Anaerobic fermentation process soft measurement modeling method based on BOASSA-CNN-SVM

Abstract

The invention discloses an anaerobic fermentation soft measurement modeling method based on BOASSA-CNN-SVM. The method is used for accurately predicting VFA concentration and methane concentration which are difficult to directly measure in the anaerobic fermentation process. Comprising the following steps: 1) after data acquisition, screening sampling data of an anaerobic fermentation process with a time tag sequence of a process obtained through field operation or experiments by using a convolutional neural network (CNN) algorithm; 2) establishing a support vector machine (SVM) soft measurement model for an anaerobic fermentation process, and optimizing SVM model parameters by utilizing BOASSA; and 3) inputting the feature variables selected by the CNN into a BOASSA-SVM soft measurement model for modeling. The modeling method provided by the invention has the advantages of good estimation performance, high accuracy and strong generalization ability, and is also suitable for soft measurement modeling in other complex chemical reaction processes.

Description

Technical Field

[0001] The present invention relates to the field of soft sensing in the anaerobic fermentation process, and more specifically, to a soft sensing model for the anaerobic fermentation process based on BOASSA-CNN-SVM. Background Art

[0002] Energy shortage and environmental pollution caused by the combustion of fossil fuels have become global problems, and it has become extremely urgent to develop and utilize renewable clean energy. As a large agricultural country, China is rich in straw resources, and the potential of anaerobic digestion (AD) of straw lignocellulosic biomass into bio-methane is huge. This is an environmentally friendly alternative energy with high calorific value and combustion efficiency. This process is crucial for reducing straw burning, reducing carbon emissions, and solving the energy shortage problem.

[0003] The AD process includes four interrelated stages: hydrolysis, acidification, acetogenesis, and methanogenesis. Each stage requires the complex cooperation of various microorganisms. The biological process involves a series of reactions. Therefore, the performance of AD is affected by operating conditions, microbial types, and quantities. These factors are extremely complex and non-linear. Therefore, it is difficult to monitor the AD process.

[0004] With the development of computational algorithms and the accessibility of computing power, machine learning (ML) has become a new data mining technology and modeling tool and has been applied to the prediction of the AD process. It can achieve output prediction based on the potential interaction between input and output variables. Predicting AD performance through ML does not require knowledge of the process mechanism. Therefore, it is an effective method for predicting biogas production. Currently, various ML algorithms, including artificial neural networks (ANN), adaptive neuro-fuzzy inference systems (ANFIS), and random forests (RF), have been used to simulate the complex non-linear relationships in AD progress.

[0005] Previous studies have demonstrated the positive role of machine learning (ML) in predicting and monitoring the anaerobic fermentation (AD) process, but there are still deficiencies. First, existing research mainly focuses on the performance evaluation of the ML model itself or the performance comparison between different models, while less attention is paid to how to further improve the prediction ability of the ML model through optimization algorithms. This results in limited optimization of the model in practical applications and makes it difficult to fully explore its prediction potential. Second, most current studies only validate the performance of the ML model based on a single or specific dataset, and this approach may limit the generalization ability of the model in different scenarios and conditions. Finally, most studies only focus on the prediction of bio-methane production in the AD process, while ignoring the importance of building a comprehensive early warning and monitoring system for the operation of the AD system, which to a certain extent limits the comprehensive application potential of ML technology in AD process management.

[0006] To address these deficiencies, the present invention proposes an innovative solution, namely a CNN-SVM model combined with a Sparrow Search Algorithm improved by Butterfly Optimization Algorithm (BOASSA). As an efficient optimization algorithm, BOASSA can significantly enhance the parameter optimization effect of ML models, thereby improving the prediction accuracy and robustness of the models. At the same time, the present invention adopts a cross-validation strategy to enhance the stability and reliability of the model on different datasets, thus overcoming the problem of model one-sidedness that may be caused by a single dataset. In addition, the present invention not only focuses on the prediction of biogas production, but also endeavors to construct a comprehensive early warning and monitoring system for the operation of the AD system. This system can monitor the key parameters in real time during the AD process, give early warnings of abnormal states, and comprehensively evaluate the performance of the system, providing more comprehensive and accurate technical support for the efficient, safe, and intelligent management of the AD process. Summary of the Invention

[0007] The object of the present invention is to overcome the deficiencies of the prior art and provide a soft-sensing modeling method for anaerobic fermentation process based on BOASSA-CNN-SVM.

