Search

Integrated control and process design (ICPD) practices focus on an interaction between process and control design. The paper investigates ICPD design in circulating fluidized bed (CFB) power plants, which face increasing load change, efficiency and emission requirements. The state of ICPD research is examined and a classification of its methodologies is provided. The applicability of ICPD to large-scale CFB boilers is discussed for the first time based on this classification. Two ICPD case studies with a simple steam path mass storage model are presented for an industrial CFB boiler, with the aim of illustrating possibilities and challenges related to boiler ICPD. The steam mass storage amounts of the boiler superheating and evaporation sections are modified based on the dynamic relative gain array and closed-loop process optimization to generate processes with improved constant pressure mode output power setpoint tracking performance.

Oxy combustion has been introduced in power plants for facilitated CO2 capture and storage (CCS). Existing process controls for air-fired circulating fluidized bed (CFB) boilers are not necessarily optimal for oxy-firing. New degrees of freedom and alternative structures for control should be carefully analyzed for performance and flexibility. In this paper, the application of self-optimizing control (SOC) and partial relative gain (PRG) methods is presented for the early phases of control structure design. Controlled variable (CV) candidate sets are studied using the SOC concept. For the proposed CV sets, PRG analysis is used to reveal possible SISO loops with manipulated variables (MVs). Structures with promising results from both analyses were discovered.

The problem of curse-of-dimensionality in finite state and action Markov decision processes is considered using iterative clustering of closed-loop data and repeated discretization of the state space process model. The performance of the control design approach is demonstrated using a multi-input van der Vusse continuous stirred tank reactor control benchmark. It is demonstrated that the finite state description provides a simple implementation of Bayesian state estimation using cell filters, and dynamic programming gives means to conduct optimization of closed-loop performance of a nonlinear stochastic multidimensional chemical plant.

Model-based optimization tools for demand side management using supermarket refrigeration systems are considered. A model for medium-sized supermarket freezer and cooler temperatures is developed, as well as for the CO 2 compressor cycle power consumption. Model parameters are estimated from measured data. The formulation of an optimization problem vs. various electricity markets is briefly discussed, and an illustrative numerical simulation of demand side management optimization in a supermarket is provided. The developed model provides a feasible tool for including the supermarket thermal capacity dynamics into the optimization problem.

A wealth of control designs and experience are available for traditional air combustion in circulating fluidized bed (CFB) boilers. For the novel process of oxy combustion (for facilitated CO 2 capture) input gas compositions and flows can be adjusted independently, which decouples fluidization and oxygen carrying tasks and introduces new degrees of freedom and alternatives for control. The self-optimizing control approach (as formulated by Skogestad and colleagues in the 2000s) was used with steady-state approximations of a validated dynamic model for a pilot-size CFB combustor to study how the added degrees of freedom should be used. Instead of centralized online optimization of setpoints, self-optimizing control searches for a set of controlled variables which can be kept at constant setpoints despite disturbances and measurement errors, resulting in performance with acceptable steady-state loss. Results for air firing support method validity by suggesting the common practice in control; keeping power, flue gas O 2 and primary air/fuel feed ratio constant. For oxy firing, various control structures could satisfactorily compensate for studied disturbances and errors. Results suggest direct oxidant O 2 % control or simpler feed-forward solutions in line with current industrial CFB control, or alternatively using the added degrees of freedom for controlling variables such as furnace temperatures. Differences in, e.g. controllability, dynamic performance and implementation cost are relevant in further studies. The results serve as a first step in oxy-CFB control studies, suggesting candidate structures for dynamic analysis.

