3.1 Characterization of biomass samples
Table 1 offers a detailed comparison of key parameters among various fuel sources, namely coal, charcoal (pinewood charcoal), pine cone, and pine needle. Notably, charcoal, derived specifically from pinewood, distinguishes itself with a high fixed carbon content of 79.14%. In terms of volatile matter, the pine cone and pine needle showcase notably elevated values (77.43% and 74.53%, respectively) compared to Coal (20.93%) and Charcoal (18.9%). Conversely, charcoal stands out with the lowest ash content at 1.96%. Pine cone (0.81%) and pine needle (4.13%) also have very low ash content whereas coal has the highest ash content of 60.31% [19, 20]. HHV varies across the fuels, with charcoal (Pinewood Charcoal) exhibiting the highest value at 7327 MJ/kg, followed by pine needle (4300 MJ/kg), pine cone (4200 MJ/kg), and coal (2597 MJ/kg). The table further delves into the chemical composition of various fuel sources, providing additional insights into specific elements. Notably, the Fluoride content is highest in coal at 264 mg/kg, while pinewood charcoal, pine cone, and pine needle show significantly lower values at 19.45 mg/kg, 11.4 mg/kg, and 9.35 mg/kg, respectively. In terms of Chloride concentration, coal displays a low value of 42.63 mg/kg, comparable to that of pine cone (53.03 mg/kg). Pinewood charcoal and pine needle, on the other hand, exhibit higher values at 157.9 mg/kg, and 1285.55 mg/kg, respectively. Bromide levels also differ among the fuels, with coal having the highest concentration at 25.93 mg/kg, followed by pinewood charcoal, pine cone, and pine needle. Interestingly, Sulphur content varies significantly across the fuels, with Coal presenting the highest value at 5383.1 mg/kg, while Pinewood charcoal, Pine cone, and pine needle exhibit lower concentrations at 1903.75 mg/kg, 713.23 mg/kg, and 1951.8 mg/kg, respectively. Figure 2 shows the graphical representation of the Dry ash (%), HHV (kcal/kg), sulphur (mg/kg) and chloride (mg/kg) content of the various fuel sources used in this study.
Table 1
Fuel Characteristics of pure samples
| Coal | Pinewood Charcoal | Pine cone | Pine needle |
Proximate analysis* |
Volatile Matter (%) | 20.93 ± 0.54 | 18.9 ± 0.47 | 77.43 ± 1.027 | 74.53 ± 1.524 |
Fixed Carbon (%) | 18.76 ± 0.4 | 79.14 ± 0.45 | 21.76 ± 0.92 | 21.34 ± 1.25 |
Ash (%) | 60.31 ± 0.15 | 1.96 ± 0.1 | 0.81 ± 0.17 | 4.13 ± 0.29 |
Heating value analysis |
Higher Heating Value (kcal/kg) | 2597 ± 103.12 | 7327 ± 146.33 | 4200 ± 122.05 | 4300 ± 89.42 |
Ultimate analysis |
Carbon (%) | 23.15 ± 0.25 | 33.18 ± 0.37 | 41.27 ± 0.12 | 46.96 ± 0.62 |
Hydrogen (%) | 2.14 ± 0.01 | 2.33 ± 0.04 | 5.75 ± 0.02 | 6.03 ± 0.02 |
Nitrogen (%) | 0.28 ± 0.05 | 0.1 ± 0.02 | 0.34 ± 0.01 | 1.73 ± 0.04 |
Combustion ion chromatography analysis |
Fluoride (mg/kg) | 264 ± 10.75 | 19.45 ± 1.48 | 11.4 ± 0.44 | 9.35 ± 0.21 |
Chloride (mg/kg) | 42.63 ± 0.64 | 157.9 ± 2.12 | 53.03 ± 4.05 | 1285.55 ± 53.10 |
Bromide (mg/kg) | 25.93 ± 1.33 | 54.55 ± 2.33 | 35.35 ± 1.34 | 43.35 ± 6.86 |
Sulphur (mg/kg) | 5383.1 ± 794.34 | 1903.75 ± 370.74 | 713.23 ± 83.1 | 1951.8 ± 63.22 |
* dry basis |
3.2 Statistical analysis
Table 2 shows the values of output parameters (experimentally measured) namely, Dry Ash content, Chloride content, Sulphur content, and Higher Heating Value (HHV) corresponding to each of the 20 separate mixtures. Table 3 presents the results of a comprehensive statistical analysis conducted to understand the influence of the four components, denoted as A, B, C, and D on various output parameters. Each factor's significance and impact are evaluated through a series of statistical measures and tests.
