The power dispatching of a solid oxide fuel cell (SOFC)/gas turbine (GT)-based cogeneration system, which is a parallel combination of natural gas-to-power and coal-to-power units, is investigated. To address the economic dispatch (ED) problem, the AI-based forecasting model such as a Gaussian process regression (GPR) and an integration of GPR and neural networks (NN) models is employed to predict natural gas/coal prices and load demand. The economic dispatch (ED) optimization algorithm for minimizing total operating costs of the cogeneration system is employed to determine the monthly inlet flowrates of natural gas and coal. Since the forecasting errors for purchasing natural gas/coal are inevitable and the high carbon tax is needed to meet the 2015 Paris Agreement, carbon tax with US$60/ton CO₂ and ±10% forecasting errors of purchasing monthly fuel prices are taken into consideration. The flexible fuel purchasing strategies (FFPS) with 1024 approaches are aided to improve the power dispatch performance. It is validated that the worst approach of FFPS ensures the total profit improvement (TPI) with 0.30% at least and the best approach of FFPS increases the TPI to 7.33%.

研究了基于固体氧化物燃料电池（SOFC）/燃气轮机（GT）的热电联产系统的功率分配，该系统是天然气制动力和煤制动力的并联组合。为了解决经济调度（ED）问题，采用了基于AI的预测模型，例如高斯过程回归（GPR）以及GPR和神经网络（NN）模型的集成来预测天然气/煤炭价格和负荷需求。采用经济调度（ED）优化算法来最小化热电联产系统的总运营成本，以确定天然气和煤炭的每月进口流量。由于购买天然气/煤炭的预测误差不可避免，并且需要高碳税才能满足2015年《巴黎协定》的要求，因此每吨CO _2需缴纳60美元的碳税并考虑了购买每月燃油价格的±10％的预测误差。具有1024种方法的灵活燃料购买策略（FFPS）有助于改善动力分配性能。事实证明，最差的FFPS方法至少可确保总利润提高（TPI）为0.30％，而最佳的FFPS方法可将TPI提高到7.33％。