Deep sequence to sequence Bi-LSTM neural networks for day-ahead peak load forecasting

•First attempt on using deep Bi-LSTM S2S regression for “peak” load forecasting.•First attempt on forecasting peak demand of a residential area supplied by FESCO.•Challenge of accurately forecasting load on special days is effectively addressed.•Model is effective for grossly limited training data.•Useful for developing countries with small total load, especially at feeder level.