# Estimation of the Solid Circulation Rate in Circulating Fluidized Bed System Using Adaptive Neuro-Fuzzy Algorithm

^{1}

^{2}

^{3}

^{*}

## Abstract

**:**

## 1. Introduction

_{2}which makes environment clean. A survey for biomass energy consumption was conducted in 2017 which describes that 9% to 13% of energy is obtained from biomass which is approximately 60 MJ in total amount of energy consumed [3]. There are various forms of energy, categorized as renewable energy, non-renewable energy and nuclear energy.

- Solid
- Liquid
- Gas

## 2. Materials and Methods

#### 2.1. Experimental Setup and Data Collection

#### 2.2. Granulated Sample and Different Parameters in CFBC

#### 2.3. ANN Model Development and Its Working Principal

#### 2.3.1. Parameters Selection for Developing ANN

_{ij}represent weight of ith input and jth neuron of the hidden layer and b

_{j}is bias of jth layer. A tansigmoid function has been used for training of model. The proposed ANN model uses Levenberg–Marquardt feed forward back propagation (BP) learning algorithm with tansigmoid function for training expressed in Equation (2). The architecture of ANN model used is shown in Figure 3.

#### 2.3.2. Tuning of ANN Model

- Batch training operation in which noises are included to train the model.
- Levenberg Marquardt (LM) with second order method which gives more accuracy in comparison of first order as time consumed by second order is less, while noises are also included in this operation.

#### 2.4. Adaptive Neuro Fuzzy Inference System (ANFIS)

#### 2.5. Parameters Selection and Tuning for Developing ANFIS

## 3. Results

^{2}) of ANN and ANFIS can be determined by Equation (9).

^{2}values for training, testing and validation of data for ANN are represented in Table 5. Figure 10 compares the values predicted by ANN with the (experimental) target values while Figure 11 shows the error produced by ANN.

^{2}should approach to 1 for accurate results.

^{2}values in training and testing are close to one another. Furthermore, if the MSE value touches approximately zero, and then it indicates that tuning of the proposed model is excellent for all data sets, also the results are very close to each other in between predicted and experimental values. The difference between network output and the intended outputs is found to be very low which can be neglected. Such results are of utmost satisfaction and quite acceptable on the basis of R

^{2}and MSE values. Figure 12 compares the values predicted by ANFIS with the (experimental) target values while Figure 13 shows the error produced by ANFIS. There might be some fraction of noise in 100th sample due to any sensors while collecting experimental data set. As the ANFIS is trained for training data samples only, any noise can therefore contribute to error. It can be one of the main reasons that error between the experimental results and the predicted values starts to become more noticeable at this specific sample. Yet, the error variation is not too high if we see the scale on Y-axis and it does not significantly affect the accuracy of the results. The experimental findings for the prediction of solid circulation rate at high pressure in CFBC through ANN and ANFIS model are mentioned in Table 8. The ANFIS model predicts mass flow rate of solid better than artificial neural network as can be seen in Table 9.

## 4. Discussions

^{2}. We have compared our findings with Table 1 in the introduction part. We found that proposed ANFIS model gives more accurate results and value of MSE error is very low and regression analysis is approaching to 1. The local sensitivity analysis is performed using one-at-a-time method to check that how much output variable is sensitive to any change in the input variable. The results of the sensitivity analysis are summarized in Table 10. It has been observed that output is most sensitive to any change in the total value opening, then riser dp, pressure and SMD. Pyrolysis process usually takes place at 700–1000 °Cand to get high heat efficiency the temperature must be maintained at 800–1000 °C. In order to improve the efficiency of gasifier, a balance between flow of mass and produced energy must be maintained to achieve the desired values of temperature and pressure. The pressure, temperature, valve opening and height of the reactor bed are the most important hydrodynamic parameters which ensure smooth flow of solid particles. Therefore, accurate estimation of solid circulation rate on the basis of these input parameters greatly affects the efficiency and performance of CFB gasifier. The CFB usually achieves combustion efficiency of 95% which can be further improved by employing proposed ANN and ANFIS algorithms.

