To address theoretical challenges and accuracy limitations in estimating line losses for low-voltage substations ,arising from complex transmission lines ,multiple users ,and data acquisition difficulties ,we devised an innovative calculation approach in this study. Our method merges an enhanced K-means+ + algorithm with an Elman neural network. We initially conducted an in-depth analysis of factors influencing line losses in low-voltage substations and identified key indicators through correlation analysis. Employing principal component analysis (PCA) ,we reduced data dimensionality and complexity. Utilizing an enhanced K-means+ + algorithm ,we efficiently clustered the dataset and optimized model training. Integration of particleswarm optimization algorithmsfurtherboosted the Elman neuralnetworks,performance. Simulation verification using actual data affirmed the method,s superior performance in training efficiency and computationalaccuracy.