Fermentation equipment cooperative control method based on distributed control architecture

Abstract

The invention relates to a fermentation equipment cooperative control method based on a distributed control architecture. Comprising the steps that each fermentation equipment control unit broadcasts node information of the fermentation equipment control unit through a distributed protocol including Zigbee or LoRaWAN, and the node information includes equipment types, control capability and state parameters; the master control node receives and integrates the broadcast information, dynamically generates a distributed control topological graph, and marks the roles of each device including a master control, an auxiliary and a terminal; the master control node formulates a global cooperative control target according to a production target comprising a metabolite concentration or fermentation time; each node monitors fermentation parameters in real time and uploads the fermentation parameters to the master control node through a distributed data transmission protocol; the master control node generates a globally optimized control parameter set by using a collaborative optimization algorithm including a Lagrangian relaxation method; each distributed node adjusts the operation state of the equipment according to the issued parameter set; each node collects fermentation environment data including temperature mutation and pH value fluctuation in real time, and rapid analysis is carried out through an edge calculation unit.

Description

Technical Field

[0001] The invention relates to a fermentation equipment collaborative control method based on a distributed control architecture. Background Art

[0002] By comparing the fermentation control method, device and related technical solutions described in Chinese invention patent 2023107564589, it can be found that the current cooperative control method for fermentation equipment based on distributed control architecture has the following problems in design and implementation: First, the fermentation control method mainly relies on the motor running current and the fermentation liquid flow rate as monitoring parameters to determine whether the fermentation state is completed or not. Although this method can reflect the fermentation process through simple physical parameters, it ignores the dynamic changes of key biological and chemical parameters such as metabolite concentration, dissolved oxygen level, and pH value during the fermentation process. These parameters are crucial for the precise control of complex fermentation processes. Only using current and flow rate as the basis for judgment may lead to inaccurate identification of the fermentation state. For example, in some cases, even if the motor running current drops or the fermentation liquid flow rate reaches the threshold, the fermentation process may not be actually completed due to fluctuations in environmental conditions or changes in the activity of the fermenting bacteria.

[0003] Therefore, this control method based on a single physical parameter lacks the ability to comprehensively analyze multi-dimensional data and cannot meet the needs of complex fermentation processes for high-precision regulation. Secondly, this method adopts a fixed power operation mode for the operation process of the fermentation equipment, that is, the motor drives the stirring device with a fixed power during the entire fermentation process. Although this fixed power design simplifies the control logic, it cannot adapt to the actual needs of different fermentation stages. For example, in the early stage of fermentation, the viscosity of the mixture is high, and the fixed power may lead to insufficient stirring efficiency and affect the uniform distribution of the fermentation bacteria; in the later stage of fermentation, when the viscosity decreases, the fixed power may lead to unnecessary energy consumption. In contrast, dynamic power regulation can optimize the stirring parameters according to the real-time monitored characteristics of the fermentation liquid, significantly improve the energy efficiency and economy of equipment operation, but this patent fails to achieve this.

[0004] Third, the method relies too much on fixed thresholds for judging the fermentation status. For example, the second preset operating current threshold of the operating current and the first speed threshold of the fermentation liquid flow rate are both static settings, and fail to be dynamically adjusted in combination with the real-time environment or process requirements. This static threshold setting may fail when environmental conditions fluctuate, resulting in inaccurate judgment of the fermentation state. For example, when the temperature fluctuates or the characteristics of the raw materials change, the current or flow rate may deviate from the set threshold, but the actual fermentation process may still meet expectations. The introduction of dynamic thresholds, combined with real-time data analysis, can automatically adjust the judgment criteria according to changes in the fermentation environment, thereby improving the flexibility and accuracy of control, but this is not designed in this method.

[0005] Fourth, the feedback mechanism of this method has a lag. Its control logic is based on a "detection-judgment-execution" cycle, and it must complete the current feedback cycle before responding to new data. This feedback mechanism can meet basic needs when the data collection and processing delay is small, but it may not respond in time in a rapidly changing fermentation process (such as sudden temperature changes or pH fluctuations). Summary of the invention

[0006] The purpose of the present invention is to provide a fermentation equipment collaborative control method based on a distributed control architecture, thereby solving some of the drawbacks and shortcomings pointed out in the background technology.

[0007] The present invention solves the above-mentioned technical problems by adopting the following technical solution, comprising the following steps:

[0008] S1. Dynamic generation of topology of distributed control nodes:

[0009] S1.1. Each fermentation equipment control unit broadcasts its own node information, including equipment type, control capability, and status parameters, through a distributed protocol including Zigbee or LoRaWAN;

[0010] S1.2, the master control node receives and integrates broadcast information, dynamically generates a distributed control topology diagram, and marks the roles of each device including master, auxiliary, and terminal;

[0011] S2. Task decomposition based on global collaborative goals:

[0012] S2.1. The master control node formulates global collaborative control targets based on production targets including metabolite concentration or fermentation time;

[0013] S2.2. Using the task decomposition algorithm, the global goal is decomposed into multiple subtasks including temperature control, dissolved oxygen regulation, and nutrient supplementation;

[0014] S3. Multi-parameter coordinated control mechanism based on distributed control:

[0015] S3.1, each node monitors the fermentation parameters in real time and uploads them to the master control node through a distributed data transmission protocol;

[0016] S3.2, the master control node generates a globally optimized control parameter set using a collaborative optimization algorithm including a Lagrangian relaxation method;

[0017] S3.3, each distributed node adjusts the device operation status according to the parameter set issued;

[0018] S4, Distributed adaptive control based on real-time feedback:

[0019] S4.1. Each node collects fermentation environment data including temperature mutation and pH fluctuation in real time, and performs rapid analysis through the edge computing unit;

[0020] S4.2. Establish a distributed closed-loop feedback model in the master control node to calculate the error in real time and generate an adjustment strategy;

[0021] S4.3, the adjustment strategy is sent to the nodes in a distributed manner;

[0022] S5. Fault detection based on dynamic collaboration between nodes:

[0023] S5.1. Each node periodically sends status data including sensor working status and actuator deviation through a distributed health monitoring protocol;

[0024] S5.2, the master control node uses anomaly detection algorithm to identify abnormal nodes and mark their status;

[0025] S5.3, dynamically adjust the topology, redistribute tasks to healthy nodes, and isolate faulty nodes;

[0026] S6. Multi-objective collaborative optimization control based on distributed architecture:

[0027] S6.1. Define a multi-objective optimization model, taking the energy consumption, operation efficiency and product quality of the fermentation equipment as objective functions;

[0028] S6.2, each distributed node calculates the local target value based on its own parameters, and coordinates the calculation of the global target through the master control node;

[0029] S6.3, the optimization results are synchronized to each node through a distributed protocol to guide the real-time control of each device.

