Search

The flue gas temperature is an important parameter for safe and economic operation of thermal power plants. However, in high temperature areas, it is difficult to measure the temperature directly because the sensors are subject to flying ash erosion. This paper establishes a dynamic heat transfer model for the furnace and the semi-radiative heat exchangers, and proposes an estimation approach of the flue gas temperature in the boiler’s horizontal flue. The boiler-side heat absorption models, which cover the heat exchangers, the metal wall and the heat losses are established to calculate the heat absorption rate of the working fluid in the boiler. By solving the optimization problems which are developed based on the dynamic heat transfer models, the flue gas temperature at the outlet of the furnace and the semi-radiative heat exchangers are obtained. Pseudo-online simulations are performed with the history data of an ultra-supercritical unit to verify the approach. The relative errors between the calculated and the designed flue gas temperature with both designed and calibrated feed coal are found less than 3.5%. Additionally, the proportion of the convective heat in the total heat absorption for each semi-radiative heat exchanger is obtainable with the same approach.

The operation conditions of the cold-end system have significant impact on the unit thermal economy. However, due to the difficulty of online determination of some key parameters, the condenser pressure optimization has been a challenging task for a long time. This paper proposes an online applicable approach to optimize the condenser pressure with variable speed pumps, taking the mass flow rate of the circulating water as the manipulating variable, to achieve better thermal economy. After the exhaust steam wetness fraction is online identified, the condenser thermodynamic characteristics under varying working conditions are investigated based on the effectiveness and steady-state energy balance of the condenser. By maximizing the net power benefit, defined as the difference between the unit power increment and the pump power consumption increment, the optimal mass flow rate of the circulating water is derived. To validate the approach, pseudo-online simulations are conducted with the history data from an ultra-supercritical unit. The retention time during which the set value of the mass flow rate of the circulating water remains constant is studied in context with the implementation of the manipulation strategy under on-site scenario. Simulation results reveal the energy-saving potential of condenser pressure optimization with the proposed approach.

The load following of nuclear power plants (NPPs) has been a contentious issue in the control field. In this paper, we propose a chattering-free higher order sliding mode control scheme with a high-gain observer for the load following of a pressurized water reactor (PWR) in the presence of lumped disturbances owing to model uncertainties and external disturbances. The mathematical model of the PWR system is first set up in the form of an affine nonlinear equation. Subsequently, a high-gain observer, which achieves accurate estimations of the unmeasured state and lumped disturbances, is designed. Based on the outputs of the high-gain observer, we develop a chattering-free higher order sliding mode controller to improve the load-following performance while dealing with lumped disturbances and estimation errors of the high-gain observer. In contrast to some previous sliding mode controllers for the load following of NPPs, the proposed controller is completely free from chattering effects because the control input is obtained after integration. The asymptotic stability of the overall control scheme is demonstrated by combining the Lyapunov stability theory with backstepping technology. Finally, the simulation results reveal that the maximum absolute power error is less than 1 × 10 4 W with the proposed control scheme, 5 × 10 5 W with a PID controller, and 4 × 10 5 W with a conventional sliding mode controller. In addition, in contrast to the conventional sliding mode controller, the proposed control scheme produces smooth control input without the chattering phenomenon. Thus, the proposed chattering-free higher order sliding mode control scheme with a high-gain observer provides smoother control input, higher load-following accuracy, and stronger robustness against lumped disturbances than the PID controller and the conventional sliding mode controller.

Online performance monitoring is crucial for the safe and economical operation of thermal power plants. In this study, dynamic models of the whole process are established for the subcritical coal-fired power plant in terms of thermodynamic balances, which covers both boiler-side and turbine-side processes. Based on the whole process models, a monitoring platform is constructed with the functions of online estimation of key state variables and performance evaluation of the unit. The evaluation results of the unit efficiency with the time period of every 5 s may be provided. Model validations are carried out with the history operation data derived from the distributed control system (DCS) of a real 300 MW power plant. Comparisons between the simulation results and designed values of the unit demonstrate the high accuracy of the whole process models. The monitoring platform may also play the role as an offline simulator for the purpose of double check of the unit’s operation during specific time period and accidental cases.

Thermodynamic properties of the working fluid and the flue gas play an important role in the thermodynamic calculation for the boiler design and the operational optimization in power plants. In this study, a generic approach to online calculate the thermodynamic properties of the flue gas is proposed based on its composition estimation. It covers the full operation scope of the flue gas, including the two-phase state when the temperature becomes lower than the dew point. The composition of the flue gas is online estimated based on the routinely offline assays of the coal samples and the online measured oxygen mole fraction in the flue gas. The relative error of the proposed approach is found less than 1% when the standard data set of the dry and humid air and the typical flue gas is used for validation. Also, the sensitivity analysis of the individual component and the influence of the measurement error of the oxygen mole fraction on the thermodynamic properties of the flue gas are presented.