Table 2
Fuel Characteristics of mixed coal-biomass samples
S. No. | Coal (%) | Pinewood charcoal (%) | Pine cone (%) | Pine needle (%) | Dry Ash (%) | HHV (kCal/kg) | Chloride (mg/kg) | Sulphur (mg/kg) |
1 | 60 | 10 | 10 | 20 | 37.54 ± 1 | 3676 ± 113 | 286.4 ± 4.49 | 4491.57 ± 490.57 |
2 | 50 | 20 | 20 | 10 | 32.29 ± 1.2 | 4125 ± 101 | 171.77 ± 15.07 | 4796.2 ± 311.79 |
3 | 60 | 10 | 10 | 20 | 38.91 ± 0.2 | 3641 ± 97 | 262.6 ± 8.17 | 4529.1 ± 145.08 |
4 | 50 | 20 | 10 | 20 | 32.5 ± 0.12 | 4192 ± 130 | 295.07 ± 21.66 | 3969.5 ± 495.12 |
5 | 50 | 10 | 10 | 30 | 30.33 ± 0.36 | 3917 ± 116 | 383.7 ± 38.85 | 3793.07 ± 351.18 |
6 | 53 | 23 | 12 | 12 | 33.72 ± 0.34 | 4178 ± 106 | 193.17 ± 5.46 | 4077.73 ± 330.54 |
7 | 60 | 20 | 10 | 10 | 38.45 ± 0.15 | 3972 ± 199 | 173.6 ± 30.12 | 3874.55 ± 586.69 |
8 | 60 | 10 | 20 | 10 | 38.94 ± 0.14 | 3682 ± 91 | 163.2 ± 17.52 | 3979.63 ± 344.34 |
9 | 60 | 10 | 20 | 10 | 37.81 ± 0.45 | 3723 ± 103 | 163.43 ± 9.97 | 4147.23 ± 545.97 |
10 | 63 | 12 | 12 | 13 | 40.04 ± 2.45 | 3654 ± 102 | 206.97 ± 5.71 | 3720.97 ± 154.3 |
11 | 60 | 20 | 10 | 10 | 37.68 ± 0.03 | 3840 ± 107 | 160.97 ± 10.19 | 4623.97 ± 235.49 |
12 | 50 | 30 | 10 | 10 | 31.34 ± 0.27 | 4544 ± 144 | 181.43 ± 7.14 | 4977.3 ± 746.19 |
13 | 50 | 20 | 20 | 10 | 31.45 ± 0.21 | 4115 ± 181 | 169.43 ± 10.26 | 4943.63 ± 221.89 |
14 | 50 | 20 | 10 | 20 | 32.52 ± 0.16 | 4228 ± 145 | 277.43 ± 8.75 | 4461.3 ± 344.27 |
15 | 55 | 15 | 15 | 15 | 34.94 ± 0.25 | 3909 ± 196 | 235.1 ± 8.17 | 5576.53 ± 312.92 |
16 | 50 | 10 | 30 | 10 | 32.41 ± 0.93 | 3882 ± 106 | 159.33 ± 15.41 | 4286.07 ± 171.91 |
17 | 50 | 10 | 20 | 20 | 32.29 ± 0.63 | 3875 ± 122 | 266.23 ± 27.38 | 4367.4 ± 477.31 |
18 | 70 | 10 | 10 | 10 | 43.34 ± 1.24 | 3535 ± 95 | 163.73 ± 6.99 | 3760.33 ± 345.11 |
19 | 52 | 12 | 23 | 13 | 32.18 ± 0.84 | 3864 ± 80 | 205.53 ± 6.5 | 4649.47 ± 213.24 |
20 | 52 | 12 | 13 | 23 | 34.12 ± 0.33 | 3915 ± 129 | 321.5 ± 32.85 | 4964.57 ± 569.55 |
Table 3
Statistical analysis of experimental responses
| Dry Ash (%) | Chloride (mg/kg) | Sulphur (mg/kg) | HHV (kCal/kg) |
Linear Mixture | Significant | Significant | Significant | Significant |
R² | 0.96 | 0.99 | 0.82 | 0.97 |
Adj. R² | 0.96 | 0.98 | 0.79 | 0.96 |
Pred. R² | 0.93 | 0.98 | 0.71 | 0.94 |
P–Value | < 0.0001 | < 0.0001 | < 0.0001 | < 0.0001 |
F–value | 145.03 | 402.62 | 24.52 | 149.17 |
C.V. % | 2.12 | 3.66 | 2.54 | 1.28 |
Std. Dev. | 0.75 | 8.13 | 93.61 | 50.29 |
Lack of Fit P – Value | 0.39 | 0.86 | 0.68 | 0.43 |
Adequate Precision | 34.72 | 62.28 | 17.42 | 44 |
Actual equation | 0.60*A + 0.02*B + 0.01*-0.001*D | 0.46*A + 1.23*B + 0.36*C + 11.68*D | 50.23*A + 26.07*B + 13.76*C + 21.14*D | 25.60*A + 75.09*B + 45.09*C + 47.44*D |
Where, A = Coal, B = Pinewood charcoal, C = Pine cone and D = Pine needle composition |