[0030] Furthermore, the task decomposition method based on the global collaborative goal:

[0031] By constructing a metabolic kinetic model, the concentration dynamics of the target metabolite is directly associated with the nonlinear relationship of various process parameters to predict and optimize the fermentation process; the metabolic kinetic model is defined as follows:

[0032]

[0033] Where M(t) represents the change of the concentration of the target metabolite over time t; C x(t) is the active cell concentration function, which reflects the law of change of the number of active cells over time during the fermentation process; Φ(T,O2,N) is the interaction function of process parameters, which specifically describes the dynamic relationship between temperature T, dissolved oxygen concentration O2 and nutrient supply rate N in the fermentation environment, including the enhancement or inhibition of metabolic rate by changes in dissolved oxygen concentration; Γ(G) is the global target correction factor, which is used to adjust the metabolic kinetic model according to the set fermentation target G.

[0034] Furthermore, the task decomposition method based on the global collaborative goal:

[0035] The task decomposition algorithm is used to decompose complex targets into multiple specific subtasks, including temperature control, dissolved oxygen regulation, and nutrient supplementation. The task decomposition process integrates the independence of subtasks, their mutual coupling relationships, and process constraints, and the global task decomposition model is defined as:

[0036]

[0037] Among them, J represents the optimization goal of global task decomposition, which comprehensively measures the control accuracy and resource utilization of the fermentation system after task decomposition; n is the total number of distributed control nodes; α i is the priority weight of the i-th node, reflecting the importance of the node task in the global goal, They respectively represent the influence gradients of the temperature, dissolved oxygen concentration and nutrient supply rate of the i-th node on the target product concentration P; Ψ(L) is the load balancing function, L represents the workload distribution of each node in the system; β is the balance factor between global optimization and load balancing.

[0038] Furthermore, the task decomposition method based on the global collaborative goal:

[0039] After task decomposition, subtasks are assigned to distributed control nodes through a dynamic allocation algorithm. At the same time, the operating status of the nodes and the task execution efficiency are evaluated in real time. The optimization execution strategy includes adaptive adjustment according to the node performance. The specific optimization model is defined as:

[0040]

[0041] Among them, U i is the execution adjustment value of the i-th node, reflecting the adjustment range of the running state of the node after optimization; Δ(x) is the adaptive adjustment function, dynamically controlling the size of the adjustment range; Θ i,opt and θ i,act are the expected performance and actual performance of the i-th node, respectively, which are used to measure the deviation of the current node task execution; Ω(ρ i ,ξ i ) is a dynamic correction function based on the node load state ρ iand health factors i Ensure the reliability of task assignment.

[0042] Furthermore, the distributed adaptive control method based on real-time feedback includes:

[0043] Key data of the fermentation environment including temperature and pH value are collected in real time and analyzed quickly. The edge computing unit deploys a lightweight algorithm to extract mutation features in environmental data through data denoising and anomaly detection. The core data feature extraction formula is defined as follows:

[0044]

[0045] Among them, S i (t) represents the data features of node i at time t; X i (t) is the real-time data collected by node i, including temperature, dissolved oxygen or pH value; ω is the dynamic weight factor, which adjusts the weight according to the importance of historical data and real-time data; f i (t) is the trend function, which captures the dynamic trend of data changes; Δt is the width of the analysis window, which is used to set the time span.

[0046] Furthermore, the distributed adaptive control method based on real-time feedback includes:

[0047] A distributed closed-loop feedback control model is established on the master control node. The global error is calculated in real time by integrating the data uploaded by the distributed nodes to generate a global optimization adjustment strategy. The error calculation formula is as follows:

[0048]

[0049] Among them, E g (t) is the global error of the system, which is used to evaluate the deviation between the current state and the target state; n is the total number of distributed nodes; Y i (t) is the real-time feedback value of node i, indicating the value of the process parameter currently monitored; R i (t) is the target reference value of node i, which indicates the process setting value that the node should achieve; γ i is the node weight factor; φ(Δ i ) is the error change rate correction function, is the error change rate of node i.

[0050] Furthermore, the distributed adaptive control method based on real-time feedback includes:

[0051] The error is divided into global error and local error. The global error is generated by the master node through the global optimization algorithm to generate adjustment parameters, while the local error is quickly processed by the distributed node through the local adjustment algorithm. The optimization strategy generation formula is as follows:

[0052]

[0053] Among them, U′ i is the optimization execution value of node i, indicating the specific operation amount of the node in the local adjustment; κ is the trade-off factor between global and local adjustments, which is used to balance the priority of global strategy and local optimization; Ψ is the global objective function, which is used to evaluate the effect of the optimization strategy in the global scope; Θ i is the current state parameter of node i; Ω(ρ i ,ξ i ) is the dynamic correction function, ρ i represents the node load status, ξ i represents the health status of the node; λ is the local optimization weight; F(Θ i ,L i ) is the local optimization function, Θ i is the node state, L i is the local error distribution.

[0054] The fermentation equipment collaborative control method based on the distributed control architecture of the present invention has the following beneficial effects:

[0055] Through real-time data collection and analysis and the coordinated optimization of distributed nodes, the present invention can dynamically adjust fermentation parameters such as temperature, dissolved oxygen concentration and nutrient supply rate to ensure that the metabolic environment is always in the best state, thereby significantly improving the yield and quality of metabolites. Using distributed architecture and edge computing technology, the present invention realizes localized data processing and rapid response at each node, reduces the delay problem of data transmission and centralized computing, and enables the system to respond to environmental fluctuations and sudden abnormalities in real time.

