Method and system of producing hydrocarbons using data-driven inferred production

A method of predicting hydrocarbon production from one or more artificial lift wells is disclosed. Test data is obtained from the artificial lift well. A decline curve model, representing well performance, is generated for one or more fluids in the artificial lift well. Measurement values are obtained from an artificial lift operation. For each of the obtained measurement values, a measurement model is generated that correlates the measurement values to the decline curve. A Kalman filter is used to predict production outputs of at least one of oil, gas, and water for the well, and to generate an uncertainty range for the predicted production outputs. The Kalman filter uses the decline curves to predict the production outputs, and uses the measurement models to correct and/or update the predicted production outputs. Hydrocarbon production activities are modified using the corrected and/or updated predicted production outputs.