The current low voltage(LV) in the distribution network is becoming more and more serious, which seriously affects the daily life of residents, and the feedback of maintenance work orders is vague and unable to accurately locate the cause. In order to solve this problem, a LV cause analysis model based on multi-task assisted learning is proposed in the paper. Firstly, raw data such as current and voltage at 96 points of the LV?users are obtained, and the pre-processing of raw data is achieved. Secondly, the deep features of the data are mined by using bidirectional gated neural network (BiGRU), at the same time, the analysis of the main cause of LV is set as the main task, and the analysis of sub-causes is set as the auxiliary tasks, and which are used to strengthen the learning of hidden features in the data and provide additional supervisory information for the main task. Multi-task joint training is used to train the main cause analysis model, assist the model to learn more robust feature representations and improve the accuracy of LV cause analysis. The experimental results show that the LV causal analysis model based on multi-task assisted learning proposed in this paper has better analysis and localization ability, and the final classification accuracy can reach 95.58%.