A feedback feeding control method and system based on multi-component concentration online detection

Abstract

The application discloses a feedback feeding control method based on multi-component concentration online detection, which is used for producing ethanol by using Saccharomyces cerevisiae and comprises the following steps: S1, in-situ online sampling is performed, and online monitoring instruments are used to detect the substrate and product concentrations in a fermentation tank in real time; S2, a kinetic model based on substrate consumption is established according to the detected substrate, product and predicted cell concentration of a soft measurement model; S3, a particle swarm algorithm is combined to optimize the parameters of the substrate consumption kinetics, and the optimized model is used as a reference track; S4, the output of a rolling optimization concentration feedback controller is optimized; and S5, the output of the controller is calculated according to a single neuron adaptive PID feedback controller based on a quadratic performance index, so that the optimal set track is quickly tracked, and the substrate concentration in the fermentation tank is maintained at a required concentration. Experimental results prove that the feedback feeding control based on the online detection of the component concentration is beneficial to the utilization rate of the substrate and the generation of the product, and meanwhile, the method can also be applied to other actual fermentations.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of microbial fermentation control, and particularly relates to a feedback feeding control method and system based on online detection of multi-component concentration for ethanol fermentation. BACKGROUND

[0002] In the past few years, many researchers have begun to invest a lot of enthusiasm in bioenergy, which is an environmentally friendly energy material produced from renewable resources, and ethanol, as an important green, clean and renewable bioenergy, has been widely used in many fields.

[0003] Ethanol production is a complex system and presents a nonlinear and dynamic process, which is affected by several key process parameters, such as temperature, pH, dissolved oxygen, cell concentration, component (sugar) concentration or product (ethanol). At present, the environmental parameters such as temperature, pressure, pH and dissolved oxygen in the fermentation process can be measured online and continuously by related sensors, and the glucose as the component of ethanol fermentation is still analyzed by offline means, so that the operator cannot easily grasp the fermentation situation in time and thus produces a lag phenomenon. The cell metabolism needs a suitable component concentration, which is a prerequisite for product synthesis. If the component concentration is too high, inhibition will occur. If the substrate concentration is too low, the product synthesis process cannot proceed due to lack of metabolic substances, so it is crucial to maintain the substrate concentration in the fermentation broth within a suitable range during the fermentation process.

[0004] With the increasing scale of fermentation production process, the traditional manual feeding control relying on experience cannot meet the needs of the fermentation process. In order to solve the shortcomings of the traditional feeding control method, truly realize online high-performance feeding control of the fermentation process, reduce the labor intensity of manual operation, improve the automation level and production efficiency, and at the same time improve the fermentation efficiency, it is urgent to adopt a new control method design scheme. SUMMARY

[0005] Therefore, the present application provides a feedback feeding control method and system based on online detection of multi-component concentration, which is used for online regulation and control of feeding rate, so that the component concentration in the fermentation tank is maintained within the required range.

[0006] The present application first provides a feedback feeding control method based on online detection of multi-component concentration, which is used for production of ethanol by Saccharomyces cerevisiae. The feedback method comprises the following steps:

[0007] S1, in-situ online sampling, real-time detection of component (glucose, ethanol) concentration in the fermentation tank by using an online monitoring instrument, and obtaining the detected substrate and product concentration;

[0008] S2, according to the detected substrate, product concentration and soft measurement model predicted cell concentration to establish based on the kinetics of substrate consumption model;

[0009] S3, combined with particle swarm algorithm to optimize the parameters of the kinetics model of S2, the optimized substrate consumption kinetics model as a reference trajectory, and real-time output, as the optimal trajectory;

[0010] S4, according to step S1 real-time detection of substrate concentration and step S3 of the optimal set trajectory error, the output of the rolling optimization concentration feedback controller;

[0011] S5, according to the single neuron adaptive PID feedback controller based on the quadratic performance index, the output u of the controller is calculated, the feed control amount u(k) should be added to the system at the current time is calculated, and is applied to the feed pump; the feed pump is connected with the fermentation tank to maintain the substrate concentration in the fermentation tank in the required concentration range.

