Parameter identification of a highly promising cleaner coal power station

https://doi.org/10.1016/j.jclepro.2021.129323Get rights and content

Highlights

  • An improved model for cleaner supercritical power plant is presented and assessed.

  • Two Multiobjective optimization techniques are used for parameter identification and compared against each other.

  • The model equations have gained simpler structure and the depicted simulations have shown higher accuracy.

  • Simulations have proved the feasibility of supercritical plants to sustain cleaner production and flexible power generation.

Abstract

It is well known that supercritical coal-firing power plants (SCPP) are much cleaner and more efficient than subcritical units. The boiler operating pressure and temperature for supercritical units are much higher, which causes improved overall cycle efficiencies, less fuel consumption, and lower undesirable emissions. Apart from the requirements of the materials to tolerate supercritical conditions, modeling of SCPP is more sophisticated than subcritical power plants. The highly complicated and multiple operational objectives in SCPP introduce further challenges in attaining sufficiently accurate results by computer modeling and simulation of the significant variables in the plant. This paper investigates the procedure of parameter identification and refinement of a real 600 MW supercritical unit. The model derivation is based on physical and mathematical principles to represent the significant variables in the plant. The model accuracy has been much refined through extensive applications of multi-objective optimization techniques in addition to deductive reasoning to calibrate some process uncertainties. Comparative simulations with a previous model version have shown two main improvements, which are more accurate results and more simplified structure or reduced order. Two advanced multi-objective optimization techniques have been investigated and compared, which are Genetic Algorithms and Particle Swarm Optimization. It has been proved that the former technique produces parameters with more accurate simulations of slow dynamics occur in the boiler, whereas the latter technique has been more accurate in the fast dynamics of the synchronous generator. Studies on dynamics responses are conducted as additional simulations to formulate the supporting arguments for the sensible behavior of the plant due to hypothetical changes in the grid demands.

Introduction

Clean coal power generation technologies can be categorized into two notable classifications: the first category includes the methods of energy efficiency improvements and the second is carbon capture and storage (CCS) (Ghosh and Prelas, 2009, Miller, 2011). Supercritical coal-fired power plants (SCPP) fall within the first category (i.e. energy-efficient power stations) and therefore, it is a highly promising clean coal technology, which can be smoothly integrated with CCS to largely reduce greenhouse gas emissions. Supercritical is a thermodynamic term, which means there is no segregation between the vapor phase and liquid phase of a substance. The SCPP boiler is working at supercritical pressure and temperature (higher than the critical region of water: 22.12 MPa pressure and 374 °C temperature) (Dipak et al., 2015) to gain higher efficiency. Thereby increasing the thermal energy harvested from the boiler for the same amount of burnt fuel. Globally, coal-fired power plants are increasingly being operated worldwide in many developed and developing countries due to the continuous and high growth of load demands (Ghosh and Prelas, 2009). However, any new thermal units should be much cleaner and efficient to satisfy the future climate targets and reduce the global warming effects. Renewable energy resources (RERs) are not adequate enough at this time to satisfy the global energy demand. In addition, the embedded RERs introduce more operational objectives of wider flexibility of conventional thermal power plants through frequent start-ups and shutdowns and grid-connected load following capabilities of energy to preserve the power system security and stability. Therefore, it is important to study the advanced clean coal generation technologies during this time as a promising option scenario for many goals including enhanced power system performance and reduction of emissions. One of the significant tools to facilitate those goals in their earlier stage is the development of trustable models for coal-fired power plants which is useful to predict and suggest the most feasible ways of safe and efficient operation of the process. Discussing all clean coal technologies is not the subject of this paper; however, the most salient clean coal technologies are depicted in Fig. 1. SCPP technology is a topic of merit due to its positive interdependency with the power system operation as well as targeted environmental enhancement.

