Modeling of Carbon Dioxide Fixation Rate by Micro Algae Using Hybrid Arti�cial Intelligence and Fuzzy Logic Methods and Optimization by Genetic Algorithm

10 In this study we are reporting a prediction model for the estimation of carbon dioxide (CO 2 ) 11 fixation based on the Adaptive Neuro-Fuzzy Inference System (ANFIS) and Genetic Algorithm 12 (GA) hybrid approach. The experimental parameters such as temperature and pH conditions of the 13 micro-algae-based carbon dioxide uptake process were taken as the input variables and the 14 CO 2 fixation rate was taken as the output variable. The optimization of ANFIS parameters and 15 formation of the model structure were performed by genetic algorithm (GA) algorithm in order to 16 achieve optimum prediction capability and industrial applicability. The best-fitting model was 17 figured out using statistical analysis parameters such as RMSE, R 2 and AARD. According to the 18 analysis, GA-ANFIS model depicted a superior prediction capability over ANFIS optimized 19 model. The Root Mean Square Error (RMSE), coefficient of determination (R 2 ) and AARD for 20 GA-ANFIS were determined as 0.000431, 0.97865 and 0.044354 in the training phase and 21 0.00056, 0.98457 and 0.032156 in the testing phase, respectively for the GA-ANFIS Model. As a 22 result, it can be concluded that the proposed GA-ANFIS model is an efficient technique having 23 very high potential to accurately calculate CO 2 fixation rate and the exploration of the industrial 24 scale-up process for commercial activities. 25


1.
Introduction 28 It is evident that the rising in the level of gas emissions caused higher global temperatures which 29 have led to the increase in frequency and the scale of the natural disasters (Xu et   The concept of natural selection and hereditary behavior based genetic algorithms (GAs) are 98 nowadays predominantly employed for the optimization than the other optimization techniques.

99
The potentiality of the binary digit control system also causes the genetic algorithms distinctive   Adaptive neuro-fuzzy inference system is the combination of quick-witted techniques of neural  Figure 3 represents the basic ANFIS model structure using MATLAB R2020a.  The representation of rules for Sugeno fuzzy inference system is given below in the Rule 1: If x is A1 and y is B1, then f1 = p1x + q1y + r1 (1) 162 Rule 2: If x is A2 and y is B2, then f2 = p2x + q2y + r2 (2) 163 Where, x and y are the two inputs. A and B are the membership function. p, q, and r linear 164 parameter. 1,2, represent the number of rules. Nodes of this layer execute the function of multiplying the inputs and passing out the product.

185
The output of various nodes depicts the strong firing of the rule.  This single node is a circular node with the symbol Σ (sigma) that is employed to calculate the 196 overall summation of all the incoming signals.  suggests that the best model is considered as the one which has got the lowest testing data error.

293
The ANFIS methodology was employed to forecast the CO2 fixation rate and to find out the 294 relation between the input variables and output variables.  Table   306 S4 (see the supporting information). The Number of fuzzy rules was 9 which are shown in Table   307 2. The values of the output membership functions for the rules are given in

316
One of the benefits of ANFIS is that it has the capability to forecast a certain parameter for different 317 inputs within the training data range. Hence, in this research, the data from various resources were 318 taken into account as ANFIS model's inputs to forecast the selectivity as shown in Figure 9   Additionally, the Figure 10 illustrates the CO2 fixation rate versus temperature and pH. As In this study, using the programming language of MATLAB R2020a, GA-ANFIS hybrid approach 337 was propounded for CO2 fixation rate prediction. As the first step, the parameters (Temperature 338 and pH) were set as input factors, and CO2 fixation rate was set as output factor for the ANFIS 339 Models. The results from ANFIS shows that Training and Testing datasets have an R 2 value of 340 0.92345 and 0.91724 respectively as shown in Table 3. 341 In order to increase the prediction potentiality of the ANFIS model, hybrid models were formed

376
The performance of a GA is affected by the diversity of the initial population as shown in Figure   377 11a and gives its genes to multiple children. reproduction and this process will advance very deliberately. Table 3 indicates the R 2 , RMSE and 405 AARD values for the given ANFIS and GA-ANFIS models. The results clearly show that GA-

406
ANFIS have better performance compared to ANFIS models as shown in Table 3 and Figure S1

410
The higher CO2 fixation rate of micro-algae is of great significance in determining the potential of 411 biomass and its possible application areas for industrial purpose. In this study, the effect of two