| 引用本文: | 李滨,王靖德,梁水莹,韦昌福.基于长短期记忆循环神经网络的AGC实时控制策略[J].电力自动化设备,2022,42(3): |
| LI Bin,WANG Jingde,LIANG Shuiying,WEI Changfu.AGC real-time control strategy based on LSTM recurrent neural network[J].Electric Power Automation Equipment,2022,42(3): |
|
| 摘要: |
| 大量新能源的接入以及电网中冲击负荷数量的剧增，使得电网对自动发电控制(AGC)策略提出了新的要求。简化AGC的一般控制流程，对比不同AGC策略的控制特性，在每个考核周期内选择控制效果更优的控制策略，并充分发挥多种控制策略在各自优势工况下的性能，以得到优秀控制数据集；在此基础上，以长短期记忆(LSTM)循环神经网络为神经元构建AGC策略深度学习模型，并提出一种基于LSTM循环神经网络的数据驱动型AGC实时控制策略。仿真结果表明，基于深度学习的控制策略的整体性能优于任何单一控制策略。 |
| 关键词: 自动发电控制 控制策略 深度学习 长短期记忆循环神经网络 数据驱动 |
| DOI：10.16081/j.epae.202111014 |
| 分类号:TM73 |
| 基金项目:国家自然科学基金资助项目(51767004) |
|
| AGC real-time control strategy based on LSTM recurrent neural network |
|
LI Bin1, WANG Jingde1, LIANG Shuiying2, WEI Changfu3
|
|
1.Guangxi Key Laboratory of Power System Optimization and Energy Technology, Guangxi University, Nanning 530004, China;2.Electric Power Research Institute of Guangxi Power Grid Co.,Ltd.,Nanning 530023, China;3.Power Dispatching Control Center of Guangxi Power Grid Co.,Ltd.,Nanning 530023, China
|
| Abstract: |
| The connection of a large number of renewable energy and the sharp increase of the number of impact loads in power grid make power grid put forward new requirements for AGC(Automatic Generation Control) strategies. The general control process of AGC is simplified, the control characteristics of different AGC strategies are compared, the control strategy with better control effect in each evaluation cycle is selected, and the performance of multiple control strategies under their respective advantageous conditions is given full play to obtain excellent control data set. On this basis, LSTM(Long Short-Term Memory) recurrent neural network is taken as the neuron to construct a deep learning model of AGC strategy, and a data-driven AGC real-time control strategy based on LSTM recurrent neural network is proposed. The simulative results show that the overall performance of the control strategy based on deep learning is better than any single control strategy. |
| Key words: automatic generation control control strategy deep learning long short-term memory recurrent neural network data-driven |
