The value of power supply coal consumption of coal-fired power units is the main index to measure the level of energy saving and consumption reduction of the units. In this paper

big data analysis technology is used to analyze the historical data of energy consumption. Back-propagation(BP) neural network method is applied to respectively establish the sensitivity analysis models of power supply coal consumption at different load intervals. The impact evaluation factor(sensitivity coefficient) of operation controllable parameters to the power supply coal consumption and prediction ability of the model under different load conditions are calculated. The results show that

both the training and prediction accuracy of the model based on the sensitivity analysis method of BP neural network are within ±0.6%

and the calculation accuracy of the model is high. The sensitivity analysis coefficient of each operation controllable parameter to power supply coal consumption is different at different loads

but has certain regularity. In actual operation

the characteristic parameters with large sensitivity coefficient should be adjusted.