Fast generalized reduced gradient algorithm based data reconciliation model

More and more industrial production companies apply computer to process control, operation optimization and performance evaluation, which significantly increases the amount of data collected. Accurate measurement data can provide solid foundation for monitoring, optimization, scheduling, and decision analysis. However, measurement data is inevitably interfered by errors from multiple processes. This kind of interference makes measurement data deviate from real value and cannot meet some of the conservation laws and process constraints, which can make the production performance severely deteriorated. This paper proposed a fast Generalized Reduced Gradient (GRG) algorithm based data reconciliation model, which focuses on nonlinear data reconciliation problems. This method is on the basis of GRG algorithm, which can stably converge close enough to the global optimal solution, and Particle Swarm Optimization (PSO) algorithm is used in the early stage to accelerate the convergence. Iteration step size and selection of base variables are also optimized to accelerate and improve GRG. The proposed method can reduce the computation time under the premise of ensuring accuracy. Experiments on actual industrial data showed that the proposed method could solve the data reconciliation problem efficiently to provide effective data support for production scheduling.