step S204, since the existing control technology has problems such as poor control accuracy, poor stability, and high energy consumption for the control of LCL type grid-connected inverters, the embodiment of the present application proposes to use reinforcement learning in a continuous "trial and error" process to optimize the embedded neural network so that its output action develops in the direction of the optimal strategy. That is, the first state is analyzed using a pre-trained deep reinforcement learning model to make a decision to obtain the target three-phase switching action of the first LCL type grid-connected inverter. Among them, the above-mentioned deep reinforcement learning model is obtained by iteratively training the deep Q network using multiple experience elements in the experience replay pool, and the experience element includes the second state of the second LCL type grid-connected inverter at the second moment, the three-phase switching action selected according to the greedy strategy, the reward obtained when the three-phase switching action is executed, and the new second state (i.e., the state at the next moment of the second moment), and the reward is the error between the two-phase current between the second LCL type grid-connected inverter and the grid and the preset grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action.
[0042] Step S206: Control the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision.
[0043] In the technical solution provided in step S206, the control system controls the level state of each phase switch output in the first LCL type grid-connected inverter according to the target three-phase switching action output by the deep reinforcement learning model decision.
[0044] The above method of this embodiment is further introduced below.
[0045] As an optional implementation, in the technical solution provided in the above step S202, the control system may obtain the first state by the following method, including:
[0046] Step S2021, obtaining the three-phase current and three-phase voltage between the first LCL type grid-connected inverter and the grid and the three-phase current and three-phase voltage inside the first LCL type grid-connected inverter at a first moment;
[0047] Step S2022, using the three-phase voltage between the first LCL type grid-connected inverter and the grid as an input signal of a phase-locked loop to obtain a phase;
[0048] Step S2023, based on the phase, use Park transformation to convert the three-phase current and three-phase voltage between the first LCL type grid-connected inverter and the grid and the three-phase current and three-phase voltage inside the first LCL type grid-connected inverter from the three-phase stationary coordinate system to the dq coordinate system that rotates synchronously with the voltage, so as to obtain the two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the grid and the two-phase current and two-phase voltage inside the first LCL type grid-connected inverter.
[0049] In this embodiment, when the three-phase grid-connected inverter (i.e., LCL type grid-connected inverter) using the LCL filter is connected to the grid, the control system can use the current transformer and the voltage transformer to collect the three-phase current (i.e., the grid-side current, denoted as i) between the first LCL type grid-connected inverter and the grid. 1,A ,i 1,A ,i 1,A ) and three-phase voltage (i.e. grid-side voltage, denoted as e A ,e B ,e C ), and at the same time, collect the three-phase current inside the first LCL type grid-connected inverter (that is, the current between the first inductor and capacitor, denoted as i 2,A ,i 2,B ,i 2,C ) and three-phase voltage (i.e., filter capacitor voltage, denoted as u c,A ,u c,B ,u c,C ).
[0050] Considering that the LCL filter is a third-order system, it will generate harmonics at certain frequencies (such as high frequencies), resulting in harmonic amplification. At the same time, the information on the grid side may change dynamically at any time, such as voltage imbalance or frequency fluctuation, making the mathematical processing relatively complicated. Therefore, in order to simplify the control algorithm, the control system can also convert the acquired three-phase current and three-phase voltage into a two-phase orthogonal coordinate system of the orthogonal axis, so as to facilitate the decoupling of current and voltage and make it easier to optimize the dynamic response.
[0051] Specifically, the control system can convert the collected three-phase current and three-phase voltage on the grid side and the three-phase current and three-phase voltage on the first LCL type grid-connected inverter side through the following steps, including:
[0052] Step S11: The three-phase voltage e between the first LCL type grid-connected inverter and the grid at the first moment is A ,e B ,e C As the input of the phase-locked loop (PLL), the voltage phase ωt of the phase-locked loop output is obtained.