[0012] Preferably, the method further comprises:

[0013] The component concentration detection of step S1 is by using the principle of electrochemical enzyme biosensor.

[0014] Preferably, the kinetics model in step S2 is described by the following differential equation:

[0015]

[0016]

[0017]

[0018]

[0019] In the formula, D, X, S, P and S f respectively represent dilution rate, cell concentration, component concentration, product concentration and feed liquid concentration, μ represents the specific growth rate of the cell, μ m , P m , K m , K in and Y x / s respectively the maximum specific growth rate of the cell, the product saturation coefficient, the component saturation coefficient, the component inhibition constant and the cell yield coefficient of the component; α, β are both kinetic parameters;

[0020] In the formula, X, S, P, V, F and S frespectively represent the cell concentration, substrate concentration, metabolite product, fermentation broth volume, feeding liquid flow rate and feeding liquid concentration. μ(S), ρ(S), γ(S) represent the specific growth rate of the cell, the specific production rate of the metabolite product and the specific consumption rate of the substrate for maintaining metabolism, respectively;

[0021] Preferably, the cell concentration in step S2 is obtained by constructing a soft measurement model based on a BP neural network through real-time detection of environmental parameters such as pH, T, Do, substrate concentration and product concentration in the fermenter, the BP neural network comprising three layers of structure of an input layer, a hidden layer and an output layer, and the mathematical expression of the network being:

[0022]

[0023] In the formula, X k represents an input vector, α i is the hidden layer output, y j is an output parameter, v ik and w ji are the connection weights of the hidden layer and the output layer, and f and g are the activation functions of the hidden layer and the output layer, respectively.

[0024] Preferably, the particle swarm algorithm in step S3 has the following specific steps:

[0025] S3-1: taking the parameters to be optimized as a particle, initializing the particle swarm size NP, the iteration number of the particle swarm algorithm NG, initializing the particle position and velocity, giving the weighting coefficient and learning factor, calculating the fitness of each particle and initializing the global optimal solution and individual optimal solution:

[0026] S3-2: updating the objective function value of each particle and taking it as the fitness value corresponding to each particle, and the objective function value formula being

[0027]

[0028] In the formula, X i , P i , S i are the calculated values of the kinetic model of the cell concentration, product concentration and substrate concentration, X′ i , P′ i , S′ i are the measured values of the online monitoring instrument in the fermentation process.

[0029] S3-3: updating the individual optimal solution pbest and the global optimal solution gbest according to the fitness function;

[0030] S3-4: updating the particle swarm, and the motion equation of the particle swarm being as follows:

[0031] s i，j(T+1) = λs i，j (T) + c1r1(p i，j -z i，j ) + c2r2(p g，j -z i，j )

[0032] z i，j (T+1) = z i，j (T) + s i，j (T+1)

[0033]

[0034] where i is the particle number i = 1, 2,..., N, j is the particle dimension, j = 1, 2,..., d; T is the frequency of iteration replacement, T max is the highest frequency of iteration replacement, λ is the inertia motion weight vector factor, λ max and λ min are the maximum and minimum inertia motion weight vector factors respectively; c1 and c2 are random constants of moving acceleration, r1 and r2 are random constants of 0-1; s i，j (T) is the moving speed of the i-th particle in the j-th dimension in the T-th iteration replacement; z i，j (T) is the position of the i-th particle in the j-th dimension in the T-th iteration replacement;

[0035] S3-5 checks the termination condition, if the maximum iteration number is reached or the optimal solution no longer changes, the iteration is terminated; otherwise, return to step S3-2 to continue repeating the optimization.

