Oil well daily output prediction method based on time sequence method

The invention relates to the field of petroleum production, in particular to an oil well daily output prediction method based on a time sequence method, which comprises the following steps: acquiring historical data influencing oil well output, and preprocessing the data; training the historical oil and gas daily output data by using SMA, EMA, ARIMA and LSTM models; carrying out autocorrelation test and white noise test on the SMA model, the EMA model, the ARIMA model and the LSTM model, and judging model quality; and constructing an elastic network regression model, taking prediction results of SMA, EMA, ARIMA and LSTM models as input features, and performing optimization by using an input feature matrix and a target variable to obtain an optimal regression coefficient. According to the method, the problem that the neural network ignores the variation trend of the yield along with time and the correlation between data is solved. 本发明涉及石油生产领域，尤其涉及一种基于时间序列方法的油井日产量预测方法，包括采集影响油井产量的历史数据，并对数据进行预处理；利用SMA、EMA、ARIMA和L