Oxy combustion in circulating fluidized bed (CFB) boilers was investigated in this paper. Oxy combustion is a carbon capture and storage technology, which uses oxygen and recirculated flue gas (RFG) instead of air as an oxidant. Air and oxy combustion were compared through physical considerations and simulations, focusing on process dynamics, transients and control. The oxidant specific heat capacity and density are elevated in oxy combustion, which leads to slower temperature dynamics. Flue gas recirculation introduces internal feedback dynamics to the process. The possibility to adjust the RFG and oxygen flows separately gives an additional degree of freedom for control. In the simulations, “direct” and “sequenced” switches between air- and oxy-firing were compared. Fast “direct” switches with simultaneous ramping of all inputs should be preferred due to the resulting smooth temperature responses. If these process input changes are unfeasible, the fuel should be altered after the gaseous flows (“sequenced” method).

The choice of a pump schedule has a great influence on the cost efficiency of a system, when the energy consumption charge changes during the day. The pump scheduling problem for a sub-system of a regional water network is considered, and the formulation of the problem as a MDP is given. The system under stochastic demand is modelled and an optimal controller is designed. Two implementations are considered: a direct application of MDP control actions, and local PI-controllers with a feedforward part. The experimental section shows computer simulations illustrating the considered approaches.

Energy efficiency is one of the main aspects in the Kraft pulping process. In Kraft pulping, the delignification of the pulp is mainly controlled by temperature and alkali. However, the quality is usually measured only in the blowline of the digester resulting in a control delay several hours. This affects to the quality control and also the energy efficiency of the plant. In this study, energy efficiency in the continuous Downflow Lo-Solids ® cooking application is studied using model-based approach. Temperature and alkali controls are considered and some suggestions are proposed.

The choice of a pump schedule has a great influence on the cost efficiency of a system, when the energy consumption charge changes during the day. The pump scheduling problem for a simplified regional water network is considered, and the formulation of the problem as a Markov decision process is given. The experimental section shows computer simulations and compares the results with a novel evolutionary approach based on neutral genetic algorithm techniques.

Cognition and control processes share many similar characteristics, and decisionmaking and learning under the paradigm of multiple models has increasingly gained attention in both fields. The controlled finite Markov chain (CFMC) approach enables to deal with a large variety of signals and systems with multivariable, nonlinear, and stochastic nature. In this article, adaptive control based on multiple models is considered. For a set of candidate plant models, CFMC models (and controllers) are constructed off-line. The outcomes of the CFMC models are compared with frequentist information obtained from on-line data. The best model (and controller) is chosen based on the Kullback–Leibler information. This approach to adaptive control emphasizes the use of physical models as the basis of reliable plant identification. Three series of simulations are conducted: to examine the performace of the developed Matlab-tools; to illustrate the approach in the control of a nonlinear nonminimum phase van der Vusse CSTR plant; and to examine the suggested model selection method for the adaptive control.

Polynomial artificial neural networks (PANN) have been shown to be powerful for forecasting nonlinear time series. The training time is small compared to the time used by other algorithms of artificial neural networks and the capacity to compute relations between the inputs and outputs represented by every term of the polynomial. In this paper a new structure of polynomial is presented that improves the performance of this type of network considering only non-integers exponents. The architecture adaptation uses genetic algorithm (GA) to find the optimal architecture for every example. Some examples of sunspots and chaotic time series are presented.

Plant-wide control in a circulating fluidized bed (CFB) power plant is considered. In particular, the self-optimizing control approach is used. The main contribution is applying the analysis in a pilot-sized CFB, which has not been reported before. This work focuses on conventional air-fired CFB combustion, for which a lot of experience on control design is already available. Therefore the results can be compared to practical knowledge from industrial practice. The work serves as an important step towards studying the control of oxy combustion in CFB, where the process is structurally similar but provides additional degrees of freedom for control design.

The goal of this paper is to design a new control algorithm for open-loop control of complex systems. This control approach is based on a genealogical decision tree for both regulation and tracking control problems. The idea behind this control strategy consists of associating Gaussian distributions to both the norms of the control actions and the tracking errors. This stochastic search model can be interpreted as a simple genetic particle evolution model with a natural birth and death interpretation. It converges on probability. A numerical example dealing with the control of a fluidized bed combustion power plant illustrates the feasibility and the performance of this control algorithm.