[0036] Preferably, the single neuron adaptive PID controller based on quadratic performance index, the output u(k) adopts the following algorithm:

[0037]

[0038] where w i (k) (i = 1, 2, 3) is a weighting coefficient,

[0039]

[0040]

[0041]

[0042] where,

[0043] x1(k) = e(k)

[0044] x2(k) = Δe(k) = e(k) - e(k-1)

[0045] x3 = Δ 2 e(k) = e(k) - 2e(k-1) + e(k-2)

[0046] b0 = y(1)

[0047] wherein η I , η P , η D are integral, proportional and differential learning rates respectively, P and Q are weighting coefficients of output error and control increment respectively, K is a proportional coefficient of neuron, w i (k) is a weighting coefficient of neuron, e(k) is a system error, b0 is an initial value of output response, w′ i (k) is a weighting coefficient of neuron, x i (k) is a weighting value, Δe(k) represents a system error increment, and y(1) represents a first value of output of the controlled system.

[0048] The application also provides a control system of the feedback feeding control method based on online detection of component concentration, which comprises a programmable controller, a fermentation tank, an in-situ sampler, an online monitoring instrument, a feeding pump and an upper monitoring station, wherein:

[0049] The programmable controller (PLC) is connected with the fermentation tank and the upper monitoring station, and is used for adjusting parameters such as temperature T, pH value and dissolved oxygen concentration Do in the fermentation process, and uploading real-time detection data to the upper monitoring station;

[0050] The fermentation tank is provided with a temperature sensor, a Do sensing electrode and a pH sensing electrode, which are respectively used for detecting environmental parameters such as temperature, dissolved oxygen concentration Do, pH value and air flow in the ethanol fermentation process;

[0051] The in-situ sampler is connected with the fermentation tank, and a ceramic membrane is arranged on the sampler, which can remove macromolecules such as bacteria, so that small molecules such as glucose and water pass through the sampler;

[0052] The online monitoring instrument is provided with electrochemical sensors of glucose and ethanol, and is connected with the in-situ sampler, which is used for obtaining the glucose concentration in the fermentation liquor;

[0053] The feeding pump is connected with the online monitoring instrument, and is used for flowing feeding medium into the fermentation tank;

[0054] The upper monitoring station as a master station is connected with the online monitoring instrument and the programmable controller, and is embedded with a control algorithm, which is used for receiving, processing and sending instructions, setting fermentation parameters, and displaying and exporting data.

[0055] Beneficial effects: the method provided by the application can establish a feedback feeding control method based on multi-component concentration online detection, which combines the real-time detection method of the online monitoring instrument with the method of predicting concentration by soft measurement model, is used for real-time optimization of the fermentation mechanism model, designs a single neuron adaptive concentration feedback controller based on a quadratic performance index, and the control performance quickly tracks the optimal trajectory concentration; in addition, a feedback control system is designed, which realizes online detection and feedback control of glucose and ethanol which are difficult to measure in the fermentation process. Through test verification, the application realizes automatic feeding control of Saccharomyces cerevisiae producing ethanol and is beneficial to the production of ethanol. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A feedback feeding control method for ethanol fermentation provided by the application is shown in the figure;

[0057] Figure 2 A BP neural network cell soft measurement model described in the application is shown in the figure;

[0058] Figure 3 A feedback feeding control system structure diagram for ethanol fermentation provided by the application is shown in the figure;

[0059] Figure 4 A simulation diagram of the step response as a tracking signal under the method of the application

[0060] Figure 5 A tracking effect diagram of the whole system under the method of the application DETAILED DESCRIPTION

[0061] The application will be further described below in combination with the drawings and specific implementation cases. The following implementation cases are used to illustrate the application, but not to limit the scope of the application.

[0062] In this implementation case, the method and system of the application are used to detect and feedback control the component concentration in the fermentation tank by using Saccharomyces cerevisiae to produce ethanol.

