| 引用本文: | 彭道刚,朱琪,车权,赵慧荣.基于CNN-LSTM神经网络的电网调度火电厂短期存煤预测[J].电力自动化设备,2021,41(6): |
| PENG Daogang,ZHU Qi,CHE Quan,ZHAO Huirong.Short-term coal storage forecasting of thermal power plant for power grid dispatching based on CNN-LSTM neural network[J].Electric Power Automation Equipment,2021,41(6): |
|
| 摘要: |
| 采用传统的回归拟合进行发电厂存煤量单变量单步预测已无法满足电网优化调度的需求，针对该问题，提出卷积神经网络(CNN)与长短期记忆(LSTM)神经网络相结合的CNN-LSTM神经网络预测模型，利用CNN良好的特征提取能力以及LSTM神经网络特殊的记忆预测功能实现对未来电厂存煤量的精准预测。为了使预测结果更符合实际存煤量，在已有预测结果的基础上进行进一步优化。实例验证结果表明，相较于传统差分自回归移动平均(ARIMA)模型和单一LSTM神经网络模型，所提模型取得的效果更好，且经过优化后的预测精度得到了进一步提高。 |
| 关键词: 短期预测 电厂存煤 深度学习 模型优化 电网调度 |
| DOI：10.16081/j.epae.202102025 |
| 分类号:TM73 |
| 基金项目:国家自然科学基金资助项目(52006131)；上海市“科技创新行动计划”高新技术领域项目(19511101600)；上海市青年科技英才扬帆计划资助项目(20YF1414900)；国网重庆市电力公司科技项目(2019渝电科技15#) |
|
| Short-term coal storage forecasting of thermal power plant for power grid dispatching based on CNN-LSTM neural network |
|
PENG Daogang1,2, ZHU Qi1,2, CHE Quan3, ZHAO Huirong1,2
|
|
1.College of Automation Engineering, Shanghai University of Electric Power, Shanghai 200090, China;2.Shanghai Engineering Research Center of Intelligent Management and Control for Power Generation Process, Shanghai 200090, China;3.State Grid Chongqing Electric Power Company, Chongqing 400014, China
|
| Abstract: |
| The univariate single-step forecasting of coal storage for power plant with traditional regression fitting cannot meet the need of optimal dispatching for power grid, for this problem, CNN(Convolutional Neural Networks)and LSTM(Long Short-Term Memory) neural network are combined, and a CNN-LSTM neural network forecasting model is proposed, which uses CNN’s good extraction capability and LSTM’s special memory forecasting function to realize accuracy forecasting of future coal storage for power plant. In order to make the forecasting results more consistent with the actual coal storage, further optimization is carried out based on the existing forecasting results. Case verification results show that, compared with the traditional ARIMA(AutoRegressive Integrated Moving Average) model and single LSTM neural network model, the proposed model obtains better effect, and the forecasting accuracy after optimization is further improved. |
| Key words: short-term forecasting coal storage of power plant deep learning model optimization power grid dispatching |