[0053] Among them, the phase-locked loop is generally composed of a phase detector (PD), a loop filter (LF), and a voltage-controlled oscillator (VCO). As a phase feedback system, it can periodically compare the phase difference between the input signal and the VCO output signal to obtain a phase error, and after the phase error is filtered by the loop filter, its output is used as a control signal input to the VCO to eliminate the phase difference between the input signal and the VCO output signal. Therefore, the above step S1 can be implemented by the following steps:
[0054] Step S111: The three-phase voltage e A ,e B ,e C Any phase in (such as e A ) as the input signal of the phase-locked loop, and the phase of this phase voltage is used as the initial reference phase of the phase-locked loop;
[0055] Step S112, calculating the difference between the two-phase voltages (such as calculating e A ,e B The difference between AB =e A -e B ), where the difference between the two-phase voltages can reflect the phase relationship between the two-phase voltages in time. Therefore, the phase change can be detected by the difference between the two-phase voltages. For example, when the differential signal passes through the zero point, it indicates that the phase of the two-phase voltage has changed by 90°;
[0056] Step S113, using the difference between the two-phase voltages calculated above as the input of the phase detector, comparing it with the reference phase of the VCO, and outputting a corresponding error signal;
[0057] Step S114, considering that the output signal of the phase detector usually includes high-frequency noise, the high-frequency noise can be removed by a loop filter to extract the low-frequency component of the phase error;
[0058] Step S115, using the output of the loop filter as a control signal of the VCO to adjust the frequency and phase of the VCO so that the output signal of the VCO is synchronized with the phase of the input signal of the phase-locked loop;
[0059] Step S116, when the phase-locked loop is locked, it means that the phase of the input signal of the VCO is consistent with the phase of the input signal of the phase-locked loop, thereby achieving locking of the phase of the input signal.
[0060] Step S12, using Park Transformation to transform the three-phase current i between the first LCL type grid-connected inverter and the grid 1,A ,i 1,B ,i 1,C and three-phase voltage e A ,e B ,e C And the three-phase current i inside the first LCL type grid-connected inverter 2,A ,i 2,B ,i 2,C And the three-phase voltage u c,A ,u c,B ,u c,C The two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the grid and the two-phase current and two-phase voltage inside the first LCL type grid-connected inverter are obtained by converting the three-phase stationary coordinate system into the voltage synchronous reference dq coordinate system.
[0061] Among them, the above Park transform can be written as:
[0062]
[0063] As an optional implementation, in the technical solution provided in the above step S204, the training process of the deep reinforcement learning model includes the following steps:
[0064] Step S21, obtaining a training sample set and a sample label set.
[0065] The training sample set includes: a plurality of training samples consisting of a second state of the second LCL type grid-connected inverter at a second moment, wherein the second state can be a two-phase current (denoted as i) between the second LCL type grid-connected inverter and the grid. 1,d ,i 1,q ) and the two-phase voltage (denoted as e d ,e q ), the two-phase current inside the first LCL type grid-connected inverter (denoted as i 2,d ,i 2,q ) and the two-phase voltage (denoted as u c,d ,u c,q ), phase (denoted as ωt), the two-phase current between the second LCL type grid-connected inverter and the grid and the preset grid reference current (denoted as i 1,d * ,i 1,q * ) of the error current (denoted as Δi 1,d,Δi 1,q ), therefore, the second state can be expressed as:
[0066] s t ={Δi 1,d ,Δi 1,q ,dΔi 1,d ,dΔi 1,q ,∫Δi 1,d ,∫Δi 1,q ,ωt,e d ,e q ,i 2,d ,i 2,q ,u c,d ,u c,q}
[0067] Where Δi 1,d =i 1,d -i 1,d * , Δi 1,q =i 1,q -i 1,q * , dΔi 1,d ,dΔi 1,q Denotes Δi 1,d ,Δi 1,q The differential value of ∫Δi 1,d ,∫Δi 1,q Denotes Δi 1,d ,Δi 1,q The integral value of .
[0068] It should be noted that the above grid reference current can be understood as: based on the instantaneous power theory, combined with the active power instruction P of the second LCL type grid-connected inverter at the second moment * and reactive power command Q * And the two-phase voltage e on the grid side at this moment d ,e q Determined grid direct axis reference current i 1,d * and the grid quadrature axis reference current i 1,q * , therefore, the active power command P * and reactive power command Q * With the grid direct axis reference current i 1,d * and the grid quadrature axis reference current i 1,q * The relationship can be written as:
[0069]
[0070] Where P * ,Q* and e d ,e q are all known quantities, so by solving the above equations, the direct axis reference current i can be obtained 1,d * and quadrature axis reference current i 1,q * .