Table 3 indicates whether a linear mixture is deemed significant for all parameters suggesting that the combination of components has a notable impact on the measured outputs. Coefficients of determination (R²), quantify the proportion of variability in the output parameters explained by the components. The R² values range from 0.82 to 0.99, indicating strong relationships between the components and the output parameters. For instance, Chloride exhibits the highest R² value of 0.99, suggesting that almost all the variability in Chloride content is accounted for by the components A, B, C, and D. Adjusted R² values give more detailed information by considering the number of predictors are in the model. The values range from 0.79 to 0.98, showing the strong explanatory power of the components while accounting for model performance. Predictive R² values offer insights into the models' predictive performance on new data. With values ranging from 0.71 to 0.98, they indicate robust predictive capabilities across the output parameters.
Statistical significance is confirmed by the low p-values (< 0.0001) and high F-values (24.52 to 402.62), indicating that the observed relationships between the components and output parameters are well-defined. The coefficient of Variation (C.V. %) (1.28 to 3.66) and Standard Deviation (0.75 to 93.61) reflect the variability within each output parameter across the different combinations of components. Lack of Fit P-Values assesses whether the models adequately fit the data. With values above 0.05, it can be suggested that the models adequately capture the variability in the data, providing a good fit. The adequate precision values range from 17.42 to 62.28, indicating that the models have relatively good predictive accuracy compared to the level of noise in the data.
Lastly, the actual equations provide insights into the relationships between the components and the output parameters, showing the relative importance of each component on the output variable and the way they affect the measured outputs.
3.3 ANN-MOGA-based optimization
The ANN protocol used in this study utilized 70% of the data set for training purposes, 15% for validation purposes, and the rest 15% data for testing purposes. Optimization of neurons in the hidden layer prompted the use of 6 neurons for the study based on minimum RMSE (Root Mean Squared Error), MAE (Mean Absolute Error), and MBE (Mean Bias Error). Table 4 shows the learning parameters of the ANN protocol used in the present study. The artificial neural network (ANN) described employs a feedforward architecture, where information moves in one direction, from input to output, without loops or cycles [18]. Utilizing the Levenberg-Marquardt training function enhances convergence speed by adapting the weights between layers. LEARNGD (Gradient Descent) serves for adaptation learning, adjusting weights iteratively based on the gradient of the error function to optimize performance. Additionally, mean square error is utilized as the performance function, measuring the average squared difference between predicted and actual outputs, indicative of model accuracy.
