CPU Computation Influence on Energy Consumption Forecasting Activities of a Building

Abstract

Scheduling forecasting activities and improving the forecasting accuracy is important to deliver energy efficiency to the customers. However, it is also important to reduce the computational effort dedicated to these forecasting activities to ensure more effective environment sustainability. This paper proposes two forecasting algorithms known as artificial neural networks and k-nearest neighbors to anticipate energy patterns of a building monitoring data from five-to-five minutes. Using a case study with an annual historic and one week test, different scenarios are defined to test the forecasting activities with both higher and lower computational effort. It is achieved to ensure energy predictions with above reasonable accuracies evaluations while decreasing the computational effort, and the respective energy consumption, dedicated to forecasting activities.