[0071] In addition, the above-mentioned sample label set includes: sample labels of multiple training samples, and the sample labels are obtained by analyzing the training samples using existing methods, such as repetitive control, sliding mode control, etc., and the target three-phase switching action of the second LCL type grid-connected inverter is obtained by decision, and the three-phase switching action can minimize the difference between the two-phase current between the second LCL type grid-connected inverter and the grid at the corresponding moment and the preset grid reference current.
[0072] Step S22, constructing an initial model based on the deep Q network.
[0073] Among them, the deep Q network improves the traditional Q learning algorithm, stores the agent's experience (state, action, reward, new state) in a data set, and then randomly extracts samples from it for learning. This process is also called experience replay. At the same time, the deep Q network uses two neural networks with the same structure but different parameters: the online network and the target network. The online network is used to select the agent's action according to the current state. It will continue to learn and update to improve the estimation of Q value as much as possible. Therefore, at each time step, the agent will select an action based on the Q value provided by the online network, and update the network parameters based on this action and environmental feedback; and the target network can update the TD target (Temporal-DifferenceTarget) based on the calculated Q value, that is, the maximum Q value of the new state. The network parameters will not be updated at each time step, but the parameters of the online network will be copied after a certain time interval. This can make the learning process more stable and avoid oscillation caused by the rapid update of the online network.
[0074] Step S23, iteratively train the initial model using the training sample set and the sample label set to obtain a deep reinforcement learning model.
[0075] Optionally, the above step S23 may be implemented by the following method, including:
[0076] Step 1: Define the state space S based on the training sample set, where each state s in the state space t ∈S represents the second state of the second LCL type grid-connected inverter at the second moment;
[0077] Step 2: Define the action space A based on the sample label set, where each action a in the action space t ∈A represents the three-phase switching action of the second LCL type grid-connected inverter;
[0078] Step 3: Define the reward function R t .
[0079] Reward function R t It is used to evaluate the second LCL type grid-connected inverter (i.e., intelligent agent) in the second state s t Next, by performing action a t Transition to the new second state s t+1 The immediate reward obtained, therefore, the expression of the reward function can be written as:
[0080] R t =αP(a t |s t )+βΩ(s t ,a t ,s t+1 )
[0081] Where P represents the prediction accuracy, Ω(s t ,a t ,s t+1 ) represents the uncertainty of the second state s transferring to the new second state s′ through action a, α, β represent weight coefficients;
[0082] Step 4: Define the state value function.
[0083] V π (s) = E π [G t |s t =s,a t =a]
[0084] Among them, V π represents the state value function based on the policy π in the Markov decision process MDP, which is defined as the expected return from following the policy in state s, G t =R t +γR t+1 +γ 2 R t+2 +…, which represents the weighted sum of all rewards (also called return) from time t to the terminal state in the Markov decision process MDP.
[0085] At the same time, in the Markov decision process MDP, due to the existence of actions, an additional action function value function needs to be defined:
[0086] Qπ (s,a)=E π [G t |s t =s,a t =a]
[0087] Among them, Q π It represents the expected reward of performing the three-phase switching action a for the current second state s when the Markov decision process MDP follows the strategy π.
[0088] The relationship between the state value function and the action value function is: when using strategy π, the value of the second state s is equal to the sum of the probabilities of taking all three-phase switch actions a based on strategy π in the second state s multiplied by the corresponding values, which can be expressed as:
[0089] V π (s)=∑ a∈A π(a|s)Q π (s,a)
[0090] When using strategy π, the value of taking the three-phase switch action a in the second state s is equal to the immediate reward plus the product of the state transition probabilities of all possible next second states s′ after attenuation and the corresponding values, which can be expressed as:
[0091] Q π (s,a)=R(s,a)+γ∑ s′∈S P(s′|a,s)V π (s′)
[0092] Therefore, the Bellman expectation equation can be derived from the relationship between the above state value function and action value function:
[0093]
[0094] Where, π(a|s)=P(s t =s,a t =a), and ∑ a∈A π(a|s)=1, π(a|s) is defined as the probability of taking the three-phase switching action a in the second state s, and γ represents the discount factor, which is a number between 0 and 1 and is used to reduce the value of future rewards. P(s′|a,s) represents the state transition probability, which represents the probability that the second LCL type grid-connected inverter takes the three-phase switching action a in the second state s and then transfers to the new state s′.
[0095] Step 5: According to the experience playback mechanism, randomly extract experience elements (s t ,a t ,r t ,s t+1), apply the deep Q network algorithm to train the initial model, and iteratively update the network parameters to optimize and minimize the error between the actual reward and the expected reward, and maximize the long-term reward by continuously optimizing the strategy π(a|s).
