Short-Term Heat Demand Prediction Using Deep Learning for Decentralized Power-to-Heat Solutions

Coupling power and heat sectors will contribute to the integration of renewable energy but also promotes decarbonization of heat sector. Better heat demand forecasting is required for reducing the costs and increasing the energy efficiency. The traditional forecasting methods faces difficulties due to the irregularities in data, computational costs and generalizing the output. Application of machine learning based data-driven techniques provides better accuracy in handling data with nonlinear characteristics. This paper proposes the application of Long Short Term Memory (LSTM) based model to forecast the heat energy consumption of a pool of residential consumers considering the real data set. The prediction accuracy is improved by using wavelet transform for data preprocessing and the proposed model gives a better performance.