[0063] The application provides a feedback feeding control method based on multi-component concentration online detection, which is used for Saccharomyces cerevisiae to produce ethanol, and the feedback method comprises the following steps:

[0064] 1. Based on the fermentation process of Saccharomyces cerevisiae producing ethanol in the past, 10 batches of historical data are counted, including the cell concentration, substrate concentration and product concentration at each time, and the temperature, pH value and Do value at the corresponding time; the fermentation data is randomly divided into two parts, one part is a training data set, and the other part is a test data set, which is used to build a BP neural network soft measurement model, and the BP neural network soft measurement model and the fermentation mechanism are modeled by Matlab software.

[0065] 2. The application provides a feedback feeding control method based on multi-component concentration online detection, as shown in the following steps: Figure 1

[0066] S1 In-situ online sampling, real-time detection of substrate (glucose) and product (ethanol) concentrations in the fermenter by online monitoring instrument;

[0067] S2 Real-time acquisition of environmental parameters in the fermentation process, establishment of a soft-sensing model based on BP neural network to predict the cell concentration, and a BP neural network soft-sensing model is shown in the following figure: Figure 2 The BP neural network includes an input layer, a hidden layer and an output layer, and the mathematical expression of the network is as follows:

[0068]

[0069] In the formula, X k represents the input vector, α i is the hidden layer output, y j is the output parameter, v ik and w ji are the connection weights between the hidden layer and the output layer, f and g are the activation functions of the hidden layer and the output layer respectively, and both are single-stage Sigmoid functions.

[0070] The input of the neural network is pH(t), dissolved oxygen Do(t), temperature T(t), substrate concentration S(t) and product concentration P(t), and the output is the cell concentration X(t). The number of hidden layer neurons of the BP neural network is usually between 1 and 10, and through performance detection of the network, the number of hidden layers is 6.

[0071] According to the collected substrate concentration, cell concentration and product concentration, the established fermentation mechanism model is updated in real time, and the particle swarm algorithm is combined to optimize the parameters of the mechanism model, and the optimized component consumption kinetic model is used as the optimal set trajectory;

[0072] In the mechanism model of ethanol fermentation, the changes of cell concentration, substrate concentration and product concentration in the fermentation process are briefly described as follows:

[0073]

[0074]

[0075]

[0076]

[0077] In the formula, D, X, S, P and S f ​respectively represent dilution rate, cell concentration, component concentration, product concentration and feed concentration, μ represents specific growth rate of the cell, μ m , P m , K m , K in and Y x / s respectively represent maximum specific growth rate of the cell, product saturation coefficient, component saturation coefficient, component inhibition constant and yield coefficient of the cell to the component; α and β are kinetic parameters;

[0078] 4. In the present application, the particle swarm algorithm is used to obtain some parameters (μ m , P m , K m , K in and Y x / s ) which need to be optimized in real time in the kinetic model. The particle swarm algorithm is used to optimize these parameters, and the optimal parameters which minimize the error are obtained.

[0079] Preferably, the particle swarm algorithm of step S3 has the following specific steps:

[0080] S3-1: the five parameters to be optimized are taken as a particle, the size of the particle swarm NP is initialized, the iteration number of the particle swarm algorithm NG is initialized, the particle position and velocity are initialized, the weighting coefficient and learning factor are given, the fitness of each particle is calculated, and the global optimal solution and individual optimal solution are initialized:

[0081] S3-2: the objective function value of each particle is updated, and it is taken as the fitness value corresponding to each particle. The objective function value formula is

[0082]

[0083] In the formula, X i , P i , S i are the calculated values of the kinetic model of cell concentration, product concentration and substrate concentration, and X′ i , P′ i , S′ i are the measured values of the fermentation process.