[0056] By introducing global optimization algorithms and load balancing mechanisms, the system can reasonably allocate tasks to avoid single-node overload or resource waste. At the same time, precise control of various parameters reduces energy consumption caused by over-operating equipment. Through node health status monitoring and dynamic topology adjustment, faulty nodes can be quickly identified and tasks can be reallocated to isolate problem areas, ensuring the overall stable operation of the system and reducing the risk of downtime caused by single-point failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 The present invention is a flow chart of the cooperative control method of fermentation equipment based on the distributed control architecture.

[0058] Figure 2 This is a flow chart of the task decomposition method based on the global collaborative goal of the present invention.

[0059] Figure 3 This is a flow chart of the distributed adaptive control method based on real-time feedback of the present invention. DETAILED DESCRIPTION

[0060] The specific implementation modes of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0061] Combination process Figure 1 As shown in Figure 1, one of the key technologies of the cooperative control method of fermentation equipment based on distributed control architecture is the dynamic topology generation of distributed control nodes. Specifically, the control unit of each fermentation equipment regularly broadcasts its own node information through a protocol that supports distributed communication (such as Zigbee or LoRaWAN). The node information includes the basic attributes of the device, such as the device type (such as temperature control unit, dissolved oxygen control unit, stirring control unit, etc.), control capabilities (such as response speed, control accuracy, etc.), and real-time status parameters (such as current operating status, load conditions, etc.). The broadcast information can be received by the master control node in the distributed control system, and the master control node automatically integrates the status and functional attributes of all devices based on the received information. Through this process, the master control node dynamically generates a distributed control topology diagram, in which each node is assigned a specific role according to its functional attributes and operating status, including the master control role (responsible for global scheduling and strategy generation), the auxiliary role (assisting the master control node to complete local data processing or control tasks), and the terminal role (mainly executing control commands and collecting data).

[0062] The master control node dynamically formulates the global collaborative control target by collecting key parameters in the fermentation process (such as metabolite concentration, fermentation time, etc.) in real time and combining the production goals of the system (such as maximizing the target product concentration or shortening the fermentation time). This global target is usually expressed in the form of a comprehensive performance indicator, such as optimizing the resource utilization efficiency of the fermentation equipment and improving the output quality of the metabolites. After clarifying the global target, the master control node uses the task decomposition algorithm to disassemble the global target and divide it into multiple subtasks with clear functions, including but not limited to core links such as temperature control, dissolved oxygen regulation, and nutritional supplementation. During the task decomposition process, the master control node will map the dynamic constraints of the global target (such as temperature range, dissolved oxygen concentration range, and nutrient supply rate) to specific control tasks based on the complexity, interdependence, and equipment operation capabilities of each subtask, and ensure the coordination and independence between the tasks. For example, the temperature control task will be assigned to a node with precise temperature control capabilities, and the dissolved oxygen regulation task will be assigned to a node related to oxygen supply. In this way, the master control node not only achieves the precise decomposition of production targets, but also ensures that tasks can be efficiently executed in a distributed architecture.

[0063] One of the core steps to achieve intelligent and efficient control of fermentation equipment is to first require distributed control nodes to monitor key parameters in the fermentation process in real time, including temperature, dissolved oxygen concentration, pH value, nutrient supply rate, etc. The real-time data obtained by each node through the sensor is uploaded to the master control node through a distributed data transmission protocol (such as MQTT or LoRaWAN). After receiving the data from all nodes, the master control node conducts a comprehensive evaluation of the state of the global fermentation system and uses a collaborative optimization algorithm (such as Lagrangian relaxation method) to generate a global optimization control parameter set. The Lagrangian relaxation method relaxes the complex constraints of the global objective function into local sub-problems by introducing Lagrangian multipliers, so that each distributed node can calculate in parallel and optimize collaboratively, thereby significantly improving the computational efficiency. In the collaborative optimization process, the master control node incorporates the dynamic coupling relationship between multiple parameters into the optimization model according to the global goal (such as maximizing the output of metabolites or minimizing energy consumption), and generates an optimization parameter set including temperature set point, dissolved oxygen concentration target value, nutrient supply rate, etc. These optimized parameter sets are sent to each distributed control node, and the node adjusts the operating state of its own equipment according to the sent parameters. For example, a temperature control node adjusts heating or cooling equipment to maintain an optimal temperature range, and a dissolved oxygen regulation node controls gas flow or stirring speed to maintain a target dissolved oxygen concentration.

[0064] Real-time feedback distributed adaptive control is a key technology for achieving precise control in a dynamic fermentation environment. First, each distributed control node is equipped with an edge computing unit to collect key data of the fermentation environment in real time, such as temperature mutations, pH fluctuations and other dynamic change data. Each node uses a locally deployed lightweight algorithm (such as sliding window average or fast anomaly detection algorithm) to quickly analyze the collected raw data, including data denoising, feature extraction and trend judgment. Through the edge computing unit, the node can locally identify abnormal fluctuations that may affect the fermentation process, and upload the processed key data to the master control node first, thereby reducing the communication load and improving the response speed. On the master control node, a distributed closed-loop feedback model is constructed, which calculates the global error of the system in real time by integrating the key data uploaded from all distributed nodes. The error calculation is based on the deviation between the current state of each node and the target state, and a dynamic correction mechanism for the error change rate is introduced to capture the impact of sudden and nonlinear changes on the stability of the system. The master control node uses a closed-loop feedback model to generate an optimization adjustment strategy, which includes optimization and adjustment of global parameters (such as temperature set points or dissolved oxygen concentration target values), as well as rapid response measures to local anomalies (such as reducing the nutrient supply rate of specific nodes). The generated adjustment strategy is sent to each control node through a distributed protocol, and the node adjusts its own operating status in real time according to the received strategy. For example, the temperature control node modifies the power of the heating equipment according to the strategy, and the stirring node dynamically changes the stirring speed to stabilize the dissolved oxygen concentration. Through the closed-loop control mechanism of real-time acquisition, rapid analysis, global optimization and distributed feedback, each node works together to effectively respond to complex and changeable environmental factors in the fermentation process, ensuring stable operation and efficient production of the system under dynamic conditions.

