Cash Flow Forecasting Model for Electricity Sale Based on Deep Recurrent Neural Network

Abstract

Daily cash flow forecasting plays a very important role in enterprise development planning and strategic deployment. This paper makes use of a deep recurrent neural network model and applies it to the forecast of daily sales cash flow. This model adopts GRU unit structure. Through analyzing and mining historical payment flow data, the neural network model is used to automatically learn and extract the internal characteristics of information, and finally the daily cash flow prediction results are obtained. This method is the first successful application of artificial intelligence algorithm in the daily cash flow prediction of power grid. Experimental results show that the model is more accurate than ARIMA method.