基于条件生成对抗网络的短期负荷预测

精准的短期负荷预测对电力系统制定合理生产计划、提高经济效益、保证电网安全运行具有重要意义.为学习非线性负荷数据中隐含的深层关系,提高短期负荷预测精度,文中提出一种基于条件生成对抗网络的短期负荷预测模型.所提模型使用卷积神经网络构建生成模型和判别模型,以负荷影响因素作为条件,并引入特征损失函数作为判别模型部分隐藏层的损失函数.然后,通过条件生成对抗网络的博弈训练,使生成模型以负荷影响因素为条件生成预测负荷数据,从而进行短期负荷预测.最后,以美国某地区3年的负荷作为实际算例,对比所提模型与其他模型的预测结果,验证了所提模型在兼顾泛化能力的同时可以提高短期负荷的预测精度.

Short-term Load Forecasting Based on Conditional Generative Adversarial Network

Accurate short-term load forecasting is of great significance for power systems to formulate rational production plans, improve economic benefits, and ensure safe operation of power grids. In order to learn the hidden deep relationship in nonlinear load data and improve the accuracy of the short-term load forecasting, this paper proposes a short-term load forecasting model based on the conditional generative adversarial network. This model uses the convolutional neural network to construct a generative model and a discriminant model, which takes load influencing factors as conditions, and introduces a feature loss function as the loss function of some hidden layers in the discriminant model. Through the game training of the conditional generative adversarial network, the generative model takes the load influencing factors as the conditions to generate the forecasting load data, and then performs short-term load forecasting. Finally, taking the three-year load in a certain area of the United States as a practical example, the forecasting results of the proposed model are compared with other models. It is verified that the proposed model can improve the accuracy of the short-term load forecasting while considering the generalization ability.