MEDIUM-TERM RESIDENTIAL ENERGY CONSUMPTION PREDICTION USING TCN AND TCN-LSTM MODELS

Abstract

: Prediction of electricity consumption is important for energy management and efficient use of energy resources. Accurate prediction of electricity consumption provides benefits to consumers and distribution companies in terms of correct management of electricity demand, electricity usage cost and waste control. This article proposes Temporal Convolutional Network (TCN) and Temporal Convolutional Network with Long Short-Term Memory (TCN-LSTM) models for medium-term electricity consumption prediction with an open access dataset in individual household electricity consumption prediction. An individual household’s 5000-hour electricity consumption prediction is made with the proposed models. The performance of the proposed models is compared with different prediction algorithms. The performance of all models used in the study is evaluated with three evaluation metrics commonly used in prediction model performance evaluation. When the prediction performances of TCN and TCN-LSTM models are compared with the performances of other prediction algorithms, it is seen that they have lower error rates and are more successful in medium-term electricity consumption prediction.