[0065] Fault detection based on dynamic coordination between nodes is a key step to ensure system stability and fault tolerance. Each distributed control node regularly sends its own operating status data to the master node through a distributed health monitoring protocol. These data include the working status of the sensor (such as whether data loss occurs, excessive signal noise or calibration deviation) and the operating deviation of the actuator (such as the difference between the execution instruction and the actual behavior). Through this monitoring mechanism, the system can continuously obtain the health status information of each node. After receiving the status data of each node, the master node uses anomaly detection algorithms to analyze and process the data. Anomaly detection algorithms can include analysis based on statistical features (such as standard deviation of data fluctuation exceeding the limit) or machine learning methods (such as anomaly classification models trained based on historical operation data) to accurately identify abnormal nodes. Nodes identified as abnormal will be marked by the master node. The marking information includes the node's abnormality type (such as sensor failure or actuator error) and the severity of the abnormality (such as complete failure or slight deviation). After the marking is completed, the master node will reallocate tasks according to the dynamic adjustment mechanism of the global topology. Specifically, the master control node will reallocate the tasks that the abnormal node is responsible for to other healthy nodes, giving priority to those nodes with low load and good health status, to ensure that the tasks can be continuously and efficiently executed. At the same time, the faulty node will be isolated by the system and will no longer participate in task allocation. The maintenance personnel will be notified through the distributed protocol to troubleshoot and repair the fault. The entire process is completed dynamically without stopping the system operation, thus ensuring the continuity and stability of the fermentation process.

[0066] Multi-objective collaborative optimization control is a key step in achieving a balance between fermentation process performance and resource utilization efficiency.

[0067] First, the system defines a multi-objective optimization model, which sets the energy consumption, operating efficiency and product quality of the fermentation equipment as the optimization objective function. There may be conflicts between these objectives. For example, it is necessary to maintain high product quality while reducing energy consumption. Therefore, the multi-objective optimization model needs to comprehensively consider these competing objectives to find the global optimal solution. Each distributed control node calculates its own local target value based on its real-time operating parameters (such as local energy consumption, equipment efficiency, product output rate, etc.). The local target value is part of the optimization model and is used to describe the contribution of the node to the global goal. Through the distributed protocol, each node uploads the calculated local target value to the master control node, which is responsible for coordinating the calculation of the global goal. The master control node uses a multi-objective optimization algorithm (such as weighted summation method, Pareto optimization algorithm or genetic algorithm) to generate a global optimal control parameter set based on the weights of each objective, node status and the dependency between tasks. This parameter set includes specific optimization strategies for each distributed node, such as energy consumption allocation of temperature control nodes, operating efficiency improvement goals of stirring nodes, and product optimization strategies of nutrient supply nodes. The optimization results are synchronized to all relevant nodes through a distributed protocol. Each node dynamically adjusts its own operating status according to the optimization parameters sent, such as modifying the equipment operation mode, adjusting energy consumption priority or reallocating resources, so as to achieve the optimal balance between energy consumption, efficiency and product quality in the whole system. In this way, the entire system can achieve real-time regulation in a dynamically changing fermentation environment, ensuring the continuity and flexibility of multi-objective optimization.

[0068] Embodiment 1:

[0069] Combination process Figure 2 As shown in the figure, in a fermentation plant for lactic acid bacteria production, the goal is to achieve efficient production of lactic acid by optimizing the control of fermentation equipment. Lactic acid is a metabolite, and its production efficiency is affected by the dynamic effects of temperature (T), dissolved oxygen concentration (O2) and nutrient supply rate (N). To achieve this goal, the plant introduced a collaborative control method for fermentation equipment based on a distributed control architecture, and used a metabolic kinetic model to predict and optimize the change in the concentration of the target metabolite lactic acid over time.

[0070] Define the metabolic kinetic model:

[0071]

[0072] M(t): lactate concentration (target variable), in g / L.

[0073] C x (t): Cell concentration of active lactic acid bacteria, as a function of C x (t) = C x0 e μt , where C x0=0.1gg / L (initial cell concentration), μ = 0.5h -1 (Specific growth rate).

[0074] Φ(T,O2,N): interaction function of process parameters, defined as

[0075] k T =0.02 (temperature influence coefficient), the range of T is 30℃≤T≤40℃.

[0076] (Dissolved oxygen influence coefficient), the range of O2 is 10%≤O2≤20%.

[0077] k N =0.1 (nutritional influence coefficient), the range of N is 1.0g / L≤N≤5.0g / L.

[0078] Γ(G): global target correction factor, defined as Γ(G) = G·(1e -t / τ ), where G = 20 g / L (target lactate concentration), τ = 10 h (time constant).

[0079] Maximize M(t) and optimize the set points of T, O2, and N while minimizing energy consumption and ensuring stable equipment operation.

[0080] 1. Assuming the initial conditions T = 35 ° C, O2 = 15%, N = 3.0 g / L, calculate the lactic acid production concentration M(10) after 10 hours:

[0081] C x (10) = 0.1e 0.5·10 =0.1e 5 ≈14.8g / L.

[0082] Φ(T,O2,N)=0.02·35·0.03·15·0.1·3.0=0.945.

[0083] Correction factor Γ(G) = 20·(1e -10 / 10 )=20·(11 / e)≈12.6g / L.

[0084]

[0085] Task decomposition and optimization

[0086] 1. Task decomposition of the master node:

[0087] Temperature control subtask: Assigned to the temperature control unit to ensure that T fluctuates within 1°C at 35°C.

[0088] Dissolved oxygen regulation subtask: Assigned to the oxygen supply unit to keep O2 at around 15%.

[0089] Nutrition supply subtask: Assigned to the nutrient pump control unit to dynamically adjust N according to the actual metabolic rate.

[0090] 2. Adjust optimization parameters:

[0091] The master control node detected that the dissolved oxygen concentration fluctuated to 12%. The system recommended adjusting O2=17% through the optimization algorithm and simultaneously optimized the temperature set point to 36°C.

[0092] 3. Predict the production concentration 10 hours after optimization:

[0093] Φ(T,O2,N)=0.02·36·0.03·17·0.1·3.0=1.1028.

