Power equipment failures have become the main reasonss for user power outages and emergency repairs when there are natural disasters. The health and resilience of power equipment are closely related to their long-term operating environment. Assessing their health level is of great significance in improving equipment reliability. This paper takes the island distribution network with the most obvious natural disaster characteristics and the drainage line with the highest fault frequency as the object. Given its operating conditions under the influence of multiple factors and strong correlation environmental characteristics, we propose a method for assessing the reliability of drainage lines based on an improved integrated neural network with double optimization. First, the maximal information coefficient (MIC) algorithm is used to filter the fault characteristics to form an historical fault data set. Secondly, to extract the fault rules of drainage lines in the island micro-weather environment, principal component analysis (PCA) combined with K-means is used to divide the fault area. Then an electrical-environmental coupling integrated neural network prediction model is constructed for life prediction by correlating island weather characteristics and drainage line fault characteristics, and is improved by an attention mechanism and dual optimization system to highlight the impact of typical island weather factors and improve prediction accuracy. Finally, based on the actual data of Zhoushan, Zhejiang Province, a calculation example is given. The results show that the proposed method can effectively assess the life of the drainage line of an island distribution network, and has important practical value for improving the reliability of distribution network equipment. This work is supported by the Innovation Action Plan of Science and Technology Commission of Shanghai Municipality (No. 8DZ1203200).