基于混合神经网络深度学习的短期负荷预测

针对能源互联网环境下用电用户数据量大、多维度这一特点,提出了一种混合神经网络深度学习的短期电力负荷预测方法。首先,考虑常见的电力系统负荷的影响因素,建立多维数据库,并进行偏相关分析,排除其他变量干扰;其次,将LSTM、GRU两种神经网络作为前端神经网络对多维数据库中数据进行处理;最后,采用随机概率剔除与Adam训练优化函数改进的BP神经网络作为末端神经网络,建立负荷预测模型。通过算例仿真对本文方法与传统BP神经网络、LSTM神经网络、GRU神经网络进行了对比,验证了所提方法的有效性。

Short-term Load Forecasting Based on Deep Learning of Hybrid Neural Networks

Aiming at the large amount and multi-dimensionality of electricity users in the energy Internet environment, this paper proposes a hybrid neural network deep learning method for short-term load forecasting. Firstly, considering the factors affecting the load in the common power systems, a multi-dimension database was established, and partial correlation analysis was performed to eliminate the interference of other variables. Secondly, LSTM and GRU were selected as front-end neural networks to handle the data in multi-dimensional database. Finally, the BP neural network improved by random probability and Adam training optimization function was used as the terminal neural network to establish load forecasting model. The proposed method was compared with the traditional BP, LSTM and GRU neural networks by numerical simulation. The results verify the effectiveness of the proposed method.