A novel Prediction Method based on Improved Deep Mixture Density Network for Electricity Consumption, Photovoltaic Generation, and Net Demand of Smart Homes: Case Study for Sydney Metropolitan Area

Abstract

The electricity consumption in the smart grids consists of an uncertainty feature. Also, an unstable atmosphere situation causes photovoltaic (PV) generation will be undefined output. With both of these problems, the net demand power of consumers can’t be a specific value. On the other hand, using consumption patterns, the consumptions and generations could be predicted for improving the operation of the power system. This paper reports the results of the differential performance of probabilistic forecasting of the residential electricity onsumption, PV power generation, and net demand related to smart buildings using the novel method of the Improved Deep Mixture Density Network (IDMDN). According to this, investigators used a strong Multi to Multi (M2M) mapping of the neural network model. They followed that they had used a kind of beta kernel to decrease the number of leakage issues. It is an attempt to generate random predictions by the method of end-to-end. It expressed a new performance of changed initiation and multiple procedures of educating to decrease or remove unsteady traits in the probabilistic problems of the beta kernel function. The results show good performance of the proposed method in comparison with other methods.

Keywords

• smart buildings
• probabilistic forecasting
• photovoltaic
• net demand
• electricity consumption
• improved deep mixture density network