In this report, three different deep learning al- gorithms: RNN, RGU and CNN are applied for time series forecasting into the distributed level system. The necessary dataset cleaning is to make sure the failure outcome due to the raw data. To determine the best methods for each network load level, the machine learning models are also compared with the MSE and MAPE of nave forecasting model to verify its feasibility. With the enhancement of the sum customer loads, the forecasting result become better and accuracy.

Yang Haolin. The Investigation of Application of Machine Learning into Multi-Level Network Load Forecasting. Academic Journal of Computing & Information Science (2018) Vol. 1: 36-52.

[1] Weicong Kong, Zhao Yang Dong, David J. Hill, Fengji Luo and Yan Xu (2018). Short-Term Residential Load Forecasting Based on Resident Behaviour Learning, IEEE Transactions On Power Systems, Vol. 33, No. 1, pp.1
[2] Liang Xiao, Xingyu Xiao, Canhuang Dai, Mugen Peng, Lichun Wang and H. Vincent Poor(2018). Reinforcement Learning-based Energy Trading for Microgrids, In Press..
[3] Huaizhi Wang, Jiaqi Ruan, Guibin Wang, Bin Zhou, Yitao Liu, Xueqian Fu, Jianchun Peng(2018). Deep Learning Based Interval State Estimation of AC Smart Grids against Sparse Cyber Attacks , In Press.
[4] L.M. Borges, F.J. Velez and A.S. Lebres (2014). Survey on the characterization and classification of wireless sensor network applications. Communications Surveys & Tutorials IEEE, vol.16, no.4, p.1860-1890.