S. Joe Qin

Experimental study of economic model predictive control in building energy systems

This paper presents results from testing an economic model predictive control (EMPC) strategy in an office building located in Milwaukee, Wisconsin, USA. The control strategy includes an economic objective function that is designed to account for the combination of energy and demand costs under a time-of-use (TOU) rate structure. A min-max optimization problem is formulated and solved as a linear program. The tests were carried out with the controller sited in a remote location and with the loop being closed over the Internet. The results show that the EMPC strategy was successful at reducing energy costs compared to the baseline case for the considered building.

Model predictive control of building energy systems with balanced model reduction

This paper presents a model reduction method based on balanced realization for thermal and power models of buildings. System identification is firstly performed to obtain high-order state-space models. The purpose of model reduction is to simplify the model structure while preserving the major input-output relations, so as to lower the computational cost in the subsequent model predictive control (MPC) scheme. An economic objective function is designed to minimize the energy and demand charges of building energy systems. The effectiveness of the presented method is shown by simulation, and it is shown that the control performance is not significantly affected by using reduced models.

Control performance monitoring via model residual assessment

Model quality is a key factor that affects the control performance of model predictive control. In this paper, a new closed-loop model assessment approach is proposed to assess model deficiency from routine closed-loop data. The proposed model quality index is a minimum variance benchmark for the model residuals obtainable from closed-loop data. From the feedback invariant principle the disturbance innovations at current instance are shown to be unaffected by the controller even if it is a nonlinear time-varying controller. Then it is shown that the disturbance innovations sequence can be estimated from closed loop data by an orthogonal projection of the current output onto the space spanned by past outputs, inputs or setpoints. With the disturbance innovations as the benchmark, a model quality index is developed by using the ratio of a quadratic form of model residuals and that of the estimated disturbance innovations. The effectiveness of the proposed methods is shown by simulation results.

An adaptive chaotic PSO for parameter optimization and feature extraction of LS-SVM based modelling

While training an LS-SVM model, two main challenges are parameter optimization and input feature extraction. The main purpose of this article is to address these two problems. Commonly used tools are PSO and BPSO, but they are not suitable for the optimization issues of many local optima owing to its random numbers used to update velocities. In this paper, an adaptive chaotic particle swarm optimization (cPSO) algorithm is proposed to enhance its global searching capability and local searching capability. The practicality of the proposed scheme is demonstrated by application to mineral process for the prediction models between production rate of the concentrated ore and the technical indexes. Compared with the original methods of grid search+PCA, GA+PCA, PSO+PCA as well as PSO+BPSO, the proposed strategy outperforms these existing methods in terms of convergence accuracy.

KPCA based multi-spectral segments feature extraction and GA based Combinatorial optimization for frequency spectrum data modeling

Mill load (ML) estimation plays a major role in improving the grinding production rate (GPR) and the product quality of the grinding process. The ML parameters, such as mineral to ball volume ratio (MBVR), pulp density (PD) and charge volume ratio (CVR), reflect the load inside the ball mill accurately. The relative amplitudes of the high-dimensional frequency spectrum of shell vibration signals contain the information about the ML parameters. In this paper, a kernel principal component analysis (KPCA) based multi-spectral segments feature extraction and genetic algorithm (GA) based Combinatorial optimization method is proposed to estimate the ML parameters. Spectral peak clustering algorithm based knowledge is first used to partition the spectrum into several segments with their physical meaning. Then, the spectral principal components (PCs) of different segments are extracted using KPCA. The candidate input features are serial combinated with mill power. At last, GA with Akaikes information criteria (AIC) is used to select the input features and the parameters for the least square-support vector machine (LS-SVM) simultaneously. Experimental results show that the proposed approach has higher accuracy and better predictive performance than the other normal approaches.

Improving industrial mpc performance with data-driven disturbance modeling

Industrial model predictive control (MPC) usually assumes a step-like disturbance model, which is insufficient when there is model mismatch in the plant or high order disturbances. In this paper, we demonstrate that a disturbance model identified from close-loop data is desirable for dynamic matrix control (DMC). We introduce a subspace based method to obtain such a model. The method estimates Markov parameters of the disturbance model using closed-loop data along with known input-output model information in the DMC controller. Simulation results are given to compare the proposed approach with traditional DMC.

Progressive Parametrization in Subspace Identification Models with finite horizons

Traditional subspace identification (SID) framework uses Kalman filter or predictor to interpret the SID models. To achieve this the horizons f, p have to approach infinity to be consistent. In practice, however, the horizons f, p are finite. We argue that for finite f and p the Kalman filter framework does not apply. In this paper, we introduce a progressive parametrization framework to interpret the models used in each step of SID methods and discuss how the progressively parametrized models lead to the recursive state space models, when additional assumptions are made. Monte- Carlo simulation is conducted on a closed-loop example to demonstrate what each step of SID contributes to the model estimate using the methods of HOARX, SSARX of Jansson [8], and that of canonical variate analysis [11]. We also state that the intermediate non-recursive models can be useful for the purpose of state estimation, fault detection, and control.

Concurrent projection to latent structures for output-relevant and input-relevant fault monitoring

When process faults occur, the process condition changes which is reflected in process variables. If these ab-normal variations are not properly annihilated in the process, poor product quality occurs as a consequence. This paper proposes a new concurrent projection to latent structures for the monitoring of output-relevant faults that affect the quality and input-relevant process faults that should be alarmed as well. The input and output data spaces are concurrently projected to five subspaces, a joint input-output subspace that captures covariations between input and output, an output-principal subspace, an output-residual subspace, an input-principal subspace, and an input-residual subspace. Process fault detection indices are developed based on the partition of subspaces for various types of fault detection alarms. The proposed monitoring method offers complete monitoring of faults that happen in the predictable output subspace and the unpredictable output residual subspace, as well as faults that affect the input spaces and could be incipient for the output. Numerical simulation examples are given to illustrate the effectiveness of the proposed methods.

Feature selection of frequency spectrum for modeling difficulty to measure process parameters

Some difficulty to measure process parameters can be obtained using the vibration and acoustical frequency spectra. The dimension of the frequency spectrum is very large. This poses a difficulty in selecting effective frequency band for modeling. In this paper, the partial least squares (PLS) algorithm is used to analyze the sensitivity of the frequency spectrum to these parameters. A sphere criterion is used to select different frequency bands from vibration and acoustical spectrum. The soft sensor model is constructed using the selected vibration and acoustical frequency band. The results show that the proposed approach has higher accuracy and better predictive performance than existing approaches.

Control performance monitoring of LP-MPC cascade systems

Traditional minimum variance control (MVC) based performance monitoring methods treats all controlled variables (CVs) the same (or with some preselected weights). However, due to the nature of soft CV constraint, CVs have priority in cascade systems of linear programming model predictive control (LP-MPC). It is desired to reduce violations of constraints for CVs at their upper or lower bounds and to keep CVs under control. In this paper, we introduce block lower triangular interactor matrix, based on which conditional MVC and corresponding performance benchmark is developed. We state that conditional MVC first consider CVs with multiple level priority and a subset of CVs in each level of priority. A simulation example is given to compare proposed method with traditional MVC methods.