Artificial Neural Network-Based Model Predictive Control Using Correlated Data

Abstract

This work addresses the problem of implementing model predictive control (MPC) in situations where the training data available for modeling contains possible correlations, and an artificial neural network (ANN)-based model is being used. In particular, we consider a problem where data sets are collected from a process that operates under the closed-loop condition in which correlations are induced between several input and output variables. In this situation, if the correlation problem is not addressed, manipulated inputs (calculated by MPC without considering the specific correlation in the input space) and independently prescribed set-points may require predictions in regions where the model is not trained, resulting in a poor closed-loop performance. To address this issue, principal component analysis (PCA)-based strategies are applied to both the input and output spaces in a way that maintains model validity. To that end, a new constraint on the squared prediction error (SPE) is incorporated into the ANN-based MPC optimization problem to make control actions follow the PCA model built using the training input data. Next, a PCA model is developed using the training output data, and then an optimization problem subject to the SPE constraint is defined to calculate set-points which are achievable. The effectiveness of the proposed ANN-based MPC to track these set-points is demonstrated using a chemical reactor example. Finally, a new autoencoder-based strategy is proposed to compute the achievable set-points. This is performed by replacing the PCA-based constraint with the autoencoder-based constraint in the optimization problem to calculate the set-points. The results indicate that the ANN-based MPC performance is improved when the autoencoder-based set-points are used.

Deep Learning Model Predictive Control Frameworks: Application to a Fluid Catalytic Cracker–Fractionator Process

“…27 In particular, novel approaches that consider correlated data (obtained from closed-loop operation) have been developed. 28,29 Past studies investigated the development of (feedforward) neural network-based control algorithms for FCC operation using generally small-scale models (i.e., models with a limited number of controlled and manipulated variables). 30−32 software/hardware.…”

“…The resulting reduced order models (dynamic and/or static) can be integrated into the process decision making structure (e.g., APC), leading to problems that are computationally tractable in practical amounts of time. − Recently, deep learning has been incorporated into APC (MPC) leading to frameworks that take into account uncertainties , or substitute the decision-making structure . In particular, novel approaches that consider correlated data (obtained from closed-loop operation) have been developed. , …”

Multi‐stage fusion regression network for quality prediction of batch process

“…[35] Hassanpour et al used an autoencoder (AE) to calculate reachable set points in model predictive control, solving the problem of nonlinear data that conventional linear methods cannot handle. [36] However, the above methods are only a single consideration of input features or quality-related features. In fact, both of them are important for the construction of the quality prediction model, and the loss of one of them may affect the quality prediction performance.…”

“…[ 35 ] Hassanpour et al used an autoencoder (AE) to calculate reachable set points in model predictive control, solving the problem of nonlinear data that conventional linear methods cannot handle. [ 36 ]…”

A hybrid machine learning approach integrating recurrent neural networks with subspace identification for modelling HVAC systems

“…Among several machine learning techniques, recurrent neural networks (RNN) have widely been used for dynamic modelling due to their capabilities to capture non‐linearities, [ 21–26 ] in general. RNN models have also been used for dynamic modelling of HVAC systems, in particular.…”