Research of time-varying performance of solar distributed thermal-power plant with neutral network prediction

Solar energy is potential in the future, but until now its commercial application is limited due to its high price. Distributed solar thermal-power plant provides electricity and heat simultaneously to the nearby users. It is competitive with traditional power plant due to its improved efficiency and low price. In this paper, a unique distributed solar thermal-power plant is designed. It is located in Nanjing Chemical Industry Park in China. Its reliability and advantage are studied through theoretical simulation. The theoretical model consists of 6 modules: solar radiation, solar collectors, thermal energy storage tanks, heat hub, bioreactors and organic rankine cycle generator. The building and bioreactors play the roles of heat and electricity users, respectively. However, both the heat source and users are time-varying, which are especially harmful to the system due to its small scale. Therefore, a thermal energy storage system and a heat hub are applied to solve the problems. Furthermore, the research focuses on the contradiction between the time-varying load requirement and the real-time response of the heat hub. It is solved by applying neural network to predict the load requirement, which enables the system to work with proper operation parameters in real time. The probability model of prediction error is built to test the system reliability with Monte Carlo method. The simulation results show how the distributed solar thermal-power plant working with neural network satisfies the time-varying load requirement. While the load requirement varies from 4472 to 21638 kW during one day, the heat hub responds with heat transfer capacity from 5552 to 22272 kW. In most of the time, the temperature deviation rate of heat hub after optimization is lower than 0.03, and the corresponding bias due to prediction error is lower than 0.01.