CN-122026393-A - Load frequency control method of performance index dependent data driving

CN122026393ACN 122026393 ACN122026393 ACN 122026393ACN-122026393-A

Abstract

The application belongs to the technical field of load frequency control, and particularly discloses a load frequency control method driven by performance index dependent data, which comprises the following steps of obtaining a statistical average value of a control performance standard CPS1 of a power system in a set sliding period and a change rate of the statistical average value, and carrying out standardized treatment on the statistical average value; the method comprises the steps of inputting processed data into a neural network model, wherein the model comprises a shared feature extraction layer and a double-branch output layer, the shared feature extraction layer is used for extracting system health states and change trend feature vectors from standardized input data, a first branch output layer in the double-branch output layer is used for outputting a controller gain value, a second branch output layer in the double-branch output layer is used for outputting an event trigger threshold value, and event trigger load frequency control is executed on an electric power system based on the controller gain value and the event trigger threshold value. The application can realize the load frequency control decision with stronger adaptability and more accuracy.

Inventors

Shangguan xingchen
OuYang Yadong
ZHANG CHUANKE
XU DA
JIN LI
WEI CHENGUANG
YANG YUANHANG
TIAN JI

Assignees

中国地质大学（武汉）

Dates

Publication Date: 20260512
Application Date: 20251209

Claims (10)

1. A performance index dependent data driven load frequency control method is characterized by comprising the following steps: S10, acquiring a statistical average value of a control performance standard CPS1 of the electric power system in a set sliding period and a change rate of the statistical average value; S20, carrying out standardized processing on the statistical mean value and the change rate to obtain standardized input data; S30, inputting the standardized input data into a pre-trained neural network model, wherein the neural network model comprises a shared feature extraction layer and a dual-branch output layer, wherein the shared feature extraction layer is used for extracting system health state and change trend feature vectors from the standardized input data; and S40, executing event trigger load frequency control on the electric power system based on the controller gain value and the event trigger threshold.
2. The method of claim 1, wherein in step S30, the shared feature extraction layer includes at least one feature extraction unit, and each feature extraction unit includes a full connection layer, a batch normalization layer, an activation function layer, and a discard layer connected in sequence.
3. The performance index dependent data driven load frequency control method according to claim 2, wherein the construction flow of the feature extraction unit is: input is Input, firstly, full connection operation is carried out, and the formula is that The method comprises the steps of representing that a full connection layer possibly exists, carrying out batch standardization through a BN layer, re-pulling the characteristic value distribution of the layer back to standard normal distribution through a data method to accelerate convergence, introducing nonlinearity through LeakyRelu activation functions after data standardization, finally carrying out dropout regularization to obtain an Output result Output of a characteristic extraction unit, and taking the Output result Output as input and entering a next characteristic extraction unit.
4. The method of claim 1, wherein in step S30, the first branch output layer and the second branch output layer each include a full connection layer, an activation function layer, and an inverse normalization layer, which are sequentially connected.
5. The performance index dependent data driven load frequency control method according to claim 1, wherein in step S20, the statistical average and the change rate are subjected to Z-Score normalization.
6. The performance index dependent data driven load frequency control method according to claim 5, wherein the calculation formula of the Z-Score normalization process is: In the formula, An original value representing the statistical mean or rate of change to be normalized, To train the mean of the corresponding features in the dataset, To train the standard deviation of the corresponding features in the dataset, Is the result after normalization.
7. The method according to claim 1, wherein in step S10, the statistical average of the control performance standard CPS1 is an average of a compliance factor of a relationship between an average of control errors per minute and an average of frequency deviations in a set sliding period, and the change rate is a relative change amount of the average at a current calculation time with respect to a previous calculation time.
8. A load frequency control system for implementing the method according to any one of claims 1 to 7, the system comprising: The data acquisition module is used for acquiring the statistical average value of a first control performance standard CPS1 of the electric power system in a set sliding period and the change rate of the statistical average value; The data processing module is used for carrying out standardized processing on the statistical mean value and the change rate to obtain standardized input data; The neural network processing module is used for storing a neural network model which is trained in advance and is used for receiving the standardized input data, wherein the neural network model comprises a shared feature extraction layer and a double-branch output layer, the shared feature extraction layer is used for extracting a system health state and a change trend feature vector from the standardized input data, a first branch output layer in the double-branch output layer is used for outputting a controller gain value based on the feature vector, and a second branch output layer in the double-branch output layer is used for outputting an event trigger threshold based on the feature vector; and the control execution module is used for executing event trigger load frequency control on the electric power system based on the controller gain value and the event trigger threshold.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1-7 when the program is executed by the processor.
10. A computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the steps of the method of any of claims 1 to 7.

