Thermal Unit Output Condition Recognition Based on Coal Blending Strategy: A Pre-trained Model Enhanced Method

Thermal power units, primarily coal-based, act as the "ballast" in China’s power system, supplying over 60% of total electricity generation even amid the accelerated construction of a new power system. However, as renewable energy integration introduces increased grid dispatch volatility, the operational flexibility of coal-fired units faces new challenges. This paper proposes a unit output condition recognition method using coal blending strategy as input to assist managers in quickly and accurately adjusting the coal blending strategy. We designed the Imitator to enable effective feature extraction and expansion of the CBS and supplementary unit parameters into a format suitable for pre-trained image classification models. Subsequently, a pre-trained image classification model is used to complete the unit output condition recognition task. A case study was conducted using real operational data from a coal-fired unit, and results show that all ResNet models achieved training accuracy above 95%, indicating high feasibility of the proposed method.