[0094]

[0095] Through task decomposition and real-time optimization, the system increased the production concentration of lactic acid from 152.6g / L to 163.2g / L, and effectively ensured the stability of the system while dynamically adjusting parameters. This verifies the feasibility and efficiency of the task decomposition method based on global collaborative objectives in fermentation process control.

[0096] Setting task decomposition requires further optimization of the task allocation of multiple distributed nodes in the system to improve the production efficiency of lactic acid while ensuring load balance and reducing energy consumption. To this end, the task decomposition algorithm is used to decompose the global goal into specific subtasks, and global optimization is performed in combination with the mutual coupling relationship between temperature control, dissolved oxygen regulation and nutrient supplementation, as well as the process constraints of the system. It is assumed that there are n = 4 distributed nodes in the system, which are responsible for temperature control (node ​​1 and node 2), dissolved oxygen regulation (node ​​3) and nutrient supplementation (node ​​4). The goal is to optimize the final concentration P of lactic acid. The initial conditions are: the temperature set points of nodes 1 and 2 are T1 = 35 ° C and T2 = 36 ° C, respectively, and the dissolved oxygen concentration of node 3 is set to O 2,3 =15%, the nutrient supply rate of node 4 is N4=3.0g / L. The process constraints are 30℃≤T≤40℃, 10%≤O2≤20%, 1.0g / L≤N≤5.0g / L.

[0097] The task decomposition model is defined as:

[0098]

[0099] J: global task decomposition optimization target. n = 4: number of distributed control nodes. α i : Priority weight, value range 0.1≤α i≤1.0, reflecting the importance of the node task. are the influence gradients of node i temperature, dissolved oxygen, and nutrition on the target concentration P, and the range of values ​​is Ψ(L): load balancing function, defined as Where L i represents the current load of node i, L max Indicates the maximum load in the system. β: Balance factor, ranging from 0.5≤β≤2.0, used to balance control accuracy and load balance.

[0100] Calculation example:

[0101] Set the current priority weights to α1=0.8, α2=0.7, α3=0.6, α4=0.5, and the initial gradient value is The current load is L1 = 30%, L2 = 40%, L3 = 50%, L4 = 20%, and L max =50%. β =1.0.

[0102] Calculate the gradient contribution:

[0103]

[0104] Calculate the load balancing function:

[0105]

[0106] Compute the global optimization objective:

[0107] J = 1.37 + 1.0 2.16 = 3.53

[0108] According to the calculation results, the system found that the dissolved oxygen regulation of node 3 contributed the most to the target concentration, and the load was high. It was recommended to adjust the priority weight α3 of point 3 to 0.4 and increase the nutrient supply rate of node 4 to share part of the production pressure. After the adjustment, the resource utilization efficiency and production performance of the system were significantly optimized. Through the task decomposition model, the system can dynamically evaluate the task importance and load of each distributed node, reasonably allocate resources and dynamically adjust the task priority.

[0109] After the task is decomposed, the subtasks are assigned to the distributed control nodes through a dynamic allocation algorithm. At the same time, the operating status of the nodes and the task execution efficiency are evaluated in real time, and optimization and adjustment are made according to the performance and actual status of the nodes. The goal of this example is to improve the operating efficiency and reliability of the entire system by optimizing the execution strategy after the subtasks are assigned.

[0110] The fermentation system still includes n = 4 distributed control nodes, nodes 1 and 2 are responsible for temperature control, node 3 is responsible for dissolved oxygen regulation, and node 4 is responsible for nutrient supply. According to the task decomposition results, the expected temperature performance of nodes 1 and 2 is set as Θ 1,opt =35℃ and Θ 2,opt =36℃, the expected value of dissolved oxygen performance of node 3 Θ 3,opt =15%, the expected value of the nutrient supply rate of node 4 is Θ 4,opt =3.5g / L. Real-time monitoring found that the actual performance of the node is Θ 1,act =34℃、Θ 2,act =37℃、Θ 3,act =13%, θ 4,act =3.0g / L. The initial load state is ρ1=30%, ρ2=40%, ρ3=50%, ρ4=25%, and the health factor ξ i Value range: 0≤ξ i ≤1, indicating the healthy state of the node, and the current values ​​are ξ1=0.9, ξ2=0.7, ξ3=0.6, ξ4=0.8.

[0111] The optimized execution strategy is based on the following model:

[0112]

[0113] Δ(x): Adaptive adjustment function, defined as Δ(x) = 1e -x , reflecting the dynamic response of the adjustment amplitude, and its value range is 0≤Δ(x)≤1.

[0114] Ω(ρ i ,ξ i ): Dynamic correction function, defined as Ω(ρ i ,ξ i )=(1ρ i )·ξ i , used to correct the execution reliability of the node.

[0115] The system goal is to maximize the U of each node i value to optimize the operating status.

[0116] 1. Calculate the execution adjustment value of node 1:

[0117]

[0118] Ω(ρ1,ξ1)=(10.3)·0.9=0.7·0.9=0.63

[0119] U1=0.636·0.63≈0.401

[0120] 2. Calculate the execution adjustment value of node 2:

[0121]

[0122] Ω(ρ2,ξ2)=(10.4)·0.7=0.6·0.7=0.42

[0123] U2=0.625·0.42≈0.263

[0124] 3. Calculate the execution adjustment value of node 3:

[0125]

[0126] Ω(ρ3,ξ3)=(10.5)·0.6=0.5·0.6=0.3

[0127] U3=0.681·0.3≈0.204

[0128] 4. Calculate the execution adjustment value of node 4:

[0129]

[0130] Ω(ρ4,ξ4)=(10.25)·0.8=0.75·0.8=0.6

[0131] U4=0.692·0.6≈0.415

[0132] Optimization results analysis:

[0133] U4=0.415 of node 4 and U1=0.401 of node 1 are relatively high, indicating that node 4 and node 1 have higher operational reliability after adjustment.