Description

Load frequency control method of performance index dependent data driving Technical Field The application belongs to the technical field of load frequency control, and particularly relates to a load frequency control method driven by performance index dependent data. Background Modern power systems are moving towards high proportion of renewable energy access and high power electronics, and their operational characteristics are becoming more complex. The strong randomness, volatility and low inertia of renewable energy sources lead to weakening of the frequency modulation capability of the power system and rapid reduction of the frequency quality of the power grid. In addition, to address the access of distributed power sources and multiple loads, the grid uses a wide area network to support the coordinated operation of the source network load storage components. However, wide area communication brings convenience and intelligence, and simultaneously introduces network induction problems such as transmission delay, bandwidth limitation, data packet loss and the like. The uncertainty of the information side and the inherent strong fluctuation of the physical side new energy are interwoven, so that the safe and stable operation of the power grid frequency is affected. The core goal of the load frequency control is to quickly stabilize the frequency deviation caused by load fluctuation and the exchange power deviation between areas, so that the safe and stable operation of the power grid is ensured. Conventional LFCs are typically designed with the highest criteria of "asymptotically stable" in frequency and link power. The design concept of pursuing absolute stability is always conservative, and needs to closely track the load change through continuous and rapid control signals, which inevitably leads to mechanical abrasion and operation and maintenance cost of the generator set due to frequent adjustment, and continuous pressure on communication bandwidth and computing resources due to uninterrupted transmission of system measurement values and control signals under a centralized control architecture. At present, cost optimization is studied, a control scheme is designed only from the viewpoint of reducing frequency modulation or communication burden, a collaborative optimization scheme is few, and the requirement of a system CPSs frequency is not considered due to asymptotically stable design. The learner uses the control performance standard to combine with the fuzzy control, and the proposed fuzzy event triggering scheme achieves the aim of the cooperative optimization of frequency modulation and communication cost. However, the fuzzy event triggering scheme relies on discrete rules formulated by expert experience, and has inherent limitations of insufficient adaptability and insufficient utilization of deep data characteristics of control performance indexes in the face of increasingly complex nonlinear system dynamics and multiple uncertainty perturbations. Therefore, how to implement a load frequency control strategy with higher adaptability and higher accuracy is a current urgent problem to be solved. Disclosure of Invention Aiming at the defects of the prior art, the application aims to provide a load frequency control method of performance index dependent data driving, which takes 'shared feature extraction and double-branch output' as cores, takes the statistical mean value of CPS1 indexes and the change rate thereof as core inputs, designs event triggering logic, and can realize control decision with stronger adaptability and more accuracy. To achieve the above object, in a first aspect, the present application provides a performance index dependent data driven load frequency control method, comprising the steps of: S10, acquiring a statistical average value of a control performance standard CPS1 of the electric power system in a set sliding period and a change rate of the statistical average value; S20, carrying out standardized processing on the statistical mean value and the change rate to obtain standardized input data; S30, inputting the standardized input data into a pre-trained neural network model, wherein the neural network model comprises a shared feature extraction layer and a dual-branch output layer, wherein the shared feature extraction layer is used for extracting system health state and change trend feature vectors from the standardized input data; and S40, executing event trigger load frequency control on the electric power system based on the controller gain value and the event trigger threshold. The performance index-dependent data-driven load frequency control method has the following effects that a special framework combining a deep feed-forward network and a double-branch output structure is designed, the design surpasses simple input-output mapping, a thinking process of an intelligent decision system is simulated, wherein a front shared layer of the network replaces a fuzzy