CN-121863529-B - Control method and control device of grid-connected converter and electric energy router

CN121863529BCN 121863529 BCN121863529 BCN 121863529BCN-121863529-B

Abstract

The application relates to a control method, a control device and an electric energy router of a grid-connected converter, which are used for acquiring the running state information of a power grid in real time, inputting the running state information into a trained decision model, dynamically outputting at least one self-adaptive coefficient for coordinating a grid-following control mode and a grid-constructing control mode by the decision model according to the running state information, fusing a first control signal from a grid-following control link and a second control signal from the grid-constructing control link according to the at least one self-adaptive coefficient to generate a control instruction, and controlling the running of the grid-connected converter according to the control instruction.

Inventors

GUAN WEIDE
CHEN YANG
LIU JIALE
LIU ZIHAO
JIANG PEILING
LIU SU
ZHOU ZAN
LU JUNJIE
ZHU YINGXI
WANG ZHENBANG

Assignees

长沙理工大学

Dates

Publication Date: 20260512
Application Date: 20260317

Claims (6)

1. A method for controlling a grid-connected inverter, the method comprising: acquiring running state information of a power grid in real time; Inputting the running state information into a trained decision model, and dynamically outputting at least one self-adaptive coefficient for coordinating a follow-up network control mode and a networking control mode by the decision model according to the running state information; According to the at least one self-adaptive coefficient, a first control signal from a network following control link and a second control signal from a network constructing control link are fused to generate a control instruction; Controlling the operation of the grid-connected converter according to the control instruction; the method further comprises the steps of: training based on a dual-delay depth deterministic strategy gradient algorithm to obtain the decision model; The decision model comprises an Actor online network, two Critic online networks, and a corresponding Actor target network and two Critic target networks; The training of the strategy gradient algorithm based on the dual-delay depth deterministic strategy to obtain the decision model comprises the following steps: Initializing parameters of the Actor online network, the two Critic online networks, the Actor target network and the two Critic target networks, and creating an experience playback pool; Inputting the acquired current training state information into the Actor online network at each time step to obtain a current state action, adding noise into the current state action to obtain a noisy action, obtaining a reward value according to the current training state information and a preset reward function, acquiring next training state information, and storing the current training state information, the current state action, the noisy action, the reward value and the next training state information as experience data into an experience playback pool; Sampling experience data from the experience playback pool, obtaining target action values based on the sampled data, namely the Actor target network and the two Critic target networks, and updating parameters of the two Critic online networks by taking the error between the action values output by the two Critic online networks according to the sampled data and the target action values as a target; Updating the Actor online network by adopting a deterministic strategy gradient every preset updating step number, and updating parameters of the Actor target network and the two Critic target networks based on a soft updating mode; repeating the steps until reaching the preset training termination condition; the at least one adaptive coefficient comprises a heel-net adaptive coefficient and a network-constructing adaptive coefficient, and the sum of the heel-net adaptive coefficient and the network-constructing adaptive coefficient is 1.
2. The method of claim 1, wherein the operating state information includes a short circuit ratio characterizing grid strength, a deviation of the grid-tied converter output power from a reference power, and a deviation of a point of common coupling voltage from a reference voltage.
3. The method of claim 1, wherein the current training state information comprises a deviation of a training output power from a reference power and a deviation of a training pcc voltage from a reference voltage; The reward value is inversely related to a deviation of the training output power from a reference power and a deviation of the training point of common coupling voltage from a reference voltage.
4. The method of claim 1, wherein the fusing the first control signal from the heel control element and the second control signal from the grid-formation control element based on the at least one adaptive coefficient comprises: And carrying out weighted summation on the first control signal from the network following control link and the second control signal from the network constructing control link according to the weight determined by the network following self-adaptive coefficient and the network constructing self-adaptive coefficient.
5. A control device for a grid-connected inverter, configured to implement the method according to any one of claims 1 to 4, wherein the control device includes: the acquisition module is used for acquiring the running state information of the power grid in real time; the decision module is used for inputting the running state information into a trained decision model, and dynamically outputting at least one self-adaptive coefficient for coordinating a network following control mode and a network constructing control mode according to the running state information by the decision model; the fusion module is used for fusing a first control signal from the following network control link and a second control signal from the networking control link according to the at least one self-adaptive coefficient to generate a control instruction; And the control module is used for controlling the running of the grid-connected converter according to the control instruction.
6. An electric energy router, comprising the control device, grid-connected converter, inverter and hybrid energy storage module of claim 5; The converter is connected between the grid-connected converter and the hybrid energy storage module and is used for realizing power transmission and electrical isolation; The hybrid energy storage module comprises a lithium battery and a super capacitor which are connected in parallel.

