基于GWO优化TCN-GRU模型的区块链电力调控平台负荷预测研究  

Research on load forecasting of blockchain power control platform based on GWO optimized TCN-GRU model

作　　者：徐鹤勇 施佳锋 刘一峰 于晓昆 李骞 XU Heyong;SHI Jiafeng;LIU Yifeng;YU Xiaokun;LI Qian(Ningxia Power Dispatching Control Center,Yinchuan 750001,China;State Grid Blockchain Technology(Beijing)Co.,Ltd.,Beijing 100053,China)

机构地区：[1]宁夏电力调度控制中心,银川750001 [2]国网区块链科技(北京)有限公司,北京100053

出　　处：《国外电子测量技术》2025年第1期148-154,共7页Foreign Electronic Measurement Technology

摘　　要：为提高区块链环境下电力调控平台负荷预测精度,提出了一种改进TCN-GRU模型的预测方法。首先,通过串联时间卷积网络(Temporal Convolutional Network,TCN)和门控循环单元(Gated Recurrent Unit,GRU)网络构建TCN-GRU预测模型;然后,采用灰狼优化算法(Grey Wolf Optimizer,GWO)对TCN-GRU预测模型卷积核大小、隐藏层数、节点数进行优化改进;最后,将改进的TCN-GRU预测模型用于电力调控平台负荷预测,实现了区块链环境下的电力调控平台预测。结果表明,该方法对区块链电力调控平台负荷预测的平均绝对百分误差和均方根误差分别为1.57%和23.44 MW;相较于标准TCN-GRU、CNN、BiLSTM等预测模型,该方法具有更优异的电力调控平台负荷预测性能。由此得出,所提预测方法可行,可为区块链环境下的电力负荷调控提供参考。To improve the accuracy of load forecasting on the power regulation platform in the blockchain environment,an improved TCN-GRU model prediction method was proposed.Firstly,a TCN-GRU prediction model was constructed by concatenating the Temporal Convolutional Network(TCN)network and Gated Recurrent Unit(GRU)network;Then,the Grey Wolf Optimizer(GWO)algorithm was used to optimize and improve the convolution kernel size,hidden layers,and number of nodes of the TCN-GRU prediction model;Finally,the improved TCN-GRU prediction model will be used for load forecasting on the power regulation platform,achieving blockchain based power regulation platform prediction.The results show that the average absolute percentage error and root mean square error of this method for load forecasting on the blockchain power regulation platform are 1.57%and 23.44 MW,respectively;Compared with standard prediction models such as TCN-GRU,CNN,BiLSTM,etc.,this method has better load forecasting performance for power regulation platforms.From this,it can be concluded that this prediction method is feasible and can provide reference for power load regulation in the blockchain environment.

关 键 词：电力调控平台 负荷预测 时间卷积网络 门控循环单元网络 灰狼优化算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]