[0134] Node 2’s U2=0.263 and Node 3’s U3=0.204 are relatively low, indicating that their current operating status and health factors are poor and require additional optimization. Based on the calculation results, the master control node preferentially assigns tasks to nodes with higher U1 and U4, while reducing the task burden of nodes 2 and 3. For example, some temperature control tasks are transferred from node 2 to node 1, and some dissolved oxygen regulation tasks are transferred from node 3 to node 4. Through dynamically adjusted task allocation, the system can further optimize resource utilization efficiency while reducing the risk of node failure due to overload or decreased health factors. Through dynamic allocation algorithms and optimized execution strategies, tasks are reasonably allocated, and the operating efficiency and reliability of nodes are significantly improved. This example demonstrates that the optimization model U based on dynamic allocation can be used to optimize resource utilization and reduce the risk of node failure due to overload or decreased health factors. i The effectiveness of the coordinated control of actual fermentation equipment was verified, and the feasibility and applicability of the method were verified through real-time adjustment.

[0135] Embodiment 2:

[0136] Combination process Figure 3 As shown in the figure, in a lactic acid bacteria fermentation plant, in order to maintain a stable metabolic environment, the fermentation process needs to strictly control the temperature and pH value, the goal is to ensure that the temperature is stable at around 35°C and the pH value is maintained at 6.5. The system is set to consist of 4 distributed nodes, of which nodes 1 and 2 are responsible for temperature acquisition and regulation, and nodes 3 and 4 are responsible for pH monitoring and regulation.

[0137] The temperature and pH data collected in real time are used to extract key features using the following formula:

[0138]

[0139] in:

[0140] S i (t): data features of node i at time t; X i (t): real-time data collected by node i (such as temperature or pH value); ω: dynamic weight factor, the value range is 0≤ω≤1, the weight is adjusted according to the importance of historical and real-time data, and the current value is set to ω=0.7; f i (t): trend function, defined as f i (t) = 1 + k sin (πt / Δt), captures the periodic change trend of the data, the coefficient k ranges from 0.1 ≤ k ≤ 1.0, and k is set to 0.5; Δt: analysis window width, the range is 5min ≤ Δt ≤ 60min, and Δt is set to 15min.

[0141] 1. Node 1 (temperature) data feature calculation:

[0142] Real-time data collection X1(t): The data in the past 15 minutes is [34.8, 35.2, 34.9, 35.1, 34.7]℃.

[0143] Dynamic weight ω=0.7.

[0144] Trend function f1(t)=1+0.5·sin(πt / 15).

[0145] calculate:

[0146]

[0147] Discrete integral:

[0148] S1(t)≈0.7·(34.8·1.25+35.2·1.15+34.9·1.10+35.1·1.05+34.7·1.00)≈123.76

[0149] 2. Node 3 (pH value) data feature calculation:

[0150] Real-time collected data X3(t): The data in the past 15 minutes is [6.6, 6.5, 6.4, 6.5, 6.7].

[0151] Dynamic weight ω=0.7.

[0152] Trend function f3(t)=1+0.5·sin(πt / 15).

[0153] calculate:

[0154] S3(t)≈0.7·(6.6·1.25+6.5·1.15+6.4·1.10+6.5·1.05+6.7·1.00)≈28.91

[0155] After receiving the characteristic values ​​of all distributed nodes, the master control node detected that the temperature characteristic S1(t)=123.76 of node 1 and the pH characteristic S3(s)=28.91 of node 3 deviated from the stable range. Through analysis, it was found that the periodic fluctuation of temperature may be caused by the control lag of the heating equipment; the fluctuation of pH value may be caused by the instability of nutrient supply rate. The master control node optimizes the adjustment strategy through the closed-loop feedback mechanism:

[0156] 1. Increase the control frequency of the heating equipment at node 1 and reduce the temperature control lag.

[0157] 2. Optimize the nutrient supply rate of node 4 and smooth the fluctuation of pH value.

[0158] After adjustment, the new eigenvalue of node 1 dropped to S1(t)=120.34, and the new eigenvalue of node 3 dropped to S3(t)=27.65, both returning to the stable range, indicating that the adjustment strategy is effective.

[0159] In actual operation, in order to further optimize the fermentation process, the master control node introduced a distributed closed-loop feedback control model, which calculated the global error in real time and generated an optimization adjustment strategy to ensure that the fermentation environment operates efficiently within the target range.

[0160] The fermentation system consists of n=4 distributed nodes. Node 1 and Node 2 are responsible for temperature control, with target temperatures R1=35°C and R2=36°C; Node 3 and Node 4 are responsible for monitoring and adjusting pH and dissolved oxygen concentration, respectively, with targets set at R3=6.5 and R4=15%. The real-time feedback values ​​are Y1=34.5°C, Y2=37°C, Y3=6.3 and Y4=13%. The weight factor γ i Indicates the importance of the node in the global error, with a value range of 0.1≤γ i ≤1.0, set to γ1=0.3,γ2=0.3,γ3=0.2,γ4=0.2.

[0161] The error change rate correction function is φ(Δ i )=1+k·|Δ i |, where k = 0.5 indicates the importance of the rate of change to the error, is the error change rate, and its value is calculated based on the change in the previous cycle.

[0162] The formula for calculating the global error of the system by the master control node is:

[0163]

[0164] First, calculate the real-time error of each node:

[0165] The error of node 1 is Y1-R1=34.5-35=-0.5; the rate of change in the previous cycle

[0166]

[0167] Correction function φ(Δ1)=1+0.5·|-0.1|=1.05.

[0168] The weighted error contribution γ1·(Y1-R1)·φ(Δ1)=0.3·-0.5·1.05=-0.1575.

[0169] The error of node 2 is Y2-R2=37-36=1.0; the rate of change in the previous cycle

[0170] Correction function φ(Δ2)=1+0.5·|0.2|=1.1.

[0171] The weighted error contribution γ2·(Y2R2)·φ(Δ2)=0.3·1.0·1.1=0.33.

[0172] The error of node 3 is Y3R3=6.36.5=-0.2; the rate of change in the previous cycle

[0173] Correction function φ(Δ3)=1+0.5·|-0.1|=1.05.

[0174] The weighted error contribution γ3·(Y3R3)·φ(Δ3)=0.2·-0.2·1.05=-0.042.

[0175] Error of node 4 Y4R4 = 1315 = -2.0; rate of change in the previous cycle

[0176] Correction function Φ(Δ4)=1+0.5·|-0.2|=1.1.

