Abstract: to quantify the randomness of equipment failures，while simplified probabilistic models fail to accurately characterize multi-equipment coupling effects and cascading failure propagation characteristics. To address this issue，this paper proposes an operational condition monitoring method for large-scale coal-fired units based on Monte Carlo simulation. A terminal sensing layer is constructed using multiple types of wireless sensors. Based on the acquired data，a joint probability distribution model is established via kernel density estimation. An adaptive weight factor derived from the historical mean of failures is introduced to optimize the sampling strategy. Furthermore，a topology-decaying Markov transition matrix is employed to quantify the risk of equipment cascading failures. The convergence of the simulation is dynamically assessed using a state entropy index. On the basis of the simulation results, the system failure probability and sensitivity are calculated，and a comprehensive risk assessment index is output. Experimental results conducted on a 600MW supercritical unit demonstrate that the proposed method achieves a transmission delay of no more than 5 seconds under load mutation conditions, and significantly outperforms comparative methods in terms of operational condition false alarm rate under low-load conditions，thereby providing a highly reliable monitoring approach for the peak-shaving operation of coal-fired units.