Technical and economic eva‎luation of the use of reinforcement learning method in optimizing the operation of an energy system

9/25/2025 12:00:00 AM

In recent years, with the increase in population in the world, on the one hand, the need for electricity production has increased, and on the other hand, fossil fuels are limited and produce pollutants. These conditions have made the discussion of simultaneous production more important than before due to its large efficiency. At first, the main and general definitions about hybrid systems and then a history of these types of systems, a review of the equipment and working methods of this system, as well as its optimization methods are presented. In this thesis, the technical, economic and environmental modeling of the integrated system of simultaneous electricity generation, heating, cooling and water consumption of residential units with hybrid equipment (renewable and non-renewable) has been done. The hybrid complex system (renewable and non-renewable) proposed for the production of electricity, heating, cooling and fresh water of isolated buildings (without connection to water, electricity and gas networks) was used in Sadaf Island. This system included photovoltaic (PV) panels, wind turbines, batteries and micro-CHP units to meet the electrical needs. Micro-CHP units and water heaters with LPG fuel and solar tubular collectors (ET) provided the necessary heating. A storage tank (HWST) was also planned for hot water storage. An electric chiller was provided for cooling and a reverse osmosis (RO) system was also considered for fresh water production. To optimize this system, the annual comparative profit (CAB) and the selection of optimal equipment according to the reduction of fossil fuel consumption were selected as two objective functions and the design variables were obtained. In fact, the objective function in this system and its optimization is related to reducing the economic cost of the system. The sensitivity analysis of the effects of changes in fuel and electricity prices as well as equipment investment costs on the optimal values of design parameters and objective functions was also investigated. In addition, to perform the modeling and optimization set faster, approximation functions in the artificial neural network (ANN) were used to estimate the electricity consumption of the chiller and RO reverse osmosis system. ANN had three input layers, eight intermediate layers and one output layer. The results showed that optimization with the help of innovative combined ANN and PSO methods reduced the execution time by about 10% to 15%. Using the DQN energy management method, by preheating the RO water, the annual electricity consumption of this equipment is reduced to 746 kilowatt hours (about 5%). Also, by choosing the proposed equipment of the hybrid system, 44% of the electric load and 45% of the heating load were produced by renewable systems and the rest by LPG fuel. which compared to the traditional system (based on the use of a gas-burning engine with LPG fuel to supply the electrical load of household electricity, electric chillers, RO systems and also a gas water heater to supply the heating load), about 70% saving in fuel consumption fossil and about 69.3% reduction in CO2 production.

انرژي خورشيدي , توربين بادي , سيستم هيبريدي(سيستم هاي تجديد پذير و غير قابل تجديد) , بهينه سازي به روش تجمع ذرات , شبكه عصبي مصنوعي , مديريت انرژي

solar energy , wind turbine , Hybrid system (renewable and non-renewable systems) , Optimization by PSO particle aggregation method , Artificial Neural Network (ANN) , DQN energy management

Mehdi Boyaghchi

Dr. Sanaye