## 5. Conclusions

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Conflicts of Interest

## Appendix A

## References

- Qi, J.; Li, H.; Han, K.; Zuo, Q.; Gao, J.; Wang, Q.; Lu, C. Influence of ammonium di-hydrogen phosphate on potassium retention and ash melting characteristics during combustion of biomass. Energy
**2016**, 102, 244–251. [Google Scholar] [CrossRef] - Chen, X. Economic potential of biomass supply from crop residues in China. Appl. Energy
**2016**, 166, 141–149. [Google Scholar] [CrossRef] - Qi, J.; Han, K.; Wang, Q.; Gao, J. Carbonization of biomass: Effect of additives on alkali metals residue, SO
_{2}and NO emission of chars during combustion. Energy**2017**, 130, 560–569. [Google Scholar] [CrossRef] [Green Version] - Ganesan, T.; Lingappan, S. A Survey on circulating fluidized bed combustion boilers. Int. J. Adv. Res. Electr. Electron. Instrum. Eng.
**2013**, 2, 4032–4042. [Google Scholar] - Sharma, S.; Ghoshal, S.K. Hydrogen the future transportation fuel: From production to applications. Renew. Sustain. Energy Rev.
**2015**, 43, 1151–1158. [Google Scholar] [CrossRef] - Ozbilen, A.; Dincer, I.; Rosen, M.A. Life cycle assessment of hydrogen production via thermochemical water splitting using multi-step Cu-Cl cycles. J. Clean. Prod.
**2012**, 33, 202–216. [Google Scholar] [CrossRef] - Ansari, A. Biomass: Energy and environmental concerns in developing country. Res. J. Environ. Sci.
**2012**, 1, 54–57. [Google Scholar] - Yılmaz, S.; Selim, H. A review on the methods for biomass to energy conversion systems design. Renew. Sustain. Energy Rev.
**2013**, 25, 420–430. [Google Scholar] [CrossRef] - Sharma, A.; Pareek, V.; Zhang, D. Biomass pyrolysis A review of modeling, process parameters and catalytic studies. Renew. Sustain. Energy Rev.
**2015**, 50, 1081–1096. [Google Scholar] [CrossRef] - George, J.; Arun, P.; Muraleedharan, C. Assessment of producer gas composition in air gasification of biomass using artificial neural network model. Int. J. Hydrog. Energy
**2018**, 43, 9558–9568. [Google Scholar] [CrossRef] - Aydinli, B.; Caglar, A. A degradation kinetic study on pyrolysis of three biomass samples and co-pyrolysis of hazelnut shell and ultra-high molecular weight polyethylene blends using a silver ındicator. Energy Source Part A Recovery Util. Environ. Eff.
**2013**, 35, 900–908. [Google Scholar] [CrossRef] - Uzun, H.; Yıldız, Z.; Goldfarb, J.L.; Ceylan, S. Improved prediction of higher heating value of biomass using an artificial neural network model based on proximate analysis. Bioresour. Technol.
**2017**, 234, 122–130. [Google Scholar] [CrossRef] [PubMed] - Wang, H.G.; Yang, W.Q.; Wu, Z.P. Measurement of External-Solids Circulation Flux in a Circulating Fluidized Bed by Electrical Capacitance Tomography and Microwave Doppler Radar. In Proceedings of the 2018 IEEE International Conference on Imaging Systems and Techniques, Krakow, Poland, 16–18 October 2018; pp. 1–5. [Google Scholar]
- Luo, G.; Cheng, L.; Li, L.; Wang, J.; Jiang, X.; Zhang, W. Effect of Connection Resistance between Cyclones and Backpass on Furnace Solids Suspension Density Profile and Circulation Rates in CFB. Energies
**2021**, 14, 6162. [Google Scholar] [CrossRef] - Gueguen, R.; Sahuquet, G.; Mer, S.; Toutant, A.; Bataille, F.; Flamant, G. Gas-Solid Flow in a Fluidized-Particle Tubular Solar Receiver: Off-Sun Experimental Flow Regimes Characterization. Energies
**2021**, 14, 7392. [Google Scholar] [CrossRef] - Sharma, B.; Venugopalan, K. Comparison of neural network training functions for Hematoma classification in brain CT images. J. Comput. Eng.
**2014**, 16, 31–35. [Google Scholar] [CrossRef] - Arnavat, M.P.; Hernandez, J.A.; Bruno, J.C.; Coronas, A. Aartificial neural network models for biomass gasification in fluidized bed gasifiers. Biomass Bioenergy
**2013**, 49, 279–289. [Google Scholar] [CrossRef] - Chen, B.; Xie, Q.; Zhou, J. Fuzzy Adaptive PID Control of Biomass Circulating Fluidized Bed Boiler. In Proceedings of the 2018 Chinese Automation Congress, Xi’an, China, 30 November–2 December 2018; pp. 3795–3800. [Google Scholar]
- Majchrzak-Kuceba, I.; Wawrzynczak, D.; Zdeb, J.; Smółka, W.; Zajchowski, A. Treatment of Flue Gas in a CO
_{2}Capture Pilot Plant for a Commercial CFB Boiler. Energies**2021**, 14, 2458. [Google Scholar] [CrossRef] - Zhang, D.; Min, T.; Jiang, M.; Yu, Y.; Zhou, Q. Numerical Simulation of Fluidized Bed Gasifier Coupled with Solid Oxide Fuel Cell Fed with Solid Carbon. Energies
**2021**, 14, 2800. [Google Scholar] [CrossRef] - Pandey, D.S.; Das, S.; Pan, I.; Leahy, J.J.; Kwapinski, W. Artificial neural network based modeling approach for municipal solid waste gasification in a fluidized bed reactor. Waste Manag.
**2016**, 58, 202–213. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Khurram, M.S.; Choi, J.H.; Won, Y.S.; Jeong, A.R.; Ryu, H.J. Relationship between solid flow rate and pressure drop in the riser of a pressurized circulating fluidized bed. J. Chem. Eng. Jpn.
**2016**, 49, 595–601. [Google Scholar] [CrossRef] - Panah, A.; Motameni, H.; Ebrahimnejad, A. An Efficient Computational Hybrid Filter to the SLAM Problem for an Autonomous Wheeled Mobile Robot. Int. J. Control. Autom. Syst.
**2021**, 19, 1–10. [Google Scholar] [CrossRef] - Anwer, A.M.O.; Omar, F.A.; Kulaksiz, A.A. Design of a Fuzzy Logic-based MPPT Controller for a PV System Employing Sensorless Control of MRAS-based PMSM. Int. J. Control. Autom. Syst.
**2020**, 18, 2788–2797. [Google Scholar] [CrossRef] - Asghar, A.B.; Liu, X. Adaptive neuro-fuzzy algorithm to estimate effective wind speed and optimal rotor speed for variable-speed wind turbine. Neurocomputing
**2018**, 272, 495–504. [Google Scholar] [CrossRef] - Truong, N.T.; Seo, T.-I.; Nguyen, S.D. Bearing Fault Online Identification Based on ANFIS. Int. J. Control. Autom. Syst.
**2021**, 19, 1703–1714. [Google Scholar] [CrossRef] - Gularte, K.H.M.; Chavez, J.J.M.; Vargas, J.A.R.; Alfaro, S.C.A. An Adaptive Neural Identifier with Applications to Financial and Welding Systems. Int. J. Control. Autom. Syst.
**2021**, 19, 1976–1987. [Google Scholar] [CrossRef] - Ozonoh, M.; Oboirien, B.O.; Higginson, A.; Higginson, M.O. Performance Evaluation of Gasification System Efficiency Using Artificial Neural Network. J. Renew. Energy Daramola
**2020**, 145, 2253–2270. [Google Scholar] [CrossRef]

