An industrial pon-based distributed control network deployment system

Abstract

The application discloses a kind of distributed control network deployment systems based on industrial PON, it includes industrial network link deployment unit, builds industrial network link and equipment, collects power production control signal and power production related data;Wireless communication unit, the power production control signal of real-time acquisition is encrypted transmission, and non-real-time power production related data is transmitted through LoRa module;Intention drive management unit includes semantic analysis module, and the natural language command of power production control signal and power production related data is parsed using BERT model, and dynamic scheduling module is calculated and distributed to link load by dynamic load balancing algorithm;Block chain storage unit, the natural language command storage of power production control signal and power production related data is realized using alliance chain;Security isolation unit, firewall is based on SM4 algorithm of national secret ALC strategy is deployed, and one-way net gate isolates control area and management information area of industrial network link.

Description

Technical Field

[0001] This invention belongs to the field of industrial control network deployment systems, and more specifically, relates to a distributed control network deployment system based on industrial PON. Background Technology

[0002] Distributed control systems (DCS) are microprocessor-based systems that integrate computer technology, communication technology, control technology, and CRT display technology. The main characteristic of DCS is that it achieves centralized management and decentralized control through real-time monitoring. Since the first DCS system was introduced in the mid-1970s, modern DCS has undergone six generations of technological iteration and has been widely applied in the field of industrial control. The reliability, stability, and real-time performance of DCS have been fully verified in power plant operation, making it an indispensable core technology for power plant automation control. Currently, domestically produced DCS systems have been successfully applied in some large domestic generator units, gradually breaking the foreign monopoly on control systems. Domestic brands such as Hollysys, NARI Group, and NARI Technology have already occupied a certain market share in power generation control systems, making significant contributions to the development of industrial automation in my country.

[0003] The existing mainstream distributed control system network architectures are mainly divided into three types: First, the traditional Ethernet architecture, such as Emerson DeltaV, Honeywell Experion, Siemens PCS7, etc.; Second, the domestic security architecture, such as NARI maxCHD, Huaneng Ruiwo T316TR DCS, Kangjisen TSx Elite, etc.; Third, the cloud-edge collaborative architecture, such as China Mobile cloud resource deployment, Beijing Kaibo Wireless HFC network, etc.

[0004] The shortcomings of existing distributed control system network architectures are that the three network architectures mentioned above only apply network state scheduling of a single industrial network link and cannot adapt to different industrial network architecture environments; and they lack supervision and control over network link load. Summary of the Invention

[0005] To address the problem that existing distributed control system network architectures lack monitoring and control of network link load, rely solely on network status scheduling of a single industrial network link, and cannot adapt to different industrial network architecture environments