Description

Control method and control device of grid-connected converter and electric energy router Technical Field The application belongs to the technical field of power electronics, and particularly relates to a control method and device of a grid-connected converter and an electric energy router. Background With the large-scale access of intermittent renewable energy sources such as wind power, photovoltaic and the like to a power system, the strength and inertia of a power grid show a decreasing trend, and higher requirements are put forward on the operation stability and control performance of a grid-connected converter, the grid-connected converter is used as a key interface between a renewable energy source generating unit and the power grid, and a control strategy of the grid-connected converter needs to meet multiple targets such as rapid synchronous tracking, dynamic stable support, fault ride-through and the like under a complex and changeable power grid environment, so that the grid-connected converter becomes a research focus in the current power electronics and system control field. In the prior art, a single grid following control or grid construction control strategy is mainly adopted in the grid following control, the grid following control depends on a phase-locked loop to track the voltage phase of a power grid in real time, and the grid following control has excellent synchronization performance under the condition of a strong power grid, but the control essence is 'power grid following', active support of voltage and frequency cannot be provided for a system, unstable oscillation is easily caused under the condition of a weak power grid or fault, the grid construction control can autonomously construct and stabilize alternating voltage by simulating the external characteristics of a synchronous generator, necessary inertia and damping are provided for the power grid, the dynamic response is relatively slow, the synchronous instability risk exists when the frequency of the power grid rapidly fluctuates or the phase is suddenly changed, and the two single control modes have inherent contradiction in performance, so that the requirements under different working conditions are difficult to be considered. Disclosure of Invention The application aims to provide a control method, a control device and an electric energy router of a grid-connected converter, which are used for solving the problem that the optimal coordination cannot be realized according to the working condition change in the prior art by introducing an intelligent model capable of independently deciding according to the real-time running state information of a power grid and fusing two control signals according to the self-adaptive coefficient dynamically output by the model. The application provides a control method of a grid-connected converter, which comprises the following steps: acquiring running state information of a power grid in real time; Inputting the running state information into a trained decision model, and dynamically outputting at least one self-adaptive coefficient for coordinating a follow-up network control mode and a networking control mode by the decision model according to the running state information; According to the at least one self-adaptive coefficient, a first control signal from a network following control link and a second control signal from a network constructing control link are fused to generate a control instruction; And controlling the running of the grid-connected converter according to the control instruction. Optionally, the operation state information comprises a short circuit ratio representing the strength of the power grid, a deviation of the output power of the grid-connected converter from a reference power and a deviation of a public coupling point voltage from a reference voltage. Optionally, the method further comprises: And training based on a dual-delay depth deterministic strategy gradient algorithm to obtain the decision model. Optionally, the decision model comprises an Actor online network and two Critic online networks, and a corresponding Actor target network and two Critic target networks. Optionally, the training of the dual-delay depth deterministic strategy gradient algorithm to obtain the decision model includes: Initializing parameters of the Actor online network, the two Critic online networks, the Actor target network and the two Critic target networks, and creating an experience playback pool; Inputting the acquired current training state information into the Actor online network at each time step to obtain a current state action, adding noise into the current state action to obtain a noisy action, obtaining a reward value according to the current training state information and a preset reward function, acquiring next training state information, and storing the current training state information, the current state action, the noisy action, the reward value and the next tr