**Figure 1.**Pressurized circulating fluidized bed (CFB) system: 1. Compressor; 2. Valve; 3. Mass flow meter controller; 4. Riser inlet; 5. Riser; 6. Pressure port; 7. Upper bubbling bed; 8. Aeration; 9. L-valve and aeration; 10. Lower bubbling bed; 11.Hopper; 12. Cyclone; 13. Pressure controller including bag filters; 14. Pressure transducer; 15. Data logger; 16. Personal computer; 17. Final bag filter. Own elaboration based on [22].

**Figure 9.**Trained MFs of the input variables, (

**a**) trained Gaussian MFs of pressure, (

**b**) trained Gaussian MFs of riser dp, (

**c**) trained Gaussian MFs of single mean diameter, (

**d**) trained Gaussian MFs of total valve opening.

Reference No. | Data Set | Errors |
---|---|---|

[10] | coconut shell, coffee husk, groundnut hell, sawdust and sugarcane bagasse | R^{2} = 0.98MSE = 0.71 |

[12] | 131 biomass samples from gasifier | MSE = 0.375 R ^{2} = 0.963 |

[28] | 315 experimental data of biomass, coal, and blends of biomass and coal from various gasifier | MISO system with R^{2} = 0.78–0.98MIMO system with R ^{2} = 0.95–0.96 |

[21] | 61 samples of biomass data set with varying hidden layer from 1 to 10 layers | MSE = 0.00057 R ^{2} = 0.98 |

Input Parameters | Dimensions | Output Parameters | Dimensions |
---|---|---|---|

Single mean diameter (m) | 1 × 217 | Mass flow rate (g/s) | 1 × 217 |

Pressure (Bar-abs) | 1 × 217 | ||

Riser dp (mmH20) | 1 × 217 | ||

Total valve opening (cm/s) | 1 × 217 |

Tuning of Parameters | Values |
---|---|

Input layer neurons | 4 |

Number of hidden layers | 3 |

Hidden layer neurons | 50 |

Momentum rate | 0.9 |

Transfer function | Hyperbolic tangent sigmoid (Tansig) |

Number of iterations (epochs) | 1000 |

Output layer neurons | 1 |

Error tolerance | 0.0001 |

Training function | Levenberg–Marquardt backpropagation (TRAINLM) |

Performance function | MSE, Regression |

ANFIS Parameters | Description/Values |
---|---|

Fuzzy structure | Sugeno-type |

FIS Generation approach | Grid partition |

I/P Membership Function | Gaussian, Bell-shaped, Sigmoidal, Gaussian2 |

Error Tolerance | 0.001 |

O/P Membership Function | Linear, constant |

No of I/P data | 4 × 217 |

No of O/P data | 1 × 217 |

No of Iteration | 250 |

No. of I/P MF | 5 |

Optimization Method | Hybrid |

Type of Data | MSE | REG (R^{2}) |
---|---|---|

Training | 0.0439 | 0.9914 |

Testing | 0.0337 | 0.9918 |

Validation | 0.0422 | 0.9934 |

**Table 6.**Comparative analysis of ANFIS performance for various types of input MFs for constant MF of output.

Type of Input MF | No. of MF | No of Epochs | MSE | RMSE | MAE | MAPE | R^{2} |
---|---|---|---|---|---|---|---|