[0096] Specifically, the fifth step may include:
[0097] Initialize a deep Q network, wherein the deep Q network includes an online network and a target network, wherein the online network is used to calculate the Q values corresponding to all three-phase switch actions corresponding to the state, and the target network is used to stabilize the target Q value in the learning process.
[0098] First, use the online network to select a second state s from the state space S t :
[0099] s t ={Δi 1,d ,Δi 1,q ,dΔi 1,d ,dΔi 1,q ,∫Δi 1,d ,∫Δi 1,q ,ωt,e d ,e q ,i 2,d ,i 2,q ,u c,d ,u c,q}
[0100] According to the state s, a greedy algorithm is used to select the three-phase switching action with a Q value greater than a preset threshold, that is, the action that maximizes Q(s,a;θ) is selected according to π(a|s).
[0101] Among them, the above ε-greedy strategy can be expressed as follows:
[0102]
[0103] Where 1-ε probability uses the action with the largest expected reward and randomly selects an action with ε probability.
[0104] Next, determine the LCL type grid-connected inverter to perform the selected three-phase switching action a t The reward r t and the new second state s t+1 , where the reward is the error between the two-phase current on the grid side and the preset grid reference current when the LCL grid-connected inverter performs the selected three-phase switching action. Therefore, the reward r t You can write:
[0105] r t =(i 1,d -i 1,d* ) 2 +(i 1,q -i 1,q * ) 2
[0106] Then, the second state s of the power system at the second moment is t , the three-phase switch action of the second LCL type grid-connected inverter a t , Reward t , the new second state s t+1 The experience element composed of the t ,a t ,r t ,s t+1 ), if the number of experience elements in the experience replay pool does not exceed the maximum experience replay pool capacity, continue to store experience elements; if it exceeds the maximum capacity, delete the earliest stored experience element and store the current experience element. When the number of experience elements in the experience replay pool is sufficient, experience elements can be randomly sampled from the experience replay pool to train the initial model.
[0107] Finally, the target network is used to determine the target Q value of each experience element in the experience replay pool, and the model parameters of the initial model are adjusted based on the Q value of each experience element and the target Q value to obtain a deep reinforcement learning model.
[0108] Specifically, the embodiments of the present application can adjust the parameters of the initial model by the following method:
[0109] In each iteration, the following steps are executed repeatedly until the preset termination condition is met:
[0110] According to the experience replay mechanism, a batch of experience elements (s) are randomly drawn from the experience replay pool. t ,a t ,r t ,s t+1 ), and use the target network to determine the extracted experience elements (s t ,a t ,r t ,s t+1 )’s target Q value;
[0111] Determine the Q(s) of each empirical element using online network analysis t ,a t ; θ) and the error between the corresponding target Q value, and the following target loss function is obtained:
[0112] y=r t +γQ(s t+1 ,argmax*(yQ(s t ,a t;θ)) 2 +)
[0113] Use the gradient descent method to minimize the above objective loss function to update the network parameters of the online network;
[0114] After at least one iteration, the network parameters of the online network are copied to the target network.
[0115] As an optional implementation, in the technical solution provided in the above step S206, the control system can control the first LCL type grid-connected inverter according to the following rules, including: when any phase switch in the target three-phase switch action is closed, the corresponding phase switch in the first LCL type grid-connected inverter is controlled to be in a high level state; when any phase switch in the target three-phase switch action is open, the corresponding phase switch in the first LCL type grid-connected inverter is controlled to be in a low level state. For example, the control strategy of the first LCL type grid-connected inverter can be expressed as the following Table 1.
[0116] Table 1
[0117] A phase level B phase level C phase level just just just just just burden just burden just just burden burden burden just just burden just burden burden burden just burden burden burden
[0118] It can be seen that in the embodiment of the present application, the control system analyzes the current state of the LCL type grid-connected inverter by adopting a deep Q network algorithm to obtain the switching action of each phase of the LCL type grid-connected inverter, and controls the level state of the corresponding phase switch output according to the switching action, so that the grid-side current can accurately and quickly follow the change of the instruction value, and suppress the grid-side current resonance to reduce the total harmonic distortion of the grid-side current in the interval, thereby solving the technical problem of poor control accuracy and control stability when the relevant nonlinear control technology controls the LCL type grid-connected inverter.