Table 4
Characteristics of ANN model and MOGA
ANN | MOGA |
Network type | Feedforward | Number of variables | 4 |
Training function | Train Levenberg-Marquardt (LM) | Number of output responses | 4 |
Adaption Learning function | LEARNGD (Gradient descent) | Population size | 200 |
Performance function | Mean square error | Crossover fraction | 0.8 |
Correlation coefficient | 0.9993 | Crossover function | @crossoverintermediate |
Input variable | 4 | Mutation rate | 0.05 |
Output response | 4 | Mutation function | @mutationadaptfeasible |
Hidden Layers | 1 | Elite count | 2 |
Neurons | 6 | Fitness function | ANN Model |
Training method | Back-propagation | Upper boundary | [50 10 10 10] |
Epochs | 1000 | Lower boundary | [70 30 30 30] |
With four input variables and four output responses, the network is designed for multi-dimensional input-output mapping. It comprises a single hidden layer with six neurons, providing flexibility in capturing complex relationships within the data. Back-propagation, a popular training method for ANNs, updates weights by propagating errors backward from the output layer to the input layer. The model undergoes 1000 epochs of training, iterating through the entire dataset multiple times to refine its predictive capabilities.
The Regression plots of training/validation/testing data for the ANN model are shown in Fig. 3.
Based on a comparative analysis of the projected output data generated by an Artificial Neural Network (ANN) and the input data derived from an optimized number of neurons, it was observed that the model exhibits trendlines with correlation coefficient (R) values of R = 0.9993 for the training data, R = 0.9993 for the validation data, and R = 0.9996 for the testing data. The collective data has a correlation coefficient, R = 0.9993, as depicted in Fig. 3. Table 5 shows the predicted values of different output variables for different mixture compositions as determined by the ANN model.
Table 5
Optimization of composition for different coal-biomass mixtures corresponding to specific criteria
A. For minimum chloride content in the mixture |
S. No. | Coal (%) | Pine wood charcoal (%) | Pine cone (%) | Pine needle (%) | Chloride (mg/kg) | HHV (kcal/kg) | Dry Ash (%) | Sulphur (mg/kg) | RSM optimization Desirability factor |
RSM predicted | ANN predicted | RSM predicted | ANN predicted | RSM predicted | ANN predicted | RSM predicted | ANN predicted |
1 | 50 | 10 | 30 | 10 | 162.61 | 173.4 | 3858.4 | 3857.6 | 32.13 | 36.1 | 3396.62 | 3465.7 | 0.99 |
2 | 60 | 10 | 20 | 10 | 163.62 | 167.9 | 3663.51 | 3787.7 | 38.10 | 37.7 | 3761.37 | 3568 | 0.98 |
3 | 56.67 | 16.67 | 16.67 | 10 | 169.1 | 151 | 3928.48 | 3938.9 | 36.03 | 38 | 3721.87 | 3645 | 0.96 |
B. For minimum sulphur content in the mixture |
1 | 50 | 10 | 30 | 10 | 162.61 | 173.4 | 3858.4 | 3857.6 | 32.13 | 36.1 | 3396.62 | 3465.7 | 0.95 |
C. For minimum dry ash content in the mixture |
1 | 50 | 10 | 10 | 30 | 389.08 | 365 | 3905.34 | 3575.2 | 31.86 | 34 | 3544.21 | 3247.3 | 0.88 |
2 | 50 | 14.08 | 10 | 25.92 | 346.41 | 329.2 | 4018.24 | 4087.7 | 31.87 | 34.8 | 3564.35 | 3625.8 | 0.88 |
3 | 50 | 30 | 10 | 10 | 180.05 | 250.1 | 4458.42 | 4503.6 | 31.89 | 32.4 | 3642.85 | 3559 | 0.88 |
D. For maximum HHV value content in the mixture |
