Distribution Long-term Load Forecast Using Long Short-Term Memory Network

Long Short-Term Memory (LSTM) network is an enhanced Recurrent Neural Network (RNN) that has gained significant attention in recent years. It solved the vanishing and exploding gradient problems standard RNN has and has been successfully applied to a variety of time-series forecasting problems. In power systems, distribution feeder long-term load forecast is a critical task many electric utility companies perform on an annual basis. The goal of this task is to forecast the load growth on existing distribution feeders for the next one to ten years. The forecasted results will be used as input in long-term feeder planning studies to determine future distribution feeder configuration, feeder tie addition and transformer capacity addition so that the distribution system can operate reliably during normal operation and contingences. This research applied LSTM network to this important and challenging task. This paper firstly explained the concept and structure of LSTM and then discussed the details of data features and dataset preparation. The dataset is then fit into a LSTM network to establish the long-term load forecast model. In the end, an application example in a Canadian Utility company is provided. The results are compared and discussed. It is proven that the proposed approach is effective for distribution feeder long-term load forecast, especially for saturated feeders and high residential percentage feeders.

[1]  Hiroshi Chin.  An analytical evaluation of top-down versus bottom-up forecast in the electricity demand , 2016, 2016 IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW).

[2]  Jürgen Schmidhuber,et al.  Learning to Forget: Continual Prediction with LSTM , 2000, Neural Computation.

[3]  Qia Ding,et al.  Long-Term Load Forecast Using Decision Tree Method , 2006, 2006 IEEE PES Power Systems Conference and Exposition.

[4]  Mohammed Ouassaid,et al.  Long term electricity demand forecasting using autoregressive integrated moving average model: Case study of Morocco , 2015, 2015 International Conference on Electrical and Information Technologies (ICEIT).

[5]  Quoc V. Le,et al.  Sequence to Sequence Learning with Neural Networks , 2014, NIPS.

[6]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[7]  Yuan Zhang,et al.  Short-Term Residential Load Forecasting Based on LSTM Recurrent Neural Network , 2019, IEEE Transactions on Smart Grid.

[8]  Bhekisipho Twala,et al.  Forecasting electricity consumption in South Africa: ARMA, neural networks and neuro-fuzzy systems , 2014, 2014 International Joint Conference on Neural Networks (IJCNN).

[9]  Siyuan Liu,et al.  Stock transaction prediction modeling and analysis based on LSTM , 2018, 2018 13th IEEE Conference on Industrial Electronics and Applications (ICIEA).

[10]  Yoshua Bengio,et al.  Learning long-term dependencies with gradient descent is difficult , 1994, IEEE Trans. Neural Networks.

[11]  C. Dortolina,et al.  Estimating future demand a top down/bottom up approach for forecasting annual growths , 2005, IEEE Power Engineering Society General Meeting, 2005.

[12]  Yoshua Bengio,et al.  Gradient Flow in Recurrent Nets: the Difficulty of Learning Long-Term Dependencies , 2001 .

[13]  S. C. Kremer,et al.  Gradient Flow in Recurrent Nets: the Difficulty of Learning Long-Term Dependencies , 2001 .

[14]  Chaker El Amrani,et al.  Sequence to Sequence Weather Forecasting with Long Short-Term Memory Recurrent Neural Networks , 2016 .