Method and system for production optimization

Method for optimizing an industrial process, the method comprising the steps of:

• controlling the industrial process using an advanced process control system (APC), the APC system being provided with at least one first model, the at least one first model including a solution vector comprising a first matrix indicating dependencies of at least one control variable (CV) on at least one manipulated variable (MV) and a set of constraints, the step of controlling comprising:
• obtaining operational data of the industrial process at intervals of a second time period, the first time period being in the order of 1 to 24 hours, the operational data including values of the at least one control variable (CV);
• indicating if the value of the at least one control variable (CV) leaves a predetermined range as indicated in the set of constraints;
• and
• adjusting at least one manipulated variable (MV) in accordance with the at least one first model to keep the at least one control variable (CV) within the predetermined range;
• improving the advanced process control (APC) system using a production optimization (RTPO) system, the step of improving comprising:
• predicting production of the industrial process during a first time period using the first model, the first time period being at least a month;
• deriving an actual or measured production during the first time period from the operational data;
• comparing the predicted production with the measured production to provide a prediction error;
• if the prediction error exceeds a minimum threshold, checking the operational data over the first time period for correlations between the at least one MV, the at least one CV, and optionally at least one intermediate variable (IV);
• analysing the prediction error taking into account the correlations;
• filtering the correlations to select the most relevant correlations;
• creating a second matrix indicating dependencies of the at least one control variable (CV) on the at least one manipulated variable (MV);
• filtering the second matrix to indicate the correlations exceeding a minimum threshold of impact;
• and
• modifying the solution vector of the APC system based on the filtered correlations.