Short-term Forecasting of Electricity Consumption in Buildings for Efficient and Optimal Distributed Energy Management

The electricity consumption profile of buildings are different from the typical load curves that represent the electricity consumption of large systems at the national or regional level. The electricity demand in buildings is many times lower than the region- or nation-wide demands. It is also much more volatile and stochastic, meaning that the conventional tools are not effective enough for straightforward application at a building demand level. In this paper, an integrated approach consisting of Hilbert-Huang Transform (HHT), Regrouping Particle Swarm Optimization (RegPSO) and Adaptive Neuro-Fuzzy Inference System (ANFIS) is devised for 24h-ahead prediction of electric power consumption in buildings. The forecasts are used as input information for smart decisions of distributed energy management systems that control the optimal bidding and scheduling of energy resources for building energy communities. The effectiveness of the proposed forecasting approach is demonstrated using actual electricity demand data from various buildings in the Otaniemi area of Espoo, Finland. The prediction performance of the proposed approach for various building types (energy customer clusters), has been examined and statistical comparisons are presented. The prediction results are presented for future days with a one- hour time interval. The validation results demonstrate that the approach is able to forecast the buildings' electricity demand with smaller error, outperforming five other approaches, and in reasonably short computation times.