1 | 50 | 30 | 10 | 10 | 180.05 | 250.1 | 4458.42 | 4503.6 | 31.89 | 32.4 | 3642.85 | 3559 | 0.92 |
E. For maximum HHV and minimum Chloride, Sulphur, and dry ash content in the mixture |
1 | 50 | 30 | 10 | 10 | 180.05 | 250.1 | 4458.42 | 4503.6 | 31.89 | 32.4 | 3642.85 | 3559 | 0.82 |
2 | 50 | 20 | 20 | 10 | 171.33 | 173.3 | 4158.41 | 4120.1 | 32.01 | 35.9 | 3519.74 | 3580.9 | 0.80 |
F. For HHV in the range of 3500–3700 kcal/kg, ash content in the range of 33–37% with minimum Chloride and Sulphur content in the mixture |
1 | 58.13 | 10 | 21.87 | 10 | 163.43 | 156.7 | 3700 | 3802.2 | 36.98 | 37.1 | 3693.07 | 3511.4 | 0.74 |
G. For minimum pinewood charcoal utilization to provide maximum HHV and minimum Chloride, Sulphur, and dry ash content in the mixture |
1 | 50 | 12.24 | 27.76 | 10 | 164.57 | 163.5 | 3925.73 | 3935.8 | 32.11 | 36.5 | 3424.25 | 3528.8 | 0.77 |
2 | 50 | 10 | 30 | 10 | 162.61 | 173.4 | 3858.41 | 3857.6 | 32.13 | 36.1 | 3396.62 | 3465.7 | 0.76 |
3 | 50 | 20 | 20 | 10 | 171.33 | 173.3 | 4158.41 | 4120.1 | 32.01 | 35.9 | 3519.74 | 3580.9 | 0.73 |
4 | 50 | 21.66 | 10 | 18.35 | 267.27 | 264.4 | 4227.65 | 4431.9 | 31.88 | 33.9 | 3601.70 | 3664.8 | 0.62 |
H. For minimum pinewood charcoal utilisation and maximum pine cone and pine needle utilization to provide maximum HHV and minimum Chloride, Sulphur, and dry ash content in the mixture |
1 | 50 | 10 | 23.49 | 16.52 | 236.39 | 191.3 | 3873.70 | 3792.5 | 32.05 | 34.7 | 3444.7 | 3341.1 | 0.63 |
I. For HHV in the range of 3500–3700 kcal/kg, ash content in the range of 33–37% with minimum Chloride and Sulphur in the mixture along with minimum pinewood charcoal utilization and maximum pine cone and pine needle utilization |
1 | 58.17 | 10 | 21.53 | 10.31 | 166.95 | 158.6 | 3700 | 3796.5 | 37 | 37.1 | 3696.72 | 3511.7 | 0.34 |
2 | 58.13 | 10 | 21.87 | 10 | 163.43 | 156.7 | 3699.96 | 3802.2 | 36.98 | 37.1 | 3693.16 | 3511.4 | 0.001 |
Using the ANN function developed as the fitness function, a multi-objective genetic algorithm (MOGA) was designed. The characteristics of MOGA are shown in Table 4. The developed Multi-Objective Genetic Algorithm (MOGA) features a 4-variable input and 4-output response configuration. With a population size of 200 individuals, the algorithm employs a crossover fraction of 0.8 and a mutation rate of 0.05. The crossover function, @crossoverintermediate, combines genetic material from parents, while @mutationadaptfeasible introduces adaptive mutations. Preserving the top two individuals as elite members maintains diversity. The fitness function integrates a previously developed Artificial Neural Network (ANN) model. Operating within specified upper and lower boundaries, the MOGA aims to optimize multi-objective functions efficiently, balancing exploration and exploitation to discover Pareto-optimal solutions.
The optimized composition of the mixture corresponding to the maximum calorific value and minimum chloride, sulphur, and ash content was determined using the multi-objective genetic algorithm. The MOGA revealed that 50% of coal, 30% of pine wood charcoal, and 10% of each pine cone and pine needle composition in the mixture will provide the maximum calorific value (4503.6 kcal/kg), and minimum sulphur (3559 mg/kg), chloride (250.1 mg/kg) and dry ash (32.4%) content attainable within the bounds of the mixture composition.
