Using the former actual line current load operation value as the training sample, the improved four layer neural network group model is put forward to predict the future current load value. For the problem that BP neural network is easy to converge to a local extremum, automatic adjusting momentum is applied. To enhance the ability in forecasting the load changing a lot in different months, the BP network group structure is put forward. Data simulation results show that the algorithm has high accuracy and can effectively estimate the current load change trend of the next time. The speed of prediction can meet the requirements of practical application. This model can be used as a large data analysis model for monitoring the change of the power load of the key units, and the early proposal is promptly proposed to power supply unit to take the corresponding measures. This model can also promote the construction of smart grid.