A complete mass transfer model was established for an internal loop airlift reactor. All sections of the reactor, including bottom, riser, top, and downcomer, were taken into account. A numerical method was developed to solve the mass transfer model. The effectiveness of the model was validated by experimental measurements. Based on numerical results of the model, the dynamical mass transfer process in the reactor was analyzed in detail. Some mass transfer characteristics of the airlift reactor were revealed. The results indicated that the proposed model well predicted the mass transfer in the internal loop airlift reactor.

Online performance monitoring of the steam turbine system is very important for the economic and safe operation of a power plant. For a subcritical steam turbine system, the wet saturated steam in the last few stages of the low pressure cylinder (LPC) increases the heat loss and threatens the unit safety. Unfortunately, online measurement of the exhaust steam wetness fraction of the LPC has been a difficult task for a long time. This paper proposes an online applicable approach based on two equivalent overall efficiency models to monitor the exhaust steam wetness fraction. Firstly, to estimate the total energy entering the steam turbine system, the lumped models of the evaporation system and the heat exchangers are established. Then two equivalent overall efficiency models of the steam turbine system are presented. Based on the same overall efficiency, the exhaust steam wetness fraction is therefore identified. The effectiveness of the proposed approach is validated through the comparison between the calculated and the designed operation data of a subcritical coal-fired unit. As the extension of the applications, the calculated wetness fraction is further used to estimate the thermal economic indices for the purpose of assessing the operation performance of the steam turbine system.

In this paper, the GTucker2 model is proposed for monitoring both even-length and uneven-length batch processes. The GTucker2 model has two prominent advantages. The first one is that it performs tensor decomposition on the three-way data array and thus avoids potential problems of information loss and “curse of dimensionality” induced by data unfolding. The second one is that it solves the uneven-length problem in a “natural” way without using batch trajectory synchronization, which prevents distorting data and fault patterns and guarantees higher modeling and monitoring precisions. An online batch process monitoring method is then developed by integrating GTucker2 with the moving data window technique. Three monitoring statistics named Q, R2, and T2 statistics are constructed for fault detection and diagnosis. The effectiveness and advantages of the GTucker2-based monitoring method are illustrated by two case studies in a benchmark fed-batch penicillin fermentation process.

A novel data projection algorithm named tensor global-local preserving projections (TGLPP) is proposed for three-way data arrays. TGLPP is able to preserve global and local structures of data simultaneously. TGLPP builds a unified framework for global structure preserving and local structure preserving. Tensor principal component analysis (TPCA) and tensor locality preserving projections (TLPP) are unified in the TGLPP framework. They are proved to be two limiting cases of TGLPP. A batch process monitoring method is then developed on the basis of TGLPP. The SPD statistic and R2 statistic are used for fault detection. The contribution plot is applied for fault identification. Two case studies are carried out to validate the efficiency of TGLPP-based monitoring method.

A novel method named tensor global–local structure analysis (TGLSA) is proposed for batch process monitoring. Different from principal component analysis (PCA) and locality preserving projections (LPP), TGLSA aims at preserving both global and local structures of data. Consequently, TGLSA has the ability to extract more meaningful information from data than PCA and LPP. Moreover, the tensor-based projection strategy makes TGLSA more applicable for the three-dimensional data than multiway-based methods, such as MPCA and MLPP. A TGLSA-based online monitoring approach is developed by combining TGLSA with a moving window technique. Two new statistics, i.e., SPD and R2 statistics, are constructed for fault detection and diagnosis. In particular, the R2 statistic is a novel monitoring statistic, which is proposed based on a support tensor domain description method. The effectiveness and advantages of the TGLSA-based monitoring approach are illustrated by a benchmark fed-batch penicillin fermentation process.

The flow regime transitions in an airlift reactor were investigated based on pressure fluctuation signals. Two time–frequency analysis methods, i.e., Wigner–Ville distribution and wavelet transform, were used to extract flow regime characteristics from pressure signals. The main frequency derived from the smoothed pseudo-Wigner–Ville distribution of the pressure signal was used to quantify flow regime transitions in the reactor. Two flow regime transition points were successfully detected from the evolution of main frequencies of pressure signals. In addition, the local dynamic characteristics of the pressure signal at different frequency bands were analyzed by use of the wavelet transform. A new flow regime identification method based on the wavelet entropy of the pressure signal was proposed. This method was confirmed to be reliable and efficient to detect flow regime transitions in the reactor.