3.4 RSM-based optimization
Table 5 presents the optimization of the composition of different biomass types used in this study to meet specific criteria in the context of thermal power generation using the desirability function. A desirability value of 1 signifies the utmost level of desirability. The concept of desirability relates to the extent of closeness between a given response and its optimal value. In this specific investigation, the desirability function was employed to determine the ideal parameter configuration that maximized the overall desirability function across four response characteristics: chloride content, dry ash content, sulfur content, and higher heating value (HHV).
To achieve the lowest chloride content in the mixture, three combinations were considered. The most desirable combination (desirability of 0.99) consisted of 50% coal, 10% pinewood charcoal, 30% pine cone, and 10% pine needle, resulting in a chloride content of 162.61 mg/kg. Further, the sulphur content was also observed as minimum with the same combination as above. The composition has particularly higher percentages of Coal and Pine cones and lower percentages of Pinewood charcoal and Pine needles. This suggests that these components contribute greatly to reducing both chloride and sulphur levels in the mixture that addresses the sustainability issues greatly.
Furthermore, to ascertain the minimum dry ash content, three combinations were considered. The most desirable combination (desirability of 0.88) consisted of 50% coal, 10% pinewood charcoal, 10% pine cone, and 30% pine needle, resulting in a dry ash content of 31.86%. For maximum HHV, the most desirable combination (desirability of 0.92) consisted of 50% coal, 30% pinewood charcoal, 10% pine cone, and 10% pine needle, resulting in an HHV of 4458.42 kcal/kg. When considering maximum HHV as well as minimum chloride, sulphur, and dry ash content, the most desirable combination (desirability of 0.82) was the same as the one for maximum HHV.
In thermal power plants, generally, coal that is burned has HHV in the range of 3600 ± 100 kcal/kg and around 35 ± 2% ash [21]. Also, as per the technical specification specified for biomass pellets to be used for co-combustion with coal in thermal power plants, HHV is specified to be mimimum 3500 kcal/kg [22]. Thus, one specific criterion was fixed where HHV was set in the range of 3500–3700 kcal/kg, ash content in the range of 33–37% with minimum chloride, sulphur, and dry ash content, one combination was obtained. The desirable combination (desirability of 0.74) consisted of 58.13% coal, 10% pinewood charcoal, 21.87% pine cone, and 10% pine needle. The combination has chloride content as 163.43 mg/kg, calorific value as 3700 kcal/kg, ash content as 36.98% and sulphur content as 3693.07 mg/kg.
For minimum pinewood charcoal utilization to provide maximum HHV and minimum chloride, sulphur, and dry ash content, four combinations were considered. The most desirable combination (desirability of 0.77) consisted of 50% coal, 12.24% pinewood charcoal, 27.76% pine cone, and 10% pine needle. As an extrapolation, it was also investigated in another criterion where minimum pinewood charcoal utilization and maximum pine cone and pine needle utilization were set to provide maximum HHV and minimum chloride, sulphur, and dry ash content. It gave the most desirable combination (desirability of 0.63) consisting of 50% coal, 10% pinewood charcoal, 23.49% pine cone, and 16.52% pine needle.
Finally, For HHV in the range of 3500–3700 kcal/kg, ash content in the range of 33–37% with minimum Chloride, Sulphur, and Dry ash content in the mixture along with minimum pinewood charcoal utilization and maximum pine cone and pine needle utilization was investigated. This investigation revealed two different combinations with the most desirable (0.34) combination consisting of 58.17% coal, 10% pinewood charcoal, 21.53% pine cone, and 10.31% pine needle. This combination had a 3700 kcal/kg calorific value, ash content of 37%, chloride content of 166.95 mg/kg, and sulphur content of 3696.72 mg/kg.
Table 5 provides a comprehensive overview of how different biomass compositions can be optimized to meet specific criteria related to thermal power generation. The results highlight the importance of carefully selecting and balancing the biomass types to achieve the desired outcomes.