基于CNN-LSTM组合模型光伏预测和负荷预测算法的研究与应用
Research and Application of PV Forecasting and Load Forecasting Algorithm Based on CNN-LSTM Combination Model
邹兵1
作者信息
- 1. 中石化胜利油田分公司生产运行管理中心,山东东营 257000
- 折叠
摘要
油田企业电网中光伏装机容量快速增长,但由于光伏发电受太阳辐照度等气象因素影响,其间歇性、波动性给油田企业电网的安全稳定运行造成严重影响.同时,以往的单一模型预测光伏发电有一定的局限性.结合卷积神经网络(CNN)和长短期记忆网络(LSTM)的特点,提出一种基于 CNN-LSTM 的组合预测模型,并在某光伏电站进行了仿真实验.实验结果表明,CNN-LSTM模型对光伏发电功率预测的均方根误差与平均绝对误差分别为 0.212 1 与 0.129 0,对负荷预测的均方根误差与平均绝对误差分别为 0.209 7 与0.115 5.验证了该模型对光伏发电功率与负荷预测的有效性.预测结果可以指导源网荷运行计划,提高光伏消纳,提升电力系统安全可靠运行水平,为油田电网的安全低碳运行提供准确的决策支撑.
Abstract
The installed photovoltaic capacity in the power grid of oilfield enterprises is growing rapidly,but due to the influence of solar irradiance and other meteorological factors,the intermittency and volatility of photovoltaic power generation have a serious impact on the safe and stable operation of the power grid of oilfield enterprises.At the same time,the previous single model prediction of photovoltaic power genera-tion has certain limitations.Combining the features of convolutional neural network(CNN)and long short-term memory network(LSTM),a combined predic-tion model based on CNN-LSTM was proposed,and simulation experiments were carried out in a photovol-taic power station.The experimental results showed that the root-mean-square error and average absolute error of CNN-LSTM model for PV power prediction were 0.212 1 and 0.129 0,respectively,and the root-mean-square error and average absolute error of load prediction were 0.209 7 and 0.115 5,respec-tively.The validity of the model for power and load prediction of photovoltaic power generation was veri-fied.The predicted results can guide the operation plan of the source network load,improve the photo-voltaic consumption,improve the safe and reliable operation level of the power system,and provide ac-curate decision-making support for the safe and low-carbon operation of the oilfield power grid.
关键词
光伏发电/功率/负荷预测/模型/安全Key words
photovoltaic power generation/power/load forecasting/model/secure引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据