Abstract: This article provides analysis characteristics of finding feasible solution domain, heuristic strategies andalgorithm of Recurrent NN (Neural Network) for the ELSP problems，which can be described in basic time method. This method produces two kinds of decision variables, one is the continuous variables for the basic period of time, and the other is the integer variables for the multiple of time. In the design algorithm for solving the ELSP problem, the characteristic information of the feasible domain determines the effectiveness of the heuristic strategies. In order to solve ELSP problems, it must be required an effective the heuristic strategies. Our research results show thatthe characteristic of feasible solution domain improve the effectiveness of the heuristic strategies. Based on advantages of evolution computing of neural network, the paper designs algorithms for solving initial values of ESLP problem, which including evolving function, network structure, evolution function and strategies, so that constrained conditions of feasible solution domain is given. The paper suggested algorithms employed feasible domain conditions and heuristic strategies, comparing with HGA and traditonal GA methods for solving ESLP problem, The test results show that the effect of computing is significantly improved and the total cost decreases.