The invention, which belongs to the converged communication network service prediction field for intelligent electricity use, provides a converged communication network service prediction method for intelligent electricity use. According to a recursive-multi-layer-sensing-network-based converged communication network service prediction method, the recursive multi-layer sensing network is a recurrent neural network; a converged communication network service for intelligent electricity use by using a multi-input multi-output recursive multi-layer sensing network method. With the method and the model, the converged communication network service for intelligent electricity use is predicted accurately and is tracked dynamically; the convergence of the recursive multi-layer sensing network is accelerated and the training performance of the network is improved by using DEKF, multi-stream training and BPTT methods.