This paper considers identification of a multiple-input single-output Wiener model of a new type of superheater integrated to a CFB boiler. The static part is based on energy- and mass balances. Simultaneous estimation of model parameters in both static and dynamic parts was based on the industrial data from an 80 MW CFB boiler. The results indicate that the approach can be satisfactorily used to capture and describe the steam temperature in a wide range of variation in the input variables.

Identification and predictive control using a probabilistic CMC-ARX model is considered. The control problem is solved by maximizing the confidence in that the selected action results in a minimum cost. Bounds for the change in confidence of the optimal control action are derived, and a learning control algorithm is then suggested.

The increasing challenges in district heating operational optimization are briefly discussed. The paper describes the first steps in a research project on minimization of short term operational costs in a full scale district heating system. Based on a test model describing a part of a real district heating network, and a chosen approximate dynamic programming technique, simulations are used to illustrate and validate the fundamentals of the modelling and optimization approaches. It is concluded that the considered methods provide an adequate set of tools for the design of optimal network loading. The project is currently continuing with building of a more realistic dynamic model of the full-scale energy network and its components.

The identification of multiple-input single-output Wiener models is considered in this paper. The non-linear memoryless part is described by a parametrized steady state model. In this paper two representations for the linear dynamic part that preserve the unit steady state gain are discussed and compared: finite step response and transfer function with a feedback polynomial. The steps necessary for estimating the parameters of the Wiener model are presented. In the simulation examples with data from a pneumatic valve model, distillation column model and a pilot pump-valve system, the performance of the approach is examined, and the use of various kinds of non-linear black-box and grey-box structure for the modelling of the static part is illustrated.

This paper concerns process modelling using fuzzy neural networks. In distributed logic processors (DLP) the rule base is parameterised. The DLP derivatives required by gradient-based training methods are given, and the recursive prediction error method is used to adjust the model parameters. The power of the approach is illustrated with a modelling example where NOx-emission data from a full-scale fluidised-bed combustion district heating plant are used. The method presented in this paper is general, and can be applied to other complex processes as well.

This paper considers the use of adaptive learning systems for multimodal functions optimization and process control. The learning system collects and processes the available data to achieve the desired control objective. The environment where the automaton operates corresponds to the function to be optimized (the process to be controlled) which is assumed to be unknown function of a single parameter x. The admissible region of x (control variable) is quantized into N levels. These levels are associated with the actions of the automaton. The set of these actions is further decomposed into nonempty subsets. The action set is changing from instant to instant. At each time an action set is selected according to a probability distribution. The action and the action set probabilities are adjusted using learning algorithms (reinforcement schemes). This optimization and control approach is tested using two numerical examples (multimodal function and a chemical reactor). However, the concept can be applied to other examples as well. Simulation results illustrate the feasibility and the performance of this adaptive automaton.

Wiener type of models consist of linear dynamics followed by a static non-linear part. In this paper, a restricted class of Wiener models is considered where the static mapping represents a steady-state model for the process. A Wiener model structure is suggested for the identification of a MISO steady-state static model with linear FIR (finite impulse response) dynamics for each input. Unit steady-state gain is obtained by using a reduced FIR model, consisting of a unit gain plus FSR (finite step response) dynamics. The necessary derivatives required by gradient-based parameter estimation techniques are given. Simulation examples with data from a distillation column model and a pump-valve system, using sigmoid neural networks to model the non-linearities, illustrate the behaviour of the approach.

A new algorithm for training the parameters of a distributed logic processor is suggested, based on learning automata. A fuzzy inference system is implemented on a multilayer perceptron platform. Various possibilities for assigning learning automata are discussed and two assignment strategies are proposed. The behaviour of the resulting algorithm is illustrated using flue-gas NO/sub x/ emission data measured from an industrial-size fluidised-bed combustor.

