Development of a Fast Operation Algorithm of a Small-Scale Fuel Cell System With Solar Reforming

Abstract

The small-scale bioethanol steam reforming system (FBSR), using sunlight applied to a heat source, is a very clean method, which can supply fuel to a fuel cell. However, it is difficult to analyze the operation planning of this system with high precision. If such an analytical algorithm is developed, the optimum operation of this system will be realized by the command of the control device. However, the difficulty of weather forecasts, such as solar radiation and outside-air-temperature, to date has made it difficult to achieve rapid and highly precise results and to analyze the system operation. In this paper, an algorithm, which analyzes the operation planning of the FBSR on arbitrary days, is developed using the neural network. The weather pattern for the past 1 year is input into this algorithm, and the operation planning of the FBSR, based on the same weather pattern, is given as a training signal. In this paper, the operation results of the system obtained via genetic algorithm (GA) were used as the training signal for the neural network. Operation planning (the amount of hydrogen production and the amount of exhaust heat storage) of the system on arbitrary days could be obtained rapidly by ensuring that input data (the weather and energy-demand patterns) are channeled into the learned neural network following this study. Moreover, in order to investigate the accuracy of the operational analysis via the proposed algorithm, it is compared with the analysis result of operation planning using the GA.