[0119] Example 2
[0120] Based on Embodiment 1 of the present application, an embodiment of an LCL type grid-connected inverter control device is also provided, and when the device is running, the LCL type grid-connected inverter control method of the above embodiment is executed. Figure 3 is a schematic structural diagram of an optional LCL type grid-connected inverter control device according to an embodiment of the present application, such as Figure 3 As shown, the LCL type grid-connected inverter control device at least includes an acquisition module 31, a decision module 32 and a control module 33, wherein:
[0121] An acquisition module 31 is used to acquire a first state of the first LCL type grid-connected inverter at a first moment, wherein the first state at least includes: a two-phase current and a two-phase voltage between the first LCL type grid-connected inverter and the grid, and a two-phase current and a two-phase voltage inside the first LCL type grid-connected inverter;
[0122] A decision module 32 is used to analyze the first state using a deep reinforcement learning model and make a decision to obtain a target three-phase switching action of the first LCL type grid-connected inverter, wherein the deep reinforcement learning model is obtained by iteratively training a deep Q network using multiple experience elements in an experience replay pool, and the experience elements include a second state of the second LCL type grid-connected inverter at the second moment, a three-phase switching action selected according to a greedy strategy, a reward obtained when the three-phase switching action is executed, and a new second state, and the reward is an error between a two-phase current between the second LCL type grid-connected inverter and the grid and a preset grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action;
[0123] The control module 33 is used to control the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision.
[0124] It should be noted that the various modules in the above-mentioned LCL type grid-connected inverter control device can be program modules (for example, a set of program instructions to implement a certain specific function) or hardware modules. For the latter, it can be expressed in the following forms, but is not limited to this: the expression form of each of the above-mentioned modules is a processor, or the functions of each of the above-mentioned modules are implemented by a processor.
[0125] Example 3
[0126] According to an embodiment of the present application, a non-volatile storage medium is also provided, in which a program is stored, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the LCL type grid-connected inverter control method in Example 1.
[0127] Optionally, the device where the non-volatile storage medium is located implements the following steps by running the program: obtaining a first state of the first LCL type grid-connected inverter at a first moment, wherein the first state at least includes: two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the power grid, and two-phase current and two-phase voltage inside the first LCL type grid-connected inverter; using a deep reinforcement learning model to analyze the first state, and making a decision to obtain a target three-phase switching action of the first LCL type grid-connected inverter, wherein the deep reinforcement learning model is obtained by iteratively training a deep Q network using multiple experience elements in an experience replay pool, and the experience elements include a second state of the second LCL type grid-connected inverter at the second moment, a three-phase switching action selected according to a greedy strategy, a reward obtained when executing the three-phase switching action, and a new second state, and the reward is an error between the two-phase current between the second LCL type grid-connected inverter and the power grid and a preset power grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action; and controlling the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision.
[0128] According to an embodiment of the present application, a computer program product is also provided, which includes a stored computer program, wherein when the computer program is executed by a processor, the LCL type grid-connected inverter control method in Example 1 is implemented.
[0129] Optionally, the computer program executes the following steps: obtaining a first state of the first LCL type grid-connected inverter at a first moment, wherein the first state includes at least: two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the grid, and two-phase current and two-phase voltage inside the first LCL type grid-connected inverter; analyzing the first state using a deep reinforcement learning model, and deciding to obtain a target three-phase switching action of the first LCL type grid-connected inverter, wherein the deep reinforcement learning model is obtained by iteratively training a deep Q network using multiple experience elements in an experience replay pool, and the experience elements include a second state of the second LCL type grid-connected inverter at the second moment, a three-phase switching action selected according to a greedy strategy, a reward obtained when executing the three-phase switching action, and a new second state, and the reward is an error between the two-phase current between the second LCL type grid-connected inverter and the grid and a preset grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action; and controlling the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision.
[0130] According to an embodiment of the present application, a processor is also provided, which is used to run a program, wherein the LCL type grid-connected inverter control method in Example 1 is executed when the program is running.