[0177] The weighted error contribution γ4·(Y4R4)·φ(Δ4)=0.2·-2.0·1.1=-0.44.

[0178] Global error E g (t) Calculation:

[0179] E g (t)=-0.1575+0.330.0420.44=-0.3095

[0180] Master node identification E g (t) = -0.3095 means that the system is currently deviating from the target in a negative trend, and we need to focus on the nodes with large negative errors. According to the calculation results:

[0181] Node 4 has the largest error, and it is recommended to increase the dissolved oxygen supply rate and adjust the target dissolved oxygen set point to R4 = 14.5%.

[0182] The temperature control lag problem of node 1 is relatively minor, but because the error change rate is close to zero, the current setting can be temporarily maintained.

[0183] The temperature control error of node 2 is positive, the control range is relatively stable, and the heating frequency is appropriately reduced to save energy.

[0184] After adjustment, the real-time feedback values ​​obtained by re-monitoring are Y1 = 34.8°C, Y2 = 36.5°C, Y3 = 6.5, and Y4 = 14.5%. Then calculate the global error:

[0185]

[0186] After optimization, the global error is significantly reduced, verifying the effectiveness of the adjustment strategy.

[0187] Through the distributed closed-loop feedback control model, the global error is calculated in real time and the adjustment strategy is generated, which can quickly locate the problem node and optimize the control parameters, significantly improving the response speed and stability of the system. This example shows that the application of this method in a dynamic fermentation environment is feasible and efficient.

[0188] In order to further improve the control accuracy and system stability, a distributed adaptive control method based on real-time feedback is adopted to divide the error into global error and local error. The global error is generated by the master node through the global optimization algorithm to adjust the parameters, while the local error is quickly processed by the distributed node through the local adjustment algorithm.

[0189] 1. System settings and initial conditions:

[0190] The fermentation system consists of n = 4 distributed nodes, where nodes 1 and 2 are responsible for temperature control, node 3 is responsible for dissolved oxygen regulation, and node 4 is responsible for nutrient supply regulation. The goal is to maximize lactic acid production (global objective function Ψ) while ensuring that temperature, dissolved oxygen, and nutrient parameters are within the target range. The initial conditions are as follows:

[0191] Temperature target values: The set values ​​for node 1 and node 2 are 35°C and 36°C respectively.

[0192] Dissolved oxygen concentration target value: The set value of node 3 is 15%.

[0193] Nutrient supply rate target value: The set value of node 4 is 3.5g / L.

[0194] Real-time feedback values: Y1=34.7℃, Y2=37℃, Y3=13%, Y4=3.0g / L.

[0195] 2. Error partitioning and global optimization formula:

[0196] The global error Ψ represents the optimization target of lactic acid production, and its influencing factors include temperature (Θ_1, Θ_2), dissolved oxygen (Θ3) and nutrition (Θ4). The formula for the master node to generate the adjustment strategy is:

[0197]

[0198] κ: The trade-off factor between global and local adjustments, with a value range of 0.5≤κ≤2.0, and κ=1.0.

[0199] The gradient of the global objective function's influence on the node state, with a range of values The specific values ​​are as follows:

[0200] Ω(ρ i ,ξ i ): Dynamic correction function, defined as Ω(ρ i ,ξ i )=(1ρ i )·ξ i Node load status ρ i and health factors i They are as follows: ρ1=0.3, ρ2=0.4, ρ3=0.5, ρ4=0.2, ξ1=0.9, ξ2=0.8, ξ3=0.7, ξ4=0.85.

[0201] λ: local optimization weight, value range 0.5≤λ≤1.5, set λ=1.0.

[0202] F(θ i ,L i): Local optimization function, set to F(Θ i ,L i )=|Θ i L i |, where L i is the local error of the node, and the initial values ​​are L1=0.3, L2=0.4, L3=0.7, L4=0.5.

[0203] 3. Calculate the optimized execution value:

[0204] Node 1:

[0205] Ω(ρ1,ξ1)=(10.3)·0.9=0.63

[0206]

[0207] ∫ t t+Δt F(Θ1,L1)dt=|Θ1L1|=|34.70.3|=34.4

[0208] U′1=1.0·0.252+1.0·34.4=34.652

[0209] Node 2:

[0210] Ω(ρ2,ξ2)=(10.4)·0.8=0.48

[0211]

[0212] ∫ t t+Δt F(Θ2,L2)dt=|Θ2L2|=|370.4|=36.6

[0213] U′2=1.0·0.144+1.0·36.6=36.744

[0214] Node 3:

[0215] Ω(ρ3,ξ3)=(10.5)·0.7=0.35

[0216]

[0217] ∫ t t+Δt F(Θ3,L3)dt=|Θ3L3|=|130.7|=12.3

[0218] U′3=1.0·0.175+1.0·12.3=12.475

[0219] Node 4:

[0220] Ω(ρ4,ξ4)=(10.2)·0.85=0.68

[0221]

[0222] ∫ t t+Δt F(Θ4,L4)dt=|Θ4L4|=|3.00.5|=2.5

[0223] U′4=1.0·0.408+1.0·2.5=2.908

[0224] 4. Optimization results analysis and adjustment:

[0225] Optimized execution value U' i This shows that node 2 has the largest adjustment demand, followed by node 1 and node 3, and node 4 has the smallest demand. The master control node will prioritize reducing the temperature control load of node 2 and lower the temperature target to 35.5°C through dynamic adjustment. At the same time, the dissolved oxygen supply of node 3 will be optimized first to make up for the current lack of dissolved oxygen.

[0226] By combining error partitioning with global-local optimization strategies, the system achieves precise control of complex fermentation environments. The optimization results verify the efficiency and feasibility of the method and improve the stability and output of the fermentation equipment.

[0227] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A fermentation equipment collaborative control method based on a distributed control architecture, characterized in that The following steps are involved: S1. Dynamic generation of topology of distributed control nodes: S1.

1. Each fermentation equipment control unit broadcasts its own node information, including equipment type, control capability, and status parameters, through a distributed protocol including Zigbee or LoRaWAN; S1.2, the master control node receives and integrates broadcast information, dynamically generates a distributed control topology diagram, and marks the roles of each device including master, auxiliary, and terminal; S2. Task decomposition based on global collaborative goals: S2.

