基于深度学习的短期电力负荷预测

Short-term load forecasting based on deep learning algorithm

电力负荷预测在智能电网中发挥着重要作用,精准的负荷预测对优化电力系统规划有着重要作用。为解决用户级负荷数据随机性和波动性较强造成短期负荷预测精度低下的问题,利用深度学习方法,基于Keras搭建负荷预测模型,提出一种新的损失函数实现短期/超短期电力负荷的高精度预测。结合某城市的用电和气象数据,采用多层感知机(MLP)和长短期记忆网络(LSTM)两种神经网络测试所提出损失函数的预测效果。仿真结果表明,采用深度学习可以有效预测短期电力负荷变化趋势,且MLP网络的预测精度高于LSTM,表明更多的先验知识有益于预测精度的提高。

The power load forecasting plays an important role in the smart grids,and an accurate load forecasting can affect the optimization of the power system planning.In order to solve the low prediction caused by the randomness and fluctuation of the user-level loads,a deep learning(DL)model with a novel loss function is proposed.By combining the typical urban power loads data and the meteorological data,the short-term forecasting based on Keras is proposed.By compare the prediction performance of the multilayer perceptron(MLP)and the long short-term memory(LSTM)with the proposed loss function,the proposed loss function can achieve better performance.The simulation results show that the DL methods can forecast the load variety trends and the simulation results also show that the MLP outperforms than LSTM by using more prior knowledge.