Volume 27: Closing Carbon Cycles – A Transformation Process Involving Technology, Economy, and Society: Part II

A Hybrid Method of Hourly Electricity Consumption Forecasting for Building Cluster Based on PSO-RF Xiaolin Chu, Peng Wang, Ruijuan Zhao, Dayong Lv

Abstract

Building energy consumption prediction is of great significance to realize intelligent decision-making of energy system and improve energy efficiency. A random forest (RF) prediction model optimized via the particle swarm optimization (PSO) algorithm is established to forecast the hourly electricity consumption of the building cluster consisting of interconnected multiple buildings. The accuracy, generalization and robustness are taken as evaluation indexes. In the case study, the building cluster located in Austin is adopted as an example to explore the predicted performance of the proposed PSO-RF model in different seasons. The results show that the hourly electricity consumption PSO-RF model of the building cluster can achieve highest accuracy, strongest generalization, and best robustness, compared with RF, decision tree (DT), XGBoost, and k-Nearest Neighbor (KNN) prediction models. Therefore, the proposed hybrid model can be used as a reliable tool for building cluster electricity consumption prediction and energy management.

Keywords Building cluster, electricity consumption, forecasting method, random forest, particle swarm optimization algorithm