Design and Implementation of Scalable Power Load Forecasting System Based on Neural Networks

With the continuous advancement of science and technology, artificial intelligence has gradually been applied in various fields, and forecasting power and load with artificial intelligence technology is also widely used in the field of electric grids. At present, the mainstream load forecasting system mainly includes data collection, data processing, and forecasting using pre-trained models. Although in this way can complete the tasks of load forecasting, because the neural network model is pre-trained, it requires a lot of historical data, which cannot be satisfied in some scenarios. At the same time, the model structure cannot be dynamically modified and optimized after the training process and deployment is completed, and for the geographical expansion scenario, a model cannot adapt to the different climate, cultural and other factors in all regions, and must ask the system provider to retrain a new model. To solve the above problems, this paper proposes an adjustable load forecasting system based on artificial intelligence. With various ways of accessing user's data, users can customize a series of dimensions of the training process such as the type, scale, input, and output of the neural network model with their own data, so that they can dynamically adjusting the load forecasting model.