[0084] S3-3: the individual optimal solution pbest and the global optimal solution gbest are updated according to the fitness;

[0085] S3-4: the particle swarm is updated, and the motion equation of the particle swarm is as follows:

[0086] s i，j (T+1)=λs i，j (T)+c1r1(p i，j -z i，j )+c2r2(p g，j -zi，j )z i，j (T+1)=z i，j (T)+s i，j (T+1)

[0087]

[0088] where i is the particle number i = 1, 2,..., N, j is the particle dimension, j = 1, 2,..., d;

[0089] T is the frequency of iteration replacement, T max is the maximum frequency of iteration replacement, λ is the inertial motion weight vector factor, λ max and λ min are the maximum and minimum inertial motion weight vector factors, respectively; c1 and c2 are random constants of moving acceleration, r1 and r2 are random constants of 0-1; s i，j (T) is the moving speed of the i-th particle in the j-th dimension in the T-th iteration replacement; z i，j (T) is the position of the i-th particle in the j-th dimension in the T-th iteration replacement;

[0090] S3-5 checks the termination condition, if the maximum iteration number is reached or the optimal solution no longer changes, the iteration is terminated; otherwise, return to step S3-2 to continue repeating the optimization.

[0091] The role of introducing the reference input trajectory is to make the system output smoothly reach the set value along the set trajectory, and the optimal reference trajectory generally adopts the following form:

[0092]

[0093] where y r (k) is the reference input at time k, y sp (k) is the value of the dynamic model at time k, is the set input value of the system at time k, and α is a regulation factor related to the convergence and robustness of the control system, 0 < α < 1.

[0094] The deviation of the system is defined as: e(k) = y r (k) - y(k)

[0095] The application of neural networks in control system design is mainly aimed at the nonlinearity, uncertainty and complexity of the system. Due to the adaptability and robustness of neural networks, the entire control system has stronger rapid tracking ability and stability. Neurons, as the basic unit of neural networks, have the ability of self-learning and self-adaptation, so combined with PID control, it can make up for the shortcomings of traditional PID regulators, such as not easy to adjust parameters online in real time and not able to effectively control some complex processes.

[0096] In optimal control theory, the quadratic performance index is used to calculate the control rate to achieve the desired optimization effect. In the neuron learning algorithm, the quadratic performance index in optimal control can be used to adjust the weighting coefficient. The weighted square sum of the output error and the control increment is minimized to adjust the weighting coefficient, thereby indirectly realizing the constraint control of the weighted output error and control increment. The single neuron PID adaptive controller realizes adaptive and self-learning functions by adjusting the weighting coefficient.

[0097] Quadratic performance index:

[0098]

[0099] S5 The output of the rolling optimization concentration feedback controller is calculated based on the error between the component concentration detected in real time in step S1 and the optimal set trajectory in step S2; the single neuron adaptive PID controller based on the quadratic performance index outputs u(k) using the following algorithm:

[0100]

[0101] where w i (k) (i = 1, 2, 3) is the weighting coefficient,

[0102]

[0103]

[0104]

[0105] wherein,

[0106] x1(k) = e(k)

[0107] x2(k) = Δe(k) = e(k) - e(k-1)

[0108] x3 = Δ 2 e(k) = e(k) - 2e(k-1) + e(k-2)

[0109] b0 = y(1)

[0110] wherein η I , η P , and η D are the learning rates of integration, proportion, and differentiation, respectively, P and Q are the weighting coefficients of the output error and the control increment, respectively, K is the proportion coefficient of the neuron, w i (k) is the weight coefficient of the neuron, e(k) is the system error, b0 is the initial value of the output response, and w′ i(k) is a weight coefficient of neuron, x i (k) is a weight value, and Deltae(k) represents a system error increment, and y(1) represents a first value of an output of a controlled system.

