Aiming at the problem that the rural load distribution is uneven, the diversity is strong, and it is difficult to predict accurately, on the basis of considering the factors of the rural development planning, economy, population and other factors, the medium-long-term load forecasting method of a rural area based on cellular automata technology and machine learning is proposed. Firstly, according to the use and characteristics of the rural electricity load, the rural load is classified. Then, according to the nature of land use in the power supply scope of station area and the distribution characteristics of the load on the block, the function of the rural land is defined, and the least squares method is used to obtain the historical load density curve of different functions. On this basis, combined with the factors of the rural development planning, economic and natural conditions, etc., using historical data to train the gated loop unit load density model, and use the cellular automata technology predicts change information of the land block. Again, based on change information of the land block and historical load density curve, the gated loop unit network model is used to predict the load density, and then the rural medium and long term load forecast results are obtained. Finally, taking a rural area as an example, the feasibility and effectiveness of the proposed strategy are verified.