基于ELM预测模型的高比例新能源电网改进频率控制策略  

Improved Frequency Control Strategy for Power Grid with High Proportion of Renewable Energy Based on ELM Prediction Model

作　　者：胡亚平[1] 聂涌泉 何宇斌 陈根军[2] 林子杰 曹武[3] HU Yaping;NIE Yongquan;HE Yubin;CHEN Genjun;LIN Zijie;CAO Wu(Power Dispatch and Control Center of China Southern Power Grid,Guangzhou 510000,Guangdong,China;Nanjing Nari-Relays Elctric Co.,Ltd,Nanjing 211106,Jiangsu,China;School of Electrical Engineering,Southeast University,Nanjing 210096,Jiangsu,China)

机构地区：[1]中国南方电网电力调度控制中心,广东广州510000 [2]南京南瑞继保电气有限公司,江苏南京211106 [3]东南大学电气工程学院,江苏南京210096

出　　处：《电网与清洁能源》2022年第7期98-106,共9页Power System and Clean Energy

基　　金：国家重点研发计划项目(2018YFB1502904)。

摘　　要：高比例新能源电网中,功率与频率变化存在很强的非线性,自动发电控制(AGC)作为电网调节频率的主要控制手段,目前的控制方式无法很好地适应强非线性特性电网的调频需求。鉴于此,提出了基于极限学习机(ELM)预测模型的高比例新能源电网改进频率控制策略。其特点在于通过ELM算法和历史运行数据,建立电网功率变化与频率变化的实时频率预测模型,进一步基于预测模型分析AGC调节机组的调频能力,按照调频能力优化AGC的区域功率控制需求功率分配。其优势在于通过机器学习拟合频率非线性调节规律,优化AGC频率控制,提高系统频率调节的快速性和可靠性,从而提高含新能源电网稳定性。最后通过电网SCADA实际数据建立预测模型并验证其准确性和实时性,并通过应用实例证明所提策略可以实现快速稳定调频。The relationship between power and frequency changes has strong non-linearity in power grids with renewable energy sources.However,as the main control means of regulating frequency in power grids,the existing automatic generation control(AGC)fails to well adapt to the regulation requirements of power grids with strong non-linearity.In order to better adapt to the non-linear characteristics of frequency and power of the power grid with renewable energy and improve frequency stability,this paper proposes a frequency control strategy for the power grid with high proportion of renewable energy based on prediction model using extreme learning machine(ELM).Based on the analysis of relevant factors of power regulation in the power grid,a real-time frequency prediction model of power and frequency variation in the power grid is established by using ELM algorithm.The sensitivity of AGC regulating unit to system frequency variation is further analyzed by the prediction model.And power distribution control is optimized according to sensitivity and AGC regional power control requirements.Its advantage lies in using machine learning to fit frequency nonlinear regulation and then improve the speed and reliability of system frequency regulation.In the end,the prediction model is established by the actual data of SCADA in the power grid and its accuracy and real-time characteristic are verified.It is also verified that the proposed strategy can achieve fast and stable frequency regulation through an application example.

关 键 词：ELM 非线性 高比例新能源 频率控制 

分 类 号：TM73[电气工程—电力系统及自动化]