1. The master control node formulates global collaborative control targets based on production targets including metabolite concentration or fermentation time; S2.

2. Using the task decomposition algorithm, the global goal is decomposed into multiple subtasks including temperature control, dissolved oxygen regulation, and nutrient supplementation; S3. Multi-parameter coordinated control mechanism based on distributed control: S3.1, each node monitors the fermentation parameters in real time and uploads them to the master control node through a distributed data transmission protocol; S3.2, the master control node generates a globally optimized control parameter set using a collaborative optimization algorithm including a Lagrangian relaxation method; S3.3, each distributed node adjusts the device operation status according to the parameter set issued; S4, Distributed adaptive control based on real-time feedback: S4.

1. Each node collects fermentation environment data including temperature mutation and pH fluctuation in real time and analyzes it through the edge computing unit; S4.

2. Establish a distributed closed-loop feedback model in the master control node to calculate the error in real time and generate an adjustment strategy; S4.3, the adjustment strategy is sent to the nodes in a distributed manner; S5. Fault detection based on dynamic collaboration between nodes: S5.

1. Each node periodically sends status data including sensor working status and actuator deviation through a distributed health monitoring protocol; S5.2, the master control node uses anomaly detection algorithm to identify abnormal nodes and mark their status; S5.3, dynamically adjust the topology, redistribute tasks to healthy nodes, and isolate faulty nodes; S6. Multi-objective collaborative optimization control based on distributed architecture: S6.

1. Define a multi-objective optimization model, taking the energy consumption, operation efficiency and product quality of the fermentation equipment as objective functions; S6.2, each distributed node calculates the local target value based on its own parameters, and coordinates the calculation of the global target through the master control node; S6.3, the optimization results are synchronized to each node through a distributed protocol to guide the real-time control of each device.

2. The fermentation equipment collaborative control method based on distributed control architecture according to claim 1 is characterized in that The task decomposition method based on global collaborative goals: By constructing a metabolic kinetic model, the concentration dynamics of the target metabolite is directly associated with the nonlinear relationship of various process parameters to predict and optimize the fermentation process; the metabolic kinetic model is defined as follows: M(t)=∫0 t [C x (t)·Φ(T,O2,N)]dt+Γ(G) Where M(t) represents the change of the concentration of the target metabolite over time t; C x (t) is the active cell concentration function, which reflects the law of change of the number of active cells over time during the fermentation process; Φ(T,O2,N) is the interaction function of process parameters, which specifically describes the dynamic relationship between temperature T, dissolved oxygen concentration O2 and nutrient supply rate N in the fermentation environment, including the enhancement or inhibition of metabolic rate by changes in dissolved oxygen concentration; Γ(G) is the global target correction factor, which is used to adjust the metabolic kinetic model according to the set fermentation target G.

3. The fermentation equipment collaborative control method based on distributed control architecture according to claim 2 is characterized in that The task decomposition method based on global collaborative goals: uses a task decomposition algorithm to decompose complex goals into multiple specific subtasks, including temperature control, dissolved oxygen regulation, and nutrient supplementation; the task decomposition process integrates the independence of the subtasks, and the mutual coupling relationship and process constraints.

4. The fermentation equipment collaborative control method based on distributed control architecture according to claim 3 is characterized in that The task decomposition method based on global collaborative goals: After task decomposition, subtasks are assigned to distributed control nodes through a dynamic allocation algorithm. At the same time, the operating status of the nodes and the task execution efficiency are evaluated in real time. The optimization execution strategy includes adaptive adjustment according to the node performance. The specific optimization model is defined as: Among them, U i is the execution adjustment value of the i-th node, reflecting the adjustment range of the running state of the node after optimization; Δ(x) is the adaptive adjustment function, dynamically controlling the size of the adjustment range; Θ i,opt and θ i,act are the expected performance and actual performance of the i-th node, respectively, which are used to measure the deviation of the current node task execution; Ω(ρ i ,ξ i ) is a dynamic correction function based on the node load state ρ i and health factors i Ensure the reliability of task assignment.

5. The fermentation equipment collaborative control method based on distributed control architecture according to claim 1 is characterized in that The distributed adaptive control method based on real-time feedback includes: Key data of the fermentation environment including temperature and pH value are collected in real time and analyzed quickly. The edge computing unit deploys a lightweight algorithm to extract mutation features in environmental data through data denoising and anomaly detection. The core data feature extraction formula is defined as follows: Among them, S i (t) represents the data features of node i at time t; X i (t) is the real-time data collected by node i, including temperature, dissolved oxygen or pH value; ω is the dynamic weight factor, which adjusts the weight according to the importance of historical data and real-time data; f i (t) is the trend function, which captures the dynamic trend of data changes; Δt is the width of the analysis window, which is used to set the time span.

6. The fermentation equipment collaborative control method based on distributed control architecture according to claim 1 is characterized in that The distributed adaptive control method based on real-time feedback includes: establishing a distributed closed-loop feedback control model in a master control node; integrating data uploaded by distributed nodes, calculating global errors in real time, and generating a global optimization adjustment strategy.

7. The fermentation equipment collaborative control method based on distributed control architecture according to claim 1, characterized in that The distributed adaptive control method based on real-time feedback includes: The error is divided into global error and local error. The global error is generated by the master node through the global optimization algorithm to generate adjustment parameters, while the local error is quickly processed by the distributed node through the local adjustment algorithm. The optimization strategy generation formula is as follows: Among them, U′ i The optimization execution value of node i represents the specific operation amount of the node in the local adjustment; κ is the trade-off factor between global and local adjustments, which is used to balance the priority of global strategy and local optimization; Ψ is the global objective function, which is used to evaluate the effect of the optimization strategy in the global scope; Θ i is the current state parameter of node i; Ω(ρ i ,ξ i ) is the dynamic correction function, ρ i represents the node load status, ξ i represents the health status of the node; λ is the local optimization weight; F(Θ i ,L i ) is the local optimization function, Θ i is the node state, L i is the local error distribution.

Images