[0111] The application provides a feedback control system device based on online detection of multi-component concentration, and a device schematic diagram is shown in the figure. Figure 3 As shown in the figure, the specific structure is as follows: the fermenter is provided with an in-situ sampler, a temperature sensor, a Do sensing electrode and a pH sensing electrode, the in-situ sampler is provided with a ceramic membrane, bacteria and other macromolecules can be removed, and small molecules such as glucose and water can pass through the membrane holes to realize online sampling; the temperature electrode is a Thermotrode electrode of Mettler, the dissolved oxygen electrode is an InPro6000 electrode of Mettler, and the pH electrode is an InPro6800 electrode of Mettler, which are respectively used for real-time detection of temperature, dissolved oxygen value and pH value in the fermentation process. The PLC control module is directly connected with the fermenter and the upper monitoring station, respectively, is used for collecting and adjusting temperature, pH and dissolved oxygen and other parameters in the fermentation process, and recording and exporting environmental parameter data; the online monitoring instrument is provided with electrochemical sensors of glucose and ethanol, which are used for real-time collection of glucose concentration and ethanol concentration in the fermentation broth, and are uploaded to the upper monitoring station through RS485, the upper monitoring station receives the data transmitted by RS485, processes the data by using a module in the upper computer, performs online feedback by using an optimal set trajectory and real-time detection data, and calculates a feed control amount to be added to the system at the next moment.

[0112] The feed execution mechanism is used to flow feed medium into the fermenter; the peristaltic pump is a S100-2B+TH15B model of Baoding Snu, and the precision can reach 0.0052 ml / min, the upper computer calculates the feed control amount to be added to the fermentation, can send a control signal to the peristaltic pump, and the peristaltic pump operates according to the instruction, so that automatic feed control is realized.

[0113] The enzyme biosensor electrode adopted in the application has good specificity, stability and anti-interference, and can be used for quantitative detection of component concentration.

[0114] The application can realize feedback control based on online detection of multi-component concentration in the fermentation process, and the test proves that the application is beneficial to improve the utilization rate of substrates and the yield of products in the fermentation process, the system runs stably, the man-machine interaction interface is friendly, the accuracy is high, and the system is easy to operate, can meet the basic requirements of the fermentation monitoring system, and can also be used for other fermentation processes.

[0115] Referring to Figure 4 The single neuron adaptive PID control method based on quadratic performance index has better tracking effect of step signal as a controlled object and strong robustness.

[0116] Referring to Figure 5 The control method is applied to the actual fermentation process of Saccharomyces cerevisiae producing ethanol, the fermenter is taken as a controlled object, and the tracking curve is set as the optimal production curve of the fermentation process.

Claims

1. A feedback feeding control method based on online detection of multi-component concentration, characterized in that... The method includes the following steps: S1. In-situ online sampling: The concentration of components in the fermenter is detected in real time using an online monitoring instrument to obtain the concentration of the substrate and product being tested. S2. Based on the environmental parameters in the fermenter and the substrate and product concentrations detected in S1, a soft sensor model is used to predict the cell concentration in the fermenter. A kinetic model based on substrate consumption is established using the cell concentration and the substrate and product concentrations detected in S1. The kinetic model in step S2 is described by the following differential equation: In the formula, These represent dilution rate, cell concentration, substrate concentration, product concentration, and feed concentration, respectively. Indicates the specific growth rate of bacterial cells. , and These represent the maximum specific growth rate of the bacterial cells, the product saturation coefficient, the component saturation coefficient, the component inhibition constant, and the bacterial cell yield coefficient for the component, respectively. All are dynamic parameters; S3. Combine the particle swarm optimization algorithm to optimize the parameters of the dynamic model of S2. Use the optimized dynamic model of substrate consumption as the reference trajectory and output it in real time as the optimal set trajectory. S4. Based on the error between the substrate concentration detected in real time in step S1 and the optimal set trajectory in step S3, the output of the concentration feedback controller is continuously optimized. S5. The output of the concentration feedback controller is input to a single-neuron adaptive PID feedback controller based on a quadratic performance index. The single-neuron adaptive PID feedback controller calculates the feed control amount that should be applied to the system at the current moment. This is applied to the feed pump; the feed pump is connected to the fermenter to maintain the substrate concentration in the fermenter within the required concentration range; the single-neuron adaptive PID controller based on quadratic performance indicators described in step S5 outputs... The following algorithm is used: in, These are weighting coefficients. In the formula, In the formula, These are the learning rates for integral, proportional, and differential equations, respectively. and These are the weighting coefficients for output error and control increment, respectively, and K is the proportional coefficient of the neuron. These are the weight coefficients of the neurons. For systematic error, The initial value for the output response. The weight coefficients of the neuron, For weighted values, Indicates the system error increment. This represents the first value output by the controlled system.

