CN-121981163-A - Model training method and energy storage system virtual impedance self-adaptive adjusting method

CN121981163ACN 121981163 ACN121981163 ACN 121981163ACN-121981163-A

Abstract

The application discloses a model training method and an energy storage system virtual impedance self-adaptive adjusting method, wherein the model training method uses a cyclic neural network as a model framework, takes an output vector of a system as input, the training network can output a dq axis component of a virtual impedance adjusting quantity according to an input output vector sequence, the trained cyclic neural network is used as a virtual impedance adjusting model, real-time output vectors of the system are acquired in actual application, the output vectors are voltage deviation and current deviation under the dq coordinate system, the virtual impedance adjusting model is utilized to identify fault characteristics, so that impedance adjusting decision is carried out, virtual impedance adjusting quantity under the dq coordinate system is predicted, virtual impedance at the next moment is calculated according to the virtual impedance at the current moment and the virtual impedance adjusting quantity, self-adaptive adjustment of the virtual impedance is realized, and stable operation of the system is ensured.

Inventors

TANG QI
CHEN XIA
ZENG QINGHUI
CAO DEFA
DENG YIFAN

Assignees

广东电网有限责任公司佛山供电局

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (10)

1. A method of model training, wherein the method trains a model for predicting a virtual impedance adjustment, comprising: acquiring a historical output vector sequence of an energy storage system, wherein the historical output vector sequence is generated by modeling a state space of the system, the historical state vector sequence comprises state vectors which are continuous in time, and the output vector comprises voltage deviation and current deviation under a dq coordinate system; Training the cyclic neural network by taking the historical output vector sequence as the input of the cyclic neural network to enable the cyclic neural network to output virtual impedance adjustment quantity under a dq coordinate system; And constructing a loss function according to the output of the cyclic neural network, updating network parameters of the cyclic neural network by using the loss function, and taking the trained cyclic neural network as a virtual impedance adjustment model.
2. The method of claim 1, wherein constructing a loss function from the output of the recurrent neural network comprises: , Wherein, the Representing the actual virtual impedance at the current time, Representing the virtual impedance at a time immediately preceding the current time, Represents the virtual impedance adjustment amount of the output of the recurrent neural network, Representing regularization coefficients.
3. The self-adaptive adjustment method for the virtual impedance of the energy storage system is characterized by comprising the following steps of: modeling a state space of the system to generate a state space model of the system; obtaining an output vector of the current moment of the system according to the state space model, wherein the output vector comprises voltage deviation and current deviation under a dq coordinate system; Inputting an output vector at the current moment into a virtual impedance adjustment model trained by the method of any one of claims 1-2, and predicting and generating a virtual impedance adjustment quantity under a dq coordinate system; And calculating the virtual impedance at the next moment according to the virtual impedance at the current moment and the virtual impedance adjustment quantity.
4. The method of claim 3, wherein modeling the state space of the system to generate the state space model of the system comprises: Defining a state vector composed of voltage, current and virtual impedance under a dq coordinate system of the system, determining that an input vector of the system comprises reference voltage and reference current under the dq coordinate system, wherein an output vector comprises voltage deviation and current deviation under the dq coordinate system, and establishing a state equation and an output equation of the system by using the state vector, the input vector and the output vector; Wherein the voltage deviation represents the deviation amount of the real-time voltage and the reference voltage, and the current deviation represents the deviation amount of the real-time current and the reference current.
5. The method of claim 4, wherein the state equation and the output equation are constructed as follows: , wherein A, B, C, D denotes the dynamic characteristics matrix of the system, Is a nonlinear disturbance term of virtual impedance to the system.
6. A method according to claim 3, wherein the calculating of the virtual impedance at the next time from the virtual impedance at the present time and the virtual impedance adjustment amount is as follows: , Wherein, the 、、 The dq-axis component and the zero sequence component representing the virtual impedance at time t, 、、 The dq-axis component and the zero sequence component representing the virtual impedance at time t +1, 、、 The dq-axis component and the zero-sequence component representing the virtual impedance adjustment quantity predicted by the virtual impedance adjustment model.
7. A model training apparatus for training a virtual impedance adjustment model, comprising: The system comprises a state vector acquisition module, a state vector generation module and a state vector generation module, wherein the state vector acquisition module acquires a historical output vector sequence of the system, the historical output vector sequence is generated by carrying out state space modeling on the system, and the output vector comprises voltage deviation and current deviation under a dq coordinate system; The network training module takes the historical output vector sequence as the input of the cyclic neural network, trains the cyclic neural network, and enables the cyclic neural network to output virtual impedance adjustment quantity under the dq coordinate system according to the input output vector sequence; And the network updating module is used for constructing a loss function according to the output of the cyclic neural network, updating network parameters of the cyclic neural network by using the loss function, and taking the trained cyclic neural network as a virtual impedance adjustment model.
8. An energy storage system virtual impedance self-adaptive adjustment device, comprising: The state space modeling module is used for carrying out state space modeling on the system and generating a state space model of the system; The state vector acquisition module acquires an output vector of the system at the current moment according to the state space model, wherein the output vector comprises voltage deviation and current deviation under a dq coordinate system; a virtual impedance prediction module for inputting the output vector at the current time into a virtual impedance adjustment model trained by the model training device according to claim 7, and predicting and generating a virtual impedance adjustment amount in the dq coordinate system; and the virtual impedance adjusting module is used for calculating the virtual impedance at the next moment according to the virtual impedance at the current moment and the virtual impedance adjusting quantity.
9. An electronic device comprising a memory storing computer executable instructions and a processor, which when executed by the processor causes the device to perform the model training method of any one of claims 1-2 and/or the energy storage system virtual impedance adaptation method of any one of claims 3-6.
10. A readable storage medium, storing a computer executable program, which when executed, implements the model training method according to any one of claims 1-2, and/or the energy storage system virtual impedance adaptive adjustment method according to any one of claims 3-6.