[0131] Optionally, when the program is running, the following steps are executed: obtaining a first state of the first LCL type grid-connected inverter at a first moment, wherein the first state at least includes: two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the grid, and two-phase current and two-phase voltage inside the first LCL type grid-connected inverter; using a deep reinforcement learning model to analyze the first state, and making a decision to obtain a target three-phase switching action of the first LCL type grid-connected inverter, wherein the deep reinforcement learning model is obtained by iteratively training a deep Q network using multiple experience elements in an experience replay pool, and the experience elements include a second state of the second LCL type grid-connected inverter at the second moment, a three-phase switching action selected according to a greedy strategy, a reward obtained when executing the three-phase switching action, and a new second state, and the reward is an error between the two-phase current between the second LCL type grid-connected inverter and the grid and a preset grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action; and controlling the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision.
[0132] According to an embodiment of the present application, an electronic device is also provided, wherein: Figure 4is a schematic diagram of the structure of an optional electronic device according to an embodiment of the present application, such as Figure 4 As shown, the electronic device includes one or more processors; a memory for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors are implemented to run the programs, wherein the programs are configured to execute the LCL type grid-connected inverter control method in the above-mentioned embodiment 1 when running.
[0133] Optionally, the processor is configured to execute the following steps through a computer program: obtaining a first state of the first LCL type grid-connected inverter at a first moment, wherein the first state includes at least: two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the power grid, and two-phase current and two-phase voltage inside the first LCL type grid-connected inverter; analyzing the first state using a deep reinforcement learning model, and deciding to obtain a target three-phase switching action of the first LCL type grid-connected inverter, wherein the deep reinforcement learning model is obtained by iteratively training a deep Q network using multiple experience elements in an experience replay pool, and the experience elements include a second state of the second LCL type grid-connected inverter at the second moment, the three-phase switching action selected according to a greedy strategy, a reward obtained when executing the three-phase switching action, and a new second state, and the reward is an error between the two-phase current between the second LCL type grid-connected inverter and the power grid and a preset power grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action; and controlling the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision.
[0134] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.
[0135] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.
[0136] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.
[0137] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.
[0138] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0139] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the relevant technology or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.
[0140] The above are only preferred implementations of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.
Claims
1. A control method for an LCL type grid-connected inverter, characterized in that: include: Acquire a first state of the first LCL type grid-connected inverter at a first moment, wherein the first state at least includes: two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the power grid, and two-phase current and two-phase voltage inside the first LCL type grid-connected inverter; The first state is analyzed by using a deep reinforcement learning model to determine a target three-phase switching action of the first LCL type grid-connected inverter, wherein the deep reinforcement learning model is obtained by iteratively training a deep Q network using multiple experience elements in an experience replay pool, and the experience elements include a second state of the second LCL type grid-connected inverter at the second moment, a three-phase switching action selected according to a greedy strategy, a reward obtained when the three-phase switching action is executed, and a new second state, and the reward is an error between a two-phase current between the second LCL type grid-connected inverter and a grid and a preset grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action; The first LCL type grid-connected inverter is controlled according to the target three-phase switching action obtained by the decision.
2. The method according to claim 1, characterized in that: Obtaining a target state of a first LCL-type grid-connected inverter at a first moment includes: Acquire the three-phase current and three-phase voltage between the first LCL type grid-connected inverter and the power grid and the three-phase current and three-phase voltage inside the first LCL type grid-connected inverter at a first moment; Using the three-phase voltage between the first LCL type grid-connected inverter and the power grid as an input signal of a phase-locked loop to obtain a phase; Based on the phase, the three-phase current and three-phase voltage between the first LCL type grid-connected inverter and the power grid and the three-phase current and three-phase voltage inside the first LCL type grid-connected inverter are respectively converted from a three-phase stationary coordinate system to a dq coordinate system that rotates synchronously with the voltage using Park transformation, so as to obtain the two-phase current and two-phase voltage between the first LCL type grid-connected inverter and the power grid and the two-phase current and two-phase voltage inside the first LCL type grid-connected inverter.
3. The method according to claim 1, characterized in that The training process of the deep reinforcement learning model includes: Acquire a training sample set and a sample label set, wherein the training sample set includes: a plurality of training samples consisting of a second state of the second LCL type grid-connected inverter at a second moment, and the sample label set includes: sample labels of the plurality of training samples, and the sample labels are used to characterize a three-phase switching action of the second LCL type grid-connected inverter that minimizes a difference between a two-phase current between the second LCL type grid-connected inverter and a grid and a preset grid reference current; Build an initial model based on a deep Q network; The initial model is iteratively trained using the training sample set and the sample label set to obtain the deep reinforcement learning model.
