Residential Energy Demand Prediction Algorithm Based on Data-Driven Conditional Hidden Markov Model

Residential energy demand prediction plays an important role in achieving efficient and reliable operation of the power system. Effective demand prediction is necessary for providing a decision for user-side energy management and helping electric utilities balance power generation and consumption as well as developing optimal energy allocation schemes. To better capture the uncertainty of residential energy demand, this paper proposes a kind of conditional hidden Markov model (CHMM) based energy demand prediction algorithm to predict the energy demand of multi-residents under time-varying conditions. The proposed algorithm first clusters the states of CHMM by K-means. Then it adopts regression models to automatically estimate the transition probability and emission probability of CHMM directly from historical energy demand data. Finally, it predicts the energy demand by state transferring. The simulation results show that the proposed algorithm has high prediction accuracy in short-term forecasting compared with other prediction algorithms.