Psigmf | 5 | 250 | 1.4282 | 1.1951 | 0.8193 | 5.3922 | 0.9799 |

Bell-shape | 5 | 250 | 0.6002 | 0.7747 | 0.5052 | 3.1644 | 0.9916 |

Gauss | 5 | 250 | 0.4507 | 0.6713 | 0.4326 | 2.6397 | 0.9937 |

Gauss2mf | 5 | 250 | 1.7482 | 1.3222 | 0.9025 | 5.8431 | 0.9754 |

**Table 7.**Comparative analysis of ANFIS performance for various types of input MFs for linear MF of output.

Type of Input MF | No of Epochs | MSE | RMSE | MAE | MAPE | R^{2} |
---|---|---|---|---|---|---|

Psigmf | 250 | 0.0075 | 0.0866 | 0.0306 | 0.2319 | 0.9999 |

Bell-shape | 250 | 0.1114 | 0.0247 | 0.0084 | 0.0675 | 1.0000 |

Gauss | 250 | 0.0519 | 0.0168 | 0.0060 | 0.0379 | 1.0000 |

Gauss2mf | 250 | 0.0392 | 0.1981 | 0.0653 | 0.4892 | 0.9994 |

Pressure | SMD | Total Valve | Riser Dip | Mass Flow Rate by Experimental | Mass Flow Rate by ANN | ANN Error | Mass Flow Rate by ANFIS | ANFIS Error |
---|---|---|---|---|---|---|---|---|

1.013 | 5.5 × 10^{−3} | 160 | 379 | 12.2075 | 12.7592 | −0.5517 | 12.2064 | 0.0011 |

1.013 | 5.5 × 10^{−3} | 280 | 102 | 12.4266 | 12.1434 | 0.2831 | 12.4262 | 0.0004 |

1.013 | 5.5 × 10^{−3} | 200 | 209 | 14.5853 | 13.7629 | 0.8223 | 14.5846 | 0.0006 |

1.013 | 5.5 × 10^{−3} | 170 | 530 | 15.6567 | 16.6004 | −0.9437 | 15.6519 | 0.0048 |

1.013 | 5.5 × 10^{−3} | 240 | 194 | 17.3973 | 17.5688 | −0.1715 | 17.3974 | −0.0001 |

1.013 | 5.5 × 10^{−3} | 210 | 295 | 18.1431 | 18.6818 | −0.5387 | 18.1422 | 0.0009 |

Types of NN | Types of Error | ||||
---|---|---|---|---|---|

MAE | R^{2} | MAPE | MSE | RMSE | |

ANN | 0.7919 | 0.9806 | 4.4612 | 1.0677 | 1.0806 |

ANFIS | 0.0060 | 1.0000 | 0.0379 | 0.0519 | 0.0168 |

Input Parameter | Sensitivity Ratio | Importance Order |
---|---|---|

Total valve opening | 4.2 | 1 |

Riser dp | 2.14 | 2 |

Pressure | 1.99 | 3 |

SMD | 1.07 | 4 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Asghar, A.B.; Farooq, S.; Khurram, M.S.; Jaffery, M.H.; Ejsmont, K.
Estimation of the Solid Circulation Rate in Circulating Fluidized Bed System Using Adaptive Neuro-Fuzzy Algorithm. *Energies* **2022**, *15*, 211.
https://doi.org/10.3390/en15010211

**AMA Style**

Asghar AB, Farooq S, Khurram MS, Jaffery MH, Ejsmont K.
Estimation of the Solid Circulation Rate in Circulating Fluidized Bed System Using Adaptive Neuro-Fuzzy Algorithm. *Energies*. 2022; 15(1):211.
https://doi.org/10.3390/en15010211

**Chicago/Turabian Style**

Asghar, Aamer Bilal, Saad Farooq, Muhammad Shahzad Khurram, Mujtaba Hussain Jaffery, and Krzysztof Ejsmont.
2022. "Estimation of the Solid Circulation Rate in Circulating Fluidized Bed System Using Adaptive Neuro-Fuzzy Algorithm" *Energies* 15, no. 1: 211.
https://doi.org/10.3390/en15010211