[0008] The soft-sensing modeling method for anaerobic fermentation based on BOASSA-CNN-SVM of the present invention includes the following steps:

[0009] 1) Obtain sampling data of the anaerobic fermentation process with time-tagged sequences through on-site operation or experiments, and extract key feature variables from the collected anaerobic fermentation data by a Convolutional Neural Network (CNN) to reduce the dimension of each data;

[0010] 2) Divide the sample data of the extracted basic feature variables into a training set and a test set, and perform normalization processing to eliminate the influence of the dimension of different feature variables;

[0011] 3) Set the initial parameters of the Sparrow Search Algorithm (SSA), namely the number of iterations and the population size, initialize the population using cubic mapping, and improve the position update strategy of the global search stage of the SSA using the Butterfly Optimization Algorithm (BOA) to modify the position formula of the discoverer in the SSA;

[0012] 4) Construct a Support Vector Machine (SVM) soft-sensing model CNN-SVM for the anaerobic fermentation process with the normalized data as the input, and use the objective function that minimizes an expression containing two terms, namely the training error and the regularization loss, as the fitness function; use BOASSA to search for the optimal parameters of the CNN-SVM model, and these parameters include the penalty coefficient, the kernel function scale, and the tolerance coefficient;

[0013] 5) The obtained optimal parameter set is substituted into CNN-SVM to form a high-precision soft sensing model of anaerobic fermentation process, BOASSA-CNN-SVM, which can predict the concentrations of volatile fatty acids (VFA) and methane and the production of biogas.

[0014] The BOASSA-CNN-SVM-based anaerobic fermentation process soft measurement modeling method comprises three stages: discoverer position update, follower position update, and reconnaissance and early warning.

[0015] During the finder position update phase, finders with better fitness values are given priority in obtaining food during the search process. Because the finder is responsible for finding food for the entire sparrow population and providing foraging directions for all followers, the finder can obtain a larger foraging search range than the followers. Using BOA's global search strategy to update the finder's position expands the search space to a certain extent and improves the algorithm's global search capability. The improved finder position update formula is as follows:

[0016]

[0017] Where R2 and ST represent the warning value and safety threshold respectively; Q is a random number that obeys the normal distribution with mean 0 and variance 1; L represents a 1×d matrix with each element being 1, and d is the dimension of the search space; when R2 <ST时,这意味着此时的觅食环境周围没有捕食者,发现者可以执行广泛的搜索操作;当R2≥ST时,这表示种群中的一些麻雀已经发现了捕食者,并向种群中其他麻雀发出了警报,此时所有麻雀都需要迅速飞到其他安全的地方进行觅食;

[0018] 2) Follower position update phase. The mathematical expression for this phase is:

[0019]

[0020] in The best position currently occupied by the discoverer; Indicates the current global worst position; A represents a 1×d matrix in which each element is randomly assigned a value of 1 or -1. + =A T (AA T ) -1 ; where is the pseudo inverse matrix. When i <n / 2时,这表明适应度值较低的第i个追随者没有获得食物,处于十分饥饿的状态,此时需要飞往其他地方觅食,以获得更多的能量;

[0021] 3) During the reconnaissance and early warning stage, when danger is sensed, the scouts in the sparrow population will exhibit anti-predation behaviors. The mathematical expression for this stage is:

[0022]

[0023] where X best is the current global optimal position; β is the step size control parameter, a random number following a normal distribution with a mean of 0 and a variance of 1; K is a random number representing the direction of sparrow movement and also the step size control parameter, and K ∈ [-1, 1]; f i is the fitness value of the current sparrow individual; f g and f w are the current global best and worst fitness values respectively; ε is the smallest constant to avoid a zero denominator. For simplicity, when f i < f g , it indicates that the sparrow is at the edge of the population at this time and is extremely vulnerable to predators; when f i = f g , this indicates that the sparrows in the middle of the population are aware of the danger and need to get closer to other sparrows to minimize their risk of being preyed upon.

[0024] In the soft sensor modeling method for anaerobic fermentation methane production process based on BOASSA-CNN-SVM, the main purpose of the support vector machine (SVM) is to find an optimal hyperplane such that the distance between the closest separating surface of different data and the hyperplane is maximized, that is, to solve the following problem:

[0025]

[0026] where ω is the normal vector of the hyperplane; x i is the training sample; y i is the category of the sample; b is the threshold determined according to the training sample; C is the penalty parameter; ε i is the slack variable introduced when it is linearly inseparable; introducing the Lagrange multiplier α i , then the Lagrangian function of this problem can be written as equation (5), and the partial derivatives of ω and b in equation (5) are set to 0.

[0027]

[0028] Therefore, its dual form is:

[0029]

[0030] In the non-linear case, introducing the kernel function mapping gives:

[0031]

[0032] The decision function derived from the above derivation is as follows:

[0033]

[0034] where y i is the corresponding expected output; x j is the input vector; n is the number of training samples; is the mapping function; K(x i , x j ) is the kernel function.

