Combined Akima Interpolation and Deep Learning LSTM Method based Building Energy Efficiency Prediction

This paper deals on a Non-intrusive Load Monitoring (NILM) method proposed for building energy efficiency data analyzing and prediction. It is based on wireless smart sensors network deployed on optimal locations of the building. One challenge is to define the suitable location of the sensors in aim to limit the global cost of the system while ensuring sufficient data collection for a relevant analysis with a view to predicting the energy consumption of the building. Another difficulties are related to the differences between sensors in terms of time step of data acquisition. The proposed NILM method combine Akima Interpolation and Deep Learning LSTM algorithms, in aim to benefit from the performance of both strategies in a complementary manner. A network of several hundred wireless sensors was deployed on several buildings in the Paris region. The data collected was used to evaluate the proposed method. This paper presents its application to building heat pump energy consumption data, cross-referenced with measured climatic data.S