This paper concerns the process modelling based on fuzzy logic neural networks. Fuzzy systems are implemented in the form of distributed logic processors. Derivatives required by gradient descent training methods are given, and recursive prediction error training method is used to adjust the model parameters. The approach is illustrated with a modelling example where nitrogen emission (x) data from a fluidized-bed combustion district heating plant is used. The method presented in this paper is general, and can be applied to other complex processes as well.

This paper concerns the adaptive selection of the optimal order of linear regression models using a variable-structure stochastic learning automaton. The Alaike criterion is derived for stationary and non-stationary cases, and it is shown that the optimal order minimizes a loss function corresponding to the evaluation of this criterion. The order of the regression model belongs to a finite set. Each order value is associated with an action of the automaton. The Bush-Mosteller reinforcement scheme with normalized automaton input is used to adjust the probability distribution. Simulation results illustrate the feasibility and performance of this model order selection approach

Artificial neural networks are studied from process engineer's point of yiew. In engineering problems, uses of neural networks can be found from modelling of non-linear systems. In on-line process modelling, the problem of measurement data that has a low information content cannot be ignored. An attempt to collect recursively an infonnationrich training data set is made using a self-organising feature map preceding a multi-layer perceptron artificial neural network. The combined configuration is introduced and a case study is presented, where models for the fluidized-bed combustor flue gas NO x -content are made using data collected from a 25 MW district heating thermal power plant

The adaptation made for the Polynomial Artificial Neural Networks (PANN) using not only integer exponentials but also fractional exponentials, have shown evidence of its better performance, especially, when it works with non-linear and chaotic time series. In this paper we show the comparison of the PANN improved model of fractional exponentials with a new logarithmic model. We show that this new model have even better performance than the last PANN improved model.

Bed fuel inventory is one of the key variables in the fluidized bed combustion process when describing the dynamic behaviour of combustors burning fuels like coal or graphite. Many of the essential features of fluidized bed combustor responses can be simulated with the dynamic model presented. The model consists of differential equations describing the dynamic behaviour of the furnace and steady-state calculations for the burning of char particles. The simulations show good agreement when compared to process measurements from a commercial-sized coal combustor.

Student's studies can be seen as consisting of chains or sequences of courses, the learning activities of a student along these lines make his study path. In this paper, methods for detection, analysis and control of learning activities are developed, based on students' credit record data. A central aim of the work is to support the supervision of students using currently available data bases.

Adaptive control based on multiple Markov transition models (MTM) is considered. MTM's are constructed off-line for a set of candidate plant models, and an optimal controller is designed for each plant model. Frequentist information obtained from measured data is compared on-line with the state transitions predicted by the plant models. The best model is selected based on Kullback-Leibler distances averaged over a time window. The manipulated input of the plant is obtained from the control policy associated with the selected model. Two simulations are conducted to illustrate the approach in adaptive (gain scheduling) control of a non-linear nonminimum phase van der Vusse continuous stirred tank reactor (CSTR). The first simulation concerns an ideal CSTR when the tempeature changes, the second simulation considers control of a non-ideal CSTR where streams in different temperatures are sampled randomly.

The increasing requirements for better quality with smaller variations are often difficult to fulfill in pulping plants. In Chemi-Thermo Mechanical Pulping (CTMP), the non-uniformity in the chemical treatment and the changes in the refiner's energy consumption often cause variations in pulp properties. The CTMP process has strong interactions between process variables and therefore necessitates decoupling of interactions in the refining control system. Together with the high variability in the process dynamics, lack of direct and reliable measurements makes it difficult to control the process. Consistency of the blow-line pulp is not always measured using an on-line sensor. If high consistency sensors are used, also redundant measurements or soft sensors may need to be developed for feedback control purposes. Pulp quality variables such as freeness and shive content are often analyzed infrequently and/or there is a long dead time and time constant between the refiner's blow line and the on-line analyzer's sensor. In this paper, pulp properties of CTMP refining are modeled and a new two-level quality control system is proposed. The nonlinear quality models are verified by simulation and compared with measurements from an existing on-line pulp quality analyzer. The two-level quality control system uses the process models for stabilizing the refiner conditions (pulp consistency and specific energy consumption) and for optimizing the process operation and the pulp quality (freeness). The stabilizing controllers are designed using multivariable Hammerstein and cascade control structures. The optimizing controller calculates the setpoint for the stabilizing controller by using the inverse of the nonlinear multivariable quality model. A simple scheme for adapting the freeness model based on the on-line measurements is presented. The performance of the control system is illustrated in simulations.

