Energy forecasting on loads plays the vital role for the energy suppliers and the customers as it allows them for building an effective plan to fulfil demands. In energy industry, decision-makers and specialists provide reliable estimation of future energy demand/load at the aggregate and particular site levels which are crucial. Through this paper, we provide a unique model TransNext and have conducted extensive experiments on UCI Power consumption dataset. Our model outperforms other SOTA models and achieves a RMSE of 0.1881, MAE of 0.07678 and MAPE of 0.15716.

Keywords – transformers, CNN, positional encoding, multi head attention, load forecasting