Application of deep learning techniques to minimize the cost of operation of a hybrid solar-biomass system in a multi-family building

Abstract

Concerns related to climate change put renewable energy at the centre of most of the policies aimed at achieving a deep decarbonisation of the building sector. The combined use of two or more renewable energy sources in the same energy system can lead to an increase in the total share of renewable energy and in the flexibility of the system. In this direction, the SolBio-Rev project aims to develop an innovative system that uses solar thermal collectors and a biomass boiler to meet energy demand in buildings in different climatic regions. An advanced control that used deep reinforcement learning techniques was considered in this paper to find an optimal control strategy for a specific SolBio-Rev system installed in a standard multi-family residential building located in Madrid. The advanced control was developed to minimize the total cost of operation of the system. The results indicated that the advanced control strategy achieved a cost reduction of 35% in winter, compared to a standard rule-based control strategy. However, the improved control was not able to produce a significant cost reduction in summer.

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 814945 (SolBio-Rev). This work is partially supported by ICREA under the ICREA Academia programme. The authors would like to thank the Departament de Recerca i Universitats of the Catalan Government for the quality accreditation given to their research group (2021 SGR 01615). GREiA is certified agent TECNIO in the category of technology developers from the Government of Catalonia. The authors would like to thank all partners of SolBio-Rev for the activities carried out within the project, and especially to CNR Institute for Advanced Energy Technologies (ITAE), AKOTEC, and Nationa Technical University of Athens (NTUA) for their support in developing some of the components models.