The flow regimes and their transitions in an internal loop airlift reactor were investigated. The Hilbert–Huang transform (HHT) was applied to analyze the energy–frequency–time distribution of the pressure signal. It was found that the Hilbert spectrum of the pressure signal was closely related to the superficial gas velocity. The stochastic behaviors of intrinsic mode functions (IMFs) extracted from the pressure signal were studied using the Hurst analysis. Two different Hurst exponents were obtained for each pressure signal: one was smaller than 0.5, while the other was larger than 0.5, representing the anti-persistent and persistent hydrodynamic behaviors, respectively. The evolution of the larger Hurst exponent clearly indicated flow regime transitions in the downcomer. The wavelet transform combined with the autocorrelation analysis were applied to extract chaotic components of the pressure signal. Two flow regime transition points were successfully detected from the evolution of chaotic parameters, i.e. the largest Lyapunov exponent, correlation dimension and Kolmogorov entropy.

Higher order statistics and Wigner higher order moment spectra were used to extract useful flow regime characteristics from wall pressure fluctuation signals in an annulus sparged internal loop airlift reactor. It is found that the pressure fluctuation in the airlift reactor is a typical nonlinear and non-stationary process, which exhibits different frequency characteristics depending on flow regimes. Analysis methods based on bispectrum and Wigner trispectrum are powerful tools to reveal frequency characteristics of pressure signals. To identify flow regime transitions in the reactor, two new characteristic quantities, namely average bispectrum and generalized average frequency, are defined from bispectrum and Wigner trispectrum of the pressure signal, respectively. Two flow regime transition points corresponding to three flow regimes in the reactor are successfully detected by using these two characteristic quantities.

The profit function is the generic criterion to describe the cost effect of a batch process. To focus on the prediction of the profit function for 2-keto-L-gulonic acid (2-KGA) cultivation, which is potentially applicable for process monitoring and optimal scheduling, rolling learning-prediction (RLP) based on a support vector machine (SVM) is applied. The RLP implies that the SVM training database is rolling updated as the batch of current interest proceeds, and the SVM learning is then repeated for the prediction. The database is further updated after termination of a batch. The updating procedures are investigated in detail. Pseudo-online prediction is carried out using the data from industrial-scale 2-KGA cultivation under actual and hypothetical inoculation sequences. The results indicate that the average relative prediction error is less than 5 % in the later phase of fermentation in all inoculation sequences.

Structured packed columns are widely used in the chemical industry for distillation and absorption. However, the understanding of the transfer mechanism behind the counter-current gas-liquid flow in structured packed columns is still limited. In this work, a three-dimensional CFD model that considers the local absorption and the local momentum transfer mechanism is developed for a film flow on a small plate with a counter-current gas flow. The model, based on the Volume of Fluid (VOF) method, is built up on the basis of a pressure drop model and the penetration theory to quantitatively investigate the instantaneous hydrodynamics and mass transfer characteristics of the liquid phase. Simulations and experiments are carried out for a system consisting of propane and toluene. A comparison of the simulation results with the experimental data for the outlet concentrations shows good agreement.

An on-line application oriented scheduling approach for cultivation periods of penicillin fed-batch fermentation is proposed to maximize the economic profit of the whole workshop. The approach consists of profit function calculation, profit potential prediction, on-line batch classification and scheduling decision-making. The objective function is to maximize the total economic profit of those batches in parallel operation by repeated updating of the termination sequence while without significant disturbance on upstream and downstream work sections. Pseudo-on-line simulations of re-scheduling on industrial penicillin cultivations demonstrate the potential benefits of the approach. The original data is from a Chinese pharmaceutical company which consists of 58 industrial batches of penicillin fed-batch cultivation under empirical operation. Re-scheduling is carried out under three different inoculation sequences and the simulation results indicate that 2–3% of total profit increase could be generally achieved, compared with the empirical one.

A batch process monitoring method using tensor factorization, tensor locality preserving projections (TLPP), is proposed. In many existing vector-based methods on batch process monitoring such as MPCA and MLPP, a batch data is represented as a vector in high-dimensional space. But vectorizing batch data will lead to information loss. Essentially, a batch data is presented as a second order tensor, or a matrix. In this case, tensor factorization may be used to deal with the two-way batch data matrix directly instead of performing vectorizing procedure. Furthermore, tensor representation has some advantages such as low memory and storage requirements and less estimated parameters for normal operating condition (NOC) model. On the other hand, different from principal component analysis (PCA) which aims at preserving the global Euclidean structure of the data, the TLPP aims to preserve the local neighborhood information and to detect the intrinsic manifold structure of the data. Consequently, TLPP may be used to find more meaningful intrinsic information hidden in the observations. The effectiveness and advantages of the TLPP monitoring approach are tested with the data from a benchmark fed-batch penicillin fermentation and two industrial fermentation processes, penicillin and cephalosporin, respectively.