It is preferred to elaborate more on the background of SCPP research topic. It can be clearly deduced that higher thermal energy can be harvested from the boiler in case of supercritical operation, which means lower CO2 emissions and lower coal consumption. This is an intensive interdisciplinary and trans-disciplinary research area that requires backgrounds from several engineering disciplines in order to achieve feasible outcomes. Research projects on SCPP modeling may be conducted interdisciplinary between two disciplines to achieve a particular research outcome, but it may be going beyond their boundaries for another particular research objective. Fig. 2 describes the trans-disciplinary perspective of supercritical power plant modeling. The mentioned research outcomes are just examples over the investigated literature and historical background about the subject.

The energy conversion processes of SCPP can be fully understood with the aid of Fig. 3. The plant has three direct inputs, feedwater flow, coal flow, and the digital-electro-hydraulic (DEH) governor valve reference. The raw coal is pulverized in a group of coal mills (usually vertical spindle type) that drive the coal powder to the furnace, where it is combusted to deliver heat energy to the water inside the boiler tubes. The heat is transferred to the water to raise it to the supercritical condition that happens in the waterwalls. The superheater deliver the superheated supercritical steam to the high pressure (HP) turbine and the steam flow output from the turbine is returned again to the reheater for further heat addition, the reheater output steam is used to energize the intermediate pressure (IP) turbine in order to have adequate steam expansion. The Synchronous generator converts the mechanical power into electrical power through the rotor that is coupled to the turbines with the same shaft. The rotor DC excitation produces constant magnetic flux but rotates with the rotor to act as a rotating magnetic field. The rotating field cuts the stator windings to induce voltages that are equal in magnitude but displaced in phase angels and the space of the machine. For grid-connected generators, the electrical power is delivered to the grid and the stator produces another rotating field that rotates at the same speed as the rotor. The exhausted steam from turbines is condensed and mixed with water supplied by the cooling tower and hence the steam cycle is repeated. The flue gas is discharged by the induced draft fans to the stack, then to the environment. The system outputs are usually taken to be the generated power, the main steam pressure, and the rotor's speed or frequency of the generator. Some targets may require the main steam temperature to be taken as output, however, the system outputs are selected according to the research objectives, and they can be easily assigned from system state variables equations.

There is a variety of purposes of modeling thermal power plants, subcritical or supercritical, including optimization of their availability (Jagtap et al. 2020, 2021), performance evaluation of the plant's components (Nikam et al., 2020), performance and production rate predictions (Kumar et al. 2019, 2020) and so on. The literature review, which focuses on cleaner coal units, shown some recent models for supercritical and ultra-supercritical units published with assertiveness on different problems on the plant and its compliance to grid codes specified by system authority. For a more detailed and critical review, it is suggested to refer to the most recent review article published by the corresponding author (Mohamed et al., 2020). However, in this article, a review of recent achievements over the past ten years is given, which will be sufficient to extract the research work significance and contributions. Mohamed et al. (2011, 2012a) have presented a physical-based model for a 600 MW supercritical unit and has been implemented in MATLAB/SIMULINK®. The parameters have been identified using GA with scattered crossover from 600 MW on-site closed-loop data. The model has been verified over the whole once-through mode of operation. Liu et al. (2015) has shown a simplified control-oriented model for a 1000 MW ultra-supercritical unit that has been verified in dynamic and steady state through comparison with measured open-loop responses. The model is based on thermodynamic principles of mass and energy conservation, built in MATLAB®, and the parameters are identified using GA to fit the operating data of the system. More recently, Deng et al. (2017) have introduced a physical-based simulator to simulate the start-up process using (APROS®) commercial software package. The model parameters are calculated via thermodynamic formulas and the model has shown its merits through comparison with design data. Taler et al. (2019) have developed an online-ready model for supercritical boiler that allows investigating the boiler dynamics during startup and small percentages of power increase. Fan et al. (2017, 2020, 2021) have built a 1000 MW ultra-supercritical plant model by physical principles and the parameters have been identified by immune genetic algorithm (IGA). Sarda et al. (2018) have offered a SCPP model built by Aspen Plus Dynamics® (APD®) simulation tool. The established model covers the flue gas path and supercritical steam cycle. No details are given optimization or identification of the parameters. Yang et al. (2021) have investigated the high-heat flux flow and heat transfer transitions in a supercritical loop by numerical verification and correlation. Some other empirical models have demonstrated different architectures of Neural Networks, which are published to address the simulation of dynamics of SCPP and ultra-supercritical plants. Muhammad Ashraf et al. (2020) have presented a feed-forward backpropagation neural network model and least square support vector machine for 660 MW for the target of performance optimization, more recently, Deep NN has been used for 1000 MW ultra-supercritical (Zhang et al., 2019; Cui et al., 2020). Some ANN architectures have shown slight improvements over other previously published ANN structures. For control system upgrades, closed-loop modeling and identification is more realistic in order to facilitate the addition of the modern controller at the supervisory level on the existing one on the regulatory level. In particular, physical-based models are usually preferred to address the control system task precisely because the model philosophy retains the physical foresight of the plant, which is necessary to reflect plant behavior under normal and emergency conditions, especially in multi-input multi-output (MIMO) with different time-scales for the outputs' dynamics. On the other hand, the Neural Networks are found to be more accurate for covering specific operating conditions and ranges, but they entirely built on the system's historical data sets, which are not necessarily informative for abnormal situations. From our proposed future research directions in the review article (Mohamed et al., 2020), some opportunities for improvements of parameters of simplified physical models are reported and thus, more accurate results and simplified physical model structures seem to be potential. Control system philosophy demands simplified structure, but simplified models lack accuracy as in detailed models. To the best of the authors' comprehension, it is very hard to gain relative simplicity and accuracy in the same model. The contribution of this paper can be then stated as follows:

1- An improved simplified physics-based model for SCPP has been presented and its unknown parameters have been identified and validated via two multi-objective optimization techniques. The derived model version in this paper has been compared with our previously published model in 2011 (Mohamed et al. 2011, 2012a) and has shown significant improvement in terms of accuracy and simplicity, thereby extended our previous version in terms of improved accuracy and lower computational requirements, which can be useful for control orientations. The main tools for improvements have resulted from rigorous effort done the manipulating of settings of the multi-objective optimization techniques in addition to some deductive and physical reasoning that are found to be useful to retain only the important variables in the system and recompense some uncertainties in the supercritical boiler process. The improved model covers the entire once-through mode of operation, which has been quantified for the plant under study from 252 MW up to 623 MW down to 300 MW for the power, and from 13.5 MPa to the supercritical 24.7 MPa for the pressure.

2- Multi-objective Genetic Algorithm (GA) with heuristic crossover and Particle Swarm Optimization (PSO) have been adopted to identify the enhanced model parameters because of their wider range applications, integrity, and robustness. The model parameters are identified and validated over the entire once-through mode of operation and the two methods have been compared. It is found that the GA is more accurate than PSO in identifying the pressure responses, whereas the PSO outperforms GA in the power responses. It can be deduced that GA is more appropriate in applications of optimization of slow dynamics (e.g. boiler dynamics), whereas PSO is preferred in case of relatively faster dynamics (e.g. electromechanical or electromagnetic dynamics). The graphical description of the 1st and 2nd contributions are shown in Fig. 4.

3- Dynamic response analysis is reported in the paper, which have been initiated by normalized changes of 6–12% in valve opening and 12% changes in coal flow and feedwater flow). This analysis has proved the suitability of SCPP to sustain a cleaner operation while satisfying power system regulations.

The paper is organized as follows: section. 2 presents the modeling philosophy adopted for SCPP, section.3 introduces the parameter identification techniques with the last version of parameters, section.4 depicts the simulation results and analyze qualitative and quantitative comparison with the previous model version, and finally, the conclusion and future research are presented in section. 5.

