Industrial Electrical Energy Consumption Forecasting by using Temporal Convolutional Neural Networks

In conjunction with the 4th Industrial Revolution, many industries are implementing systems to collect data on energy consumption be able make informed decision scheduling processes and manufacturing in factories. Companies can now use this historical forecast expected for cost management. This research proposes of a Temporal Convolutional Neural Network (TCN) dilated causal convolutional layers perform forecasting instead conventional Long-Short Term Memory (LSTM) or Recurrent Networks (RNN) as TCN exhibit lower memory computational requirements. approach is also chosen due traditional regressive methods such Autoregressive Integrated Moving Average (ARIMA) fails capture non-linear patterns features multi-step time series data. paper, electrical factory will forecasted by extract complex daily limited dataset. The neural network built using Keras TensorFlow libraries Python. training provided GoAutomate Sdn Bhd. Then, economic factors indexes KLCI included alongside determine effects economy industrial consumption. results without then compared evaluated Weighted Percentage Error (WAPE) Mean Absolute (MAPE) metrics. parameters fined tuned accordingly based accuracy error create CNN WAPE = 0.083 & MAPE 0.092, one (1) week ahead small scale dataset only 427 points, has determined that index Bursa Malaysia no meaningful impact applied factory.