Description

Model training method and energy storage system virtual impedance self-adaptive adjusting method Technical Field The invention belongs to the inverter control technology, and particularly relates to a model training method and an energy storage system virtual impedance self-adaptive adjustment method. Background In recent years, with the large-scale access of distributed energy and renewable energy, the role of energy storage systems in electric power systems has become increasingly prominent. The energy storage system not only can balance the load of the power grid and adjust the voltage fluctuation, but also can improve the disturbance rejection capability and the dynamic response speed of the system. To further improve the performance of energy storage systems in transient and fault conditions, virtual impedance techniques have been introduced to improve the dynamic characteristics of the system, enhancing system stability. The traditional virtual impedance adjustment method generally relies on offline modeling and parameter presetting of a system state, and the basic idea is to improve impedance matching between an energy storage system and a power grid by introducing virtual impedance, so that current oscillation is inhibited, and resonance problems are relieved. Most of the existing schemes adopt fixed or pre-tuned parameters to set virtual impedance, and the method can play a certain role under the condition of relatively stable system working conditions, but when the system is faced with transient disturbance or sudden fault, the fixed parameters are difficult to meet actual operation requirements due to dynamic changes of system parameters and operation environments, so that the adjusting effect is greatly reduced. Disclosure of Invention Based on the dynamic behavior of the system is described by using a state space model under a dq coordinate system, the self-adaptive adjustment of the virtual impedance is performed by using a cyclic neural network, and the cooperative control of the virtual impedance is realized for different short-circuit fault types. In a first aspect, the present invention provides a model training method, the trained model being used for predicting a virtual impedance adjustment amount, comprising: Acquiring a historical output vector sequence of the system, wherein the historical output vector sequence is generated by modeling a state space of the system, the historical state vector sequence comprises state vectors which are continuous in time, and the output vector comprises voltage deviation and current deviation under a dq coordinate system; The historical output vector sequence is used as input of a cyclic neural network, and the cyclic neural network is trained to output virtual impedance adjustment quantity under the dq coordinate system; And constructing a loss function according to the output of the cyclic neural network, updating network parameters of the cyclic neural network by using the loss function, and taking the trained cyclic neural network as a virtual impedance adjustment model. Further, constructing a loss function from the output of the recurrent neural network includes: , Wherein, the Representing the actual virtual impedance at the current time,Representing the virtual impedance at a time immediately preceding the current time,Represents the virtual impedance adjustment amount of the output of the recurrent neural network,Representing regularization coefficients. Further, the recurrent neural network adopts an LSTM network. In a second aspect, the present invention provides a method for adaptively adjusting virtual impedance of an energy storage system, including: modeling a state space of the system to generate a state space model of the system; obtaining an output vector of the current moment of the system according to the state space model, wherein the output vector comprises voltage deviation and current deviation under a dq coordinate system; Inputting the output vector at the current moment into a virtual impedance adjustment model trained by the method of the first aspect, and predicting and generating a virtual impedance adjustment quantity under a dq coordinate system; And calculating the virtual impedance at the next moment according to the virtual impedance at the current moment and the virtual impedance adjustment quantity. Further, performing state space modeling on the system, and generating a state space model of the system includes: Defining a state vector composed of voltage, current and virtual impedance under a dq coordinate system of the system, determining that an input vector of the system comprises reference voltage and reference current under the dq coordinate system, and an output vector comprises voltage deviation and current deviation under the dq coordinate system, and establishing a state equation and an output equation of the system by using the state vector, the input vector and the output vector; The voltage deviation