2. The feedback feeding control method based on online detection of multi-component concentration according to claim 1, characterized in that, The concentration detection method described in step S1 is based on the principle of electrochemical biosensors.

3. The feedback feeding control method based on online detection of multi-component concentration according to claim 1, characterized in that, The environmental parameters mentioned in step S2 include: pH, temperature T, and dissolved oxygen concentration Do; the soft sensor model is constructed based on a backpropagation (BP) neural network, which consists of three layers: an input layer, a hidden layer, and an output layer. The mathematical expression of the network is as follows: In the formula: Represents the input vector. For hidden layer output, For output parameters, and These are the connection weights between the hidden layer and the output layer. and These are the activation functions for the hidden layer and the output layer, respectively.

4. The feedback feeding control method based on online detection of multi-component concentration according to claim 1, characterized in that, The particle swarm optimization algorithm described in step S3 consists of the following steps: S3-1 treats the parameters to be optimized as particles, initializes the particle swarm size NP, the number of particle swarm algorithm iterations NG, initializes the particle positions and velocities, gives weighting coefficients and learning factors, calculates the fitness of each particle, and initializes the global optimum and individual optimum: S3-2 updates the objective function value for each particle and uses it as the fitness value for each particle. The formula for the objective function value is: In the formula, These are the kinetic model calculation values ​​for cell concentration, product concentration, and substrate concentration. These are measured values ​​from the online monitoring instrument for the fermentation process. S3-3 Updates the individual optimal solution pbest and the global optimal solution gbest according to the fitness function; S3-4 updates the particle swarm, and the equations of motion for the particle swarm are as follows: In the formula, Number of particles =1, 2, ..., N, In particle dimension, =1, 2, ..., d; The frequency of iteration replacement. The highest frequency of replacement. For inertial motion weight vector factors, and These are the maximum and minimum inertial motion weight vector factors, respectively; and Let be a random constant representing the acceleration. and A random constant between 0 and 1; For the first The iteration replaces the first The particle in the first Movement speed in the dimension; For the first The iteration replaces the first The particle in the first Position on the dimension; S3-5 checks the termination condition. If the maximum number of iterations is reached or the optimal solution no longer changes, the iteration is terminated; otherwise, return to step S3-2 to continue the optimization search.

5. A control system for implementing the feedback feeding control method based on online detection of multi-component concentration as described in claim 1, characterized in that: The control system includes a programmable logic controller (PLC), a fermenter, an in-situ sampler, an online monitoring instrument, a feed pump, and a supervisory control station, wherein: The programmable logic controller (PLC) is connected to the fermenter and the upper monitoring station to adjust the temperature (T), pH value and dissolved oxygen concentration (Do) parameters during the fermentation process, and upload the real-time detected data to the upper monitoring station. Temperature sensors, Do-sensing electrodes, and pH-sensing electrodes are installed in the fermenter to detect environmental parameters such as temperature, dissolved oxygen concentration (Do), and pH value during ethanol fermentation, respectively. The in-situ sampler is connected to the fermenter. The sampler is equipped with a ceramic membrane to remove large molecules such as bacteria, allowing small molecules, including glucose and water, to pass through the sampler. The online monitoring instrument is equipped with electrochemical sensors for glucose and ethanol, which are connected to an in-situ sampler to obtain the glucose concentration in the fermentation broth. The feed pump is connected to an online monitoring instrument and is used to add feed culture medium to the fermenter; The supervisory control station acts as the master station, connecting to the online monitor and programmable controller. It embeds control algorithms, receives, processes, and sends instructions to set fermentation parameters and display and export data.

Images