基于数据驱动的智能电网光伏能源预测方法研究  

Research on data-driven smart grid photovoltaic energy prediction method

作　　者：沈宁 Shen Ning(State Grid Qinghai Electric Power Company,Xining 810008,China)

机构地区：[1]国网青海省电力公司,青海西宁810008

出　　处：《能源与环保》2024年第11期193-197,204,共6页CHINA ENERGY AND ENVIRONMENTAL PROTECTION

基　　金：国网青海省科技项目(106000007890)。

摘　　要：光伏能源作为全球可再生能源体系中的核心组成部分,其产能波动受气候条件影响显著,从而给电力系统的管理带来了极大的挑战。为此,探索一种有效的预测工具对于智能电网的能源集成、控制与运营至关重要。提出了一个基于数据驱动的智能电网光伏能源预测模型,该模型通过多种神经网络和历史光伏发电数据精确预测未来的光伏发电量。研究中比较了多种预测算法,包括长短期记忆网络(LSTM)、前馈神经网络(FFNN)以及门控循环单元(GRU),并依据均方根误差(RMSE)、平均绝对误差(MAE)及决定系数(R2)等关键指标,全面评估了各算法的预测效果。实验结果表明,LSTM和GRU在处理复杂的时间依赖关系时表现得更好,而FFNN在特定单元数量(如100和150)时的预测效果较为出色。实验验证了所提方法在光伏能源预测领域的有效性和应用潜力,不仅为智能电网的优化提供了科学依据,也为电力系统的管理者提供了一种强有力的决策工具。Photovoltaic(PV)energy,as a core component of the global renewable energy system,has a fluctuating capacity that is significantly affected by climatic conditions,thus posing a great challenge to the management of the power system.For this reason,exploring an effective prediction tool is crucial for smart grid energy integration,control and operation.In this study,a data-driven PV energy prediction model for smart grids was proposed,which accurately predicts future PV power generation through multiple neural networks and historical PV generation data.Multiple prediction algorithms,including Long Short-Term Memory Network(LSTM),Feed-Forward Neural Network(FFNN),and Gated Recurrent Unit(GRU),were compared in the study,and the prediction effectiveness of the algorithms was comprehensively evaluated based on key metrics,such as Root Mean Square Error(RMSE),Mean Absolute Error(MAE),and Coefficient of Determination(R 2).The experimental results showed that LSTM and GRU performed better in dealing with complex temporal dependencies,while FFNN was more effective in predicting a specific number of units(such as100 and 150).The experiments verified the effectiveness and application potential of the proposed method in the field of PV energy prediction,which not only provides a scientific basis for the optimization of smart grids,but also provides a powerful decision-making tool for power system managers.

关 键 词：光伏能源预测 神经网络 数据驱动 能源管理 

分 类 号：TM615[电气工程—电力系统及自动化]