Publisher Summary Artificial intelligence is the science of intelligence and is connected to systems sciences through the basic assumptions of computational psychology. Computational intelligence includes fields such as neural networks, fuzzy systems, evolutionary computing, and artificial life. The various types of fuzzy neural networks can also be considered. Neuro-fuzzy systems are tools for engineering technical processes, having an interest in engineering applications. Models are needed for training the process operators and for the development of soft sensors. Standard modeling approaches include two main streams: the first-principle (white-box) approach and the identification of a parameterized black-box model. The first-principle approach denotes models based on laws about physical and chemical phenomena, derived on the basis of the first principles whereas in the black-box models, the model structure is a priori selected. Based on experimental data, an identification algorithm is used for estimating the unknown parameters. The neuro-fuzzy methods are most useful in the identification of nonlinear and poorly known processes when measurement data giving examples of the process behavior are available. The fuzzy features in the approach also make it useful when knowledge is available in the form of human experimental knowledge. The integration of fuzzy methods with optimization, systems identification, and parameter estimation techniques offers robust and efficient tools for the modeling of real-world industrial processes. This chapter illustrates a unified representation including logic processors (LP), and multilayer perceptron (MLP) neural networks. It discusses gradient-based parameter estimation techniques, learning-automata-based techniques, and identification of a nonlinear plant.

Wiener type of systems consist of linear dynamics followed by a static non-linear part. In this paper, a restricted class of Wiener systems is considered where the static mapping represents a steady-state model for the process. A Wiener model structure is suggested for the identification of MISO steady-state static systems with linear unit steady-state gain OE dynamics for each input. The derivatives required by gradient-based parameter estimation techniques are given. Example with MISO data from a pump-valve pilot plant, using sigmoid neural networks to model the non-linearities, illustrates the behaviour of the approach.

A new approach based on a genealogical decision tree is suggested for solving an open-loop tracking problem. The algorithm associates Gaussian distributions to both the norms of the control actions and the tracking errors. It solves the optimization problem sequentially, using random resampling from a population of solutions. This stochastic search model can be interpreted as a simple genetic particle evolution model with a natural birth and death interpretation. It converges in probability. Two numerical examples, dealing with rapid thermal processing and robotics, illustrate the feasibility and the performance of this control algorithm.

Controlled finite Markov chains are used for solving a multivariable control problem. The controlled system was the secondary air flow system in a fluidized-bed combustion power plant. A four-input four-output system control problem was formulated as a CFMC, and solved using dynamic programming. A non-linear system model, developed in earlier studies, was used in this model-based design. Three filters were applied to smoothen the (finite) control actions, for feasible plant control. These were based on averaging over finite state-space, references, and averaging over time. Simulations were conducted to illustrate the approach, and compare it with respect to SISO PID design. The simulations indicated that system control can be succesfully designed based on CFMC techniques, even for a 4×4 MIMO system.

Analysis and tuning of a circulating fluidized bed (CFB) hot-loop model are considered. Selected model outputs are fitted to measured data by allowing a set of constant-valued parameters to be considered as time-varying. These include the reaction rate, several heat transfer coefficients, and fuel moisture. Estimates of parameter distributions are created using particle filters (PF). PF provide a Bayesian approach utilizing both the system model and measured data. An illustrative example is given using data and model for a full scale plant.