The safety risk factors in power transmission and transformation projects are complex and exhibit significant nonlinear characteristics. This complexity limits the adaptability of safety risk identification in power transmission and transformation projects

leading to decreased efficiency and large error tables. A safety risk assessment method for power transmission and transformation projects based on improved random forest (RF) is proposed. This method analyzes the entire process of power transmission and transformation projects in detail

identifies and quantifies risk factors in each link

and ensures the comprehensiveness of risk identification. Using techniques such as weighted weight calculation

binary encoding

and simulated annealing algorithm

feature selection and optimization are carried out to screen out the features that have the greatest impact on security risks. The optimal feature combination obtained from the above process is used as the supervised learning sample feature

and an improved random forest algorithm is employed for security risk identification. The improved random forest algorithm introduces an optimization mechanism for decision tree weights

which dynamically adjusts the weights of each tree based on its performance in out-of-bag (OOB) error evaluation to improve the overall prediction accuracy of the model. The experimental results show that the OOB error of the method is stable at around 0.062. The recall rate reaches 92% and the accuracy rate reaches 95%. The risk level determination time is between 6.4 s and 7.7 s

which can effectively control various safety risks in power transmission and transformation projects.