4. The method according to claim 3, characterized in that: The process of determining the grid reference current includes: For each of the second LCL type grid-connected inverters, obtaining an active power command and a reactive power command of the second LCL type grid-connected inverter at a second moment; The grid reference current is determined based on the active power command and the reactive power command and the two-phase voltage between the second LCL type grid-connected inverter and the grid at the second moment.
5. The method according to claim 3, characterized in that: Iteratively training the initial model using the training sample set and the sample label set to obtain the deep reinforcement learning model includes: Defining a state space based on the training sample set, wherein each state in the state space represents a second state of the second LCL type grid-connected inverter at a second moment; Defining an action space based on the sample label set, wherein each action in the action space represents a three-phase switching action of the second LCL type grid-connected inverter; Initializing a deep Q network, wherein the deep Q network includes an online network and a target network; Selecting a second state from the state space using the online network, and selecting a three-phase switch action having a Q value greater than a preset threshold using a greedy algorithm according to the second state; determining a reward and a new second state when the second LCL type grid-connected inverter performs the selected three-phase switching action; The experience element consisting of the second state of the second LCL type grid-connected inverter at the second moment, the three-phase switching action performed by the second LCL type grid-connected inverter, the reward when the second LCL type grid-connected inverter performs the selected three-phase switching action and the new second state is stored in the experience replay pool; The target network is used to determine the target Q value of each experience element in the experience replay pool, and the model parameters of the initial model are adjusted based on the Q value of each experience element and the target Q value to obtain the deep reinforcement learning model.
6. The method according to claim 5, characterized in that The target Q value of each experience element in the experience replay pool is determined using the target network, and the model parameters of the initial model are adjusted based on the Q value of each experience element and the target Q value, including: In each iteration, the following steps are executed repeatedly until the preset termination condition is met: Randomly extracting experience elements from the experience replay pool according to the experience replay mechanism, and using the target network to determine the target Q value of each extracted experience element; Determine a loss function based on analyzing the Q value of each of the experience elements and the corresponding target Q value using an online network; Minimizing the loss function using a gradient descent method to update network parameters of the online network; After at least one iterative training process, the network parameters of the online network are copied to the target network.
7. The method according to claim 4, characterized in that Controlling the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision includes: When any phase switch in the target three-phase switching action is closed, controlling the corresponding phase switch in the first LCL type grid-connected inverter to be in a high level state; When any phase switch in the target three-phase switching action is turned on, the corresponding phase switch in the first LCL type grid-connected inverter is controlled to be in a low level state.
8. An LCL type grid-connected inverter control device, characterized in that: include: An acquisition module, configured to acquire a first state of a first LCL type grid-connected inverter at a first moment, wherein the first state at least includes: a two-phase current and a two-phase voltage between the first LCL type grid-connected inverter and a power grid, and a two-phase current and a two-phase voltage inside the first LCL type grid-connected inverter; a decision module, configured to analyze the first state using a deep reinforcement learning model, and obtain a target three-phase switching action of the first LCL type grid-connected inverter by decision, wherein the deep reinforcement learning model is obtained by iteratively training a deep Q network using a plurality of experience elements in an experience replay pool, and the experience elements include a second state of the second LCL type grid-connected inverter at a second moment, a three-phase switching action selected according to a greedy strategy, a reward obtained when the three-phase switching action is executed, and a new second state, and the reward is an error between a two-phase current between the second LCL type grid-connected inverter and a grid and a preset grid reference current when the second LCL type grid-connected inverter executes the selected three-phase switching action; A control module is used to control the first LCL type grid-connected inverter according to the target three-phase switching action obtained by the decision.
9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a computer program, wherein the device where the non-volatile storage medium is located executes the LCL type grid-connected inverter control method according to any one of claims 1 to 7 by running the computer program.
10. A computer program product, characterized in that include: A computer program, wherein when the computer program is executed by a processor, the LCL type grid-connected inverter control method according to any one of claims 1 to 7 is implemented.
Citation Information
Patent Citations
Inverter controller based on deep reinforcement learning
CN112187074A
Hybrid electric vehicle energy management method based on rule and double-depth Q network
CN113511082A
Layered multi-agent deep reinforcement learning power distribution network voltage regulation method
CN115986750A
Multi-type resource cooperative regulation and control method for three-phase imbalance treatment of distribution network
CN116454926A
LCC-HVDC inverter valve current prediction method
CN116633114A