The design of packed columns requires the detailed description of the hydrodynamics on the surface of the packings. To analyze the local flow behavior of the liquid phase, a three-dimensional Computational Fluid Dynamics (CFD) model was developed that applies to the two-phase countercurrent flow on an inclined and flat plate. This model, based on the volume-of-fluid (VOF) method, considers the gravity, the surface tension and the drag force between the two phases. The development of such a model allows investigation of the influences of the liquid and gas flow rates on the flow behavior such as the film flow and the rivulet flow. A validation of the model was performed using data from the literature and from experiments conducted in this work. Simulation and experimental results demonstrate that the specific wetted area on the plate decreases with decreasing liquid load. Moreover, CFD simulations reveal that the presence of the countercurrent gas phase tends to increase the fluctuation and the thickness of the film flow, which is in accordance with experimental data. It also affects the flow behavior of the rivulet flow and changes the velocity profiles for both film and rivulet flow behavior. On the other hand, the simulation results indicate that CFD is a potent tool for analyzing and investigating the flow phenomena in chemical engineering.

An approach for multivariate statistical process control based on multiway locality preserving projections (LPP) is presented. The recently developed LPP is a linear dimensionality reduction technique for preserving the neighborhood structure of the data set. It is characterized by capturing the intrinsic structure of the observed data and finding more meaningful low-dimensional information hidden in the high-dimensional observations compared with PCA. In this study, LPP is used to extract the intrinsic geometrical structure of the process data. Hotelling’s T 2 ( D ) and the squared prediction error (SPE or Q ) statistic charts for on-line monitoring are then presented, and the contribution plots of these two statistical indices are used for fault diagnosis. Moreover, a moving window technique is used for the implementation of on-line monitoring. Case study was carried out with the data of industrial penicillin fed-batch cultivations. As a comparison, the results obtained with the MPCA are also presented. It is concluded that the Multiway LPP (MLPP) outperforms the conventional MPCA. Finally, the robustness of the MLPP monitoring is analyzed by adding noises to the original data.

A scheduling model for a multi-product, multistage batch plant with parallel units is presented. The objective is to maximize the weighted completion times of orders in every processing stage while imposing a penalty on the slower orders. The proposed model uses the continuous-time representation mode and describes the allocations of tasks, units and stages by a set of binary variables. In order to reduce the model size and provide a more effective solution to the model, a pre-ordering approach that sorts the processing sequence of orders is developed. The pre-ordering approach identifies the infeasible assignments through which the number of binary variables is significantly reduced. Illustrative examples are provided to show that the size of the proposed model is small, and therefore, needs much less computational effort in comparison with the existing models in the literature.

A profit function is established to evaluate the economic performance of baker's yeast continuous production. A structured kinetic model for baker's yeast Saccharomyces cerevisiae is applied to estimate the key state variables of the process, such as biomass and ethanol concentration. Taking possible time variant dynamics into account, a rolling identification approach is used to identify several key model parameters online so as to adapt the changing dynamics of the process. The well predicted biomass concentration makes the profit function online predictable very well. This investigation may benefit the study on profit optimization of the process.

Support vector machines (SVM) are applied to the prediction of key state variables in bioprocesses, such as product concentration and biomass concentration, which commonly play an important role in bioprocess monitoring and control. A so-called rolling learning-prediction procedure is used to deal with the time variant property of the process, and to establish the training database for the SVM predictor, which is characterized with the rolling update of the training database. As an example, product concentration in industrial penicillin production is predicted, and a comparison is also made with three different artificial neural network architectures (FBNN, RBFN, and RNN). The test results indicate that a prediction accuracy of 1-3 % can be obtained for 4-40h ahead prediction using the SVM, which is better than the best of the three artificial neural networks (ANNs). Moreover, for noise-added training highly noisy data or small-sample learning, the SVM also clearly outper-forms FBNN, RBFN, and RNN.

Robust controllers that ensure closed-loop pole placement are designed for uncertain discrete-time switched systems. A robust switching rule is proposed combined with the switched static output feedback controllers to ensure 1) quadratic stabilisation of the resulting closed-loop system, 2) pole location inside a circle for each linear mode of operation. The switching rule is state-dependent and does not rely on any uncertainties. All the results are expressed in terms of linear matrix inequalities (LMIs), which can be easily tested with efficient algorithms.