Section snippets

Modeling supercritical pulverized coal power stations

Thermodynamics principles of mass and energy balance are adopted to model boiler and turbine stages, and the generator is modeled classically using the swing equation that is rooted from Newton's second law. Considering control volumes for the main steam path, turbine, and the reheater unit is a one control volume. The mass balance principle clarifies that the mass of material is neither destroyed nor created, but it can be transformed from one form to another. The energy balance principle also

Parameter identification

The model for the power plant is dedicated to the once-through mode in this study, real sets of data have been used to reflect the behavior of the power plant with SC boiler, which has specification or operational parameters are mentioned in Table 1 (Mohamed et al., 2011).

A Sample power data is depicted in Fig. 7, which shows multiples processes of the SCPP from the OFF state to starting to synchronize to the grid, recirculation mode, and the once-through loading up to the rated power. The

Simulation results

It is preferable to divide this section into three subsections, one is to compare the results of the enhanced model with the last version of parameters as tabulated above, the second as another comparison between PSO and multi-objective GA used for identifying the enhanced model, and the last one for dynamic response analysis using the enhanced model simulations.

Conclusion

In this paper, the procedure of parameter identification of a real 600 MW supercritical power plant is investigated, using two artificial intelligence techniques which are GA and PSO. A comparison between the results obtained by each technique is made. The model is mathematically developed to a simpler structure while attaining higher accuracy than the previously published model. GA and PSO are applied to the model and the simulation results show that both techniques can be applied to achieve

CRediT authorship contribution statement

Amal Haddad: Conceptualization, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. Omar Mohamed: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. Mustafa Zahlan: Software. Jihong Wang: Data curation, Supervision.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (41)

  • J. Cui et al.

    Deep-neural-network-based economic model predictive control for ultra-supercritical power plant

    IEEE Trans. Ind. Inform.

    (2020)
  • Y.A. Cengel et al.

    Thermodynamics an Engineering Approach

    (2015)
  • Y. Del Valle et al.

    Particle swarm optimization: basic concepts, variants and applications in power systems

    IEEE Trans. Evol. Comput.

    (2008)
  • K.S. Dipak

    Steam generators

  • R. Eberhart et al.

    Particle swarm optimization

  • C.M. Fonseca et al.

    Multiobjective genetic algorithms

  • T. Ghosh et al.

    Energy resources and systems

  • S. Gue et al.

    A new model-based approach for power plant Tube-ball mill condition monitoring and fault detection

    Energy Convers. Manag.

    (2014)
  • J.H. Holland

    Genetic algorithms

    Sci. Am.

    (1992)
  • T. Inoue et al.

    A model of fossil fueled plant with once-through boiler for power system frequency simulation studies

    IEEE Trans. Power Syst.

    (2000)
  • Cited by (8)

    • Artificial intelligence enabled efficient power generation and emissions reduction underpinning net-zero goal from the coal-based power plants

      2022, Energy Conversion and Management
      Citation Excerpt :

      Haddad et al., applied genetic algorithm and particle swarm optimization techniques for the parametric identification from a 660 MW power plant. The simulations proved the feasibility for cleaner and flexible power generation [25]. Liu et al., modeled the 1000 MW ultra-supercritical power plant operation by the significant operating parameters using fuzzy neural network methods.

    • Multiple processes modeling and identification for a cleaner supercritical power plant via Grey Wolf Optimizer

      2022, Energy
      Citation Excerpt :

      The working fluid is assumed to have uniform properties in a one-phase flow in the heat exchanger, the heat exchanger pipes are assumed to be lumped together, the three superheaters are modeled as a single superheater, finally, the flue gas dynamics have not been considered. Another feasible and viable assumption has been presented in Ref. [7] that the boiler's main steam path is considered as a single heat exchanger for the once-through mode, however, this will not be practical in dealing with multiple processes, from recirculation to shutdown, the study presented in this work addresses multiple process including the recirculation and shutdown, and that implicates changes in the working fluid path and adaptation in some heat exchangers and turbine parameters accordingly, which needs some considerations of the multi-stages or the boiler sections. Apart from the once-through mode, which has been studied extensively and the plant is likely to operate, the startup and shutdown modes will not be mathematically represented by the same parameters via bidirectional concept, but rather they should be different for the following reasons:

    View all citing articles on Scopus
    View full text