[0035] The SSA parameter settings of the described soft sensor modeling method for the anaerobic fermentation methane production process based on BOASSA-CNN-SVM include the number of iterations and the population size. The parameter settings of SVM include the penalty coefficient, the kernel function scale, and the fault tolerance coefficient. The steps for BOASSA to optimize the SVM parameters are as follows:

[0036] 1) Initialize SVM and initialize the population of SSA using cubic mapping;

[0037] 2) Use the selected variables as input and output to train SVM;

[0038] 3) According to the population size, randomly select a part of the individuals from the population to be given the role of predators, and randomly select a part of the remaining individuals as followers according to the proportion of followers. Finally, the scouts will be selected from the remaining individuals;

[0039] ]>4) Use the SSA improved by BOA for global search. The predators will actively explore new solutions in the search space according to the global search strategy and update their positions; the expression in this stage is:

[0040]

[0041] The followers will search around the high-quality solutions found by the predators according to the local search strategy and adjust their positions accordingly. The expression in this stage is:

[0042]

[0043] When aware of danger, the scouts in the sparrow population will perform anti-predation behaviors. The expression in this stage is:

[0044]

[0045] 5) Accurately calculate the fitness of each sparrow individual as the criterion for evaluating its performance. If the fitness function is satisfied, go to step 6); otherwise, repeat step 4);

[0046] 6) If the number of iterations is met, the optimal parameter solution is assigned to the SVM, otherwise repeat step 4). BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is the flow chart of the BOASSA-SVM soft sensor model;

[0048] Figure 2 It is the overall flow chart of the soft measurement system;

[0049] Figure 3 This is the methane concentration prediction result of the BOASSA-CNN-SVM soft sensing model.

[0050] Figure 4 This is the VFA concentration prediction result of the BOASSA-CNN-SVM soft sensing model. DETAILED DESCRIPTION

[0051] In conjunction with the embodiment, the technical solution of this modeling is clearly and completely described, and the soft measurement modeling method of the anaerobic fermentation process based on BOASSA-CNN-SVM is used to perform soft measurement modeling of the actual same or similar process, including the following steps:

[0052] 1) 477 sets of sampling data were collected from the anaerobic fermentation process of a food waste treatment plant through on-site operation, including percentage of solids (TS), pH, percentage of volatile suspended solids (VS), chemical oxygen demand (COD), average flow rate, alkalinity (Alk), percentage of CO2, daily gas production, percentage of CH4, and laboratory values of VFA.

[0053] 2) Using convolutional neural networks (CNNs) to extract basic features from the collected anaerobic fermentation data as input variables;

[0054] 3) Take out 314 and 163 groups of data as training set and test set respectively for the selected input variable sample data, and then normalize them to eliminate the influence of units and dimensions of different variables. Select the mapping space (0, 1) and use As a normalization criterion, where X0 is the historical data, X i is the minimum value in the historical data, X a is the maximum value, and X is the data sample after normalization;

[0055] 4) Set the parameters of the BOASSA algorithm and SVM: the number of BOASSA iterations is 10, the number of individuals in the sparrow population is 6, the optimized parameter dimension is 3, and the kernel function of the SVM is the Gaussian kernel function;

[0056] 5) Use BOASSA to optimize the parameters of SVM and assign the optimal parameter solution to SVM.

[0057] 6) Run the proposed BOASSA-CNN-SVM model to extract the time series features in the samples and use them to predict the concentrations of VFA and methane and methane production.

[0058] The soft sensor model established by this method has a good prediction accuracy for the VFA concentration in the embodiment, and the prediction results are shown in the attached figure. Figure 4 The results show that the established VFA concentration soft-sensing model can estimate VFA concentration more accurately in practical applications and has broad application prospects in the field of monitoring and controlling the anaerobic fermentation process of food waste treatment.

