关键词: 板式固体氧化物燃料电池, 广义预测控制, NARIMAX模型, Levenberg-Marquardt算法

Abstract: The working condition of solid oxide fuel cell is a chemical reaction process of high-temperature，high-rate change. The stability of gas inlet temperature directly affects fuel utilization and cell efficiency. This paper presents an improved external input of nonlinear autoregressive integrated moving average model which combines the ratio of fuel and air flow rates and the load current change to realize nonlinear generalized predictive control for the temperature of gas inlets，and uses the Levenberg-Marquardt algorithm based on nonlinear least squares to determine the parameters. The prediction values of real-time flow rates of fuel and air could be obtained accurately and quickly based on the combination of the model and transfer function model for the temperature control system，so that at higher temperature operation，the system could overcome parameter deviations due to changes caused by load current and retain the stability of output voltage.

Key words: plate solid oxide fuel cell, generalized predictive control(GPC), NARIMAX model, Levenberg-Marquardt algorithm