Claims

1. A soft sensor modeling method for anaerobic fermentation process based on BOASSA-CNN-SVM, characterized by including The following steps: 1) Obtain sampling data of the anaerobic fermentation process with a time-tagged sequence through on-site operation or experiment, and extract key feature variables from the collected anaerobic fermentation data by means of a Convolutional Neural Network (CNN) to reduce the dimension of each piece of data; 2) Divide the sample data of the extracted basic feature variables into a training set and a test set, and perform normalization processing to eliminate the influence of the dimensions of different feature variables; 3) Set the initial parameters of the Sparrow Search Algorithm (SSA), namely the number of iterations and the population size, initialize the population using cubic mapping, and improve the position update formula of the discoverer in SSA using the position update strategy in the global search stage of the Butterfly Optimization Algorithm (BOA); 4) Construct a Soft Sensor Model CNN-SVM for the anaerobic fermentation process with the normalized data as input, and use the objective function that minimizes the expression containing two items, the training error and the regularization loss, as the fitness function; use BOA-SSA to search for the optimal parameters of the CNN-SVM model, and these parameters include the penalty coefficient, the kernel function scale, and the tolerance coefficient; 5) Substitute the obtained optimal parameter set into CNN-SVM to form a high-precision Soft Sensor Model BOA-SSA-CNN-SVM for the anaerobic fermentation process, and predict the concentrations of volatile fatty acids (VFA) and methane and the biogas production.

2. The soft sensor modeling method for anaerobic fermentation process based on BOASSA-CNN-SVM according to claim 1, wherein Each piece of sampling data of the anaerobic fermentation process includes the fermentation time, the characteristics of the fermentation substrate, the concentrations of various components, the VFA concentration, and the biogas production; use CNN to extract key features from the data of the anaerobic fermentation process.

3. A soft sensor modeling method for anaerobic fermentation process based on BOASSA-CNN-SVM according to claim 1, characterized in that The normalization method adopted is to map the data to a specified interval, and the mapping space is selected as [0, 1]. is used as the normalization criterion, where X0 is the historical data, and X i is the vector composed of the minimum values of each variable in the historical data, and X a is the vector composed of the maximum values of each variable in the historical data, and X is the data sample vector after normalization.

4. A soft sensor modeling method for anaerobic fermentation process based on BOASSA-CNN-SVM according to claim 1, characterized in that The cubic mapping uses the following formula to construct an efficient and uniform initial population distribution by utilizing the characteristics of randomness, ergodicity, and regularity of the chaotic sequence, and improve the quality of the population to optimize the initialization strategy of SSA: y(n + 1) = 4y(n) 3 -3y(n) (1) where n is the number of mappings; y(n) is the value of the nth mapping, and y(n) ∈ (-1, 0) ∪ (0, 1).

5. A soft sensor modeling method for anaerobic fermentation process based on BOASSA-CNN-SVM according to claim 1, characterized in that The SSA algorithm uses the following formula to update the position of the discoverer by utilizing the global search strategy of BOA, which expands the search space to a certain extent and enhances the global search ability of the algorithm: Among them is the position of the $i$-th butterfly in the $t$-th iteration process; $X$ best is the current global optimal position; $f$ i is the smell emitted by the $i$-th butterfly; Its value depends on the size of the fitness; r is a random number in [0, 1].

6. The soft-sensing modeling method for anaerobic fermentation process based on BOASSA-CNN-SVM according to claim 1, wherein The SSA is used to optimize the SVM network parameters, and the specific steps are as follows: 1) Initialize SVM, and initialize the population of SSA using cubic mapping; 2) Use the selected variables as input and output to train SVM; 3) According to the population size, randomly select a part of the individuals from the population to be given the role of predators, and randomly select a part of the remaining individuals as followers according to the proportion of followers, and finally the scouts will be selected from the remaining individuals; 4) Use the improved SSA based on BOA for global search. The predators will actively explore new solutions in the search space according to the global search strategy and update their positions; the expression in this stage is: where R2 and ST represent the early warning value and the safety valve value respectively; Q is a random number subject to a normal distribution with a mean of 0 and a variance of 1; L represents a 1×d matrix with all elements being 1, and d is the dimension of the search space; Followers search around the high-quality solutions discovered by the predators according to the local search strategy and adjust their positions accordingly; the expression in this stage is: Where X ip is the optimal position occupied by the current discoverer; represents the worst position currently stored; A represents a 1×d matrix in which each element is randomly assigned 1 or -1, A + = A T (AA T ) -1 ; When aware of danger, scouts in the sparrow population will perform anti-predation behaviors, and the expression in this stage is: Among them is the globally optimal position currently searched for; β is a step size control parameter, a random number that follows a normal distribution with a mean of 0 and a variance of 1; K is a random number that represents the direction of sparrow movement and is also a step size control parameter, and K ∈ [-1, 1]; f i is the fitness value of the current sparrow individual; f g and f w are the current global best and worst fitness values respectively; ε is the smallest constant to avoid a zero denominator; 5) Accurately calculate the fitness of each sparrow individual as the criterion for evaluating its performance; if the fitness function is satisfied, go to step 6), otherwise repeat step 4); 6) If the iteration times are met, assign the obtained optimal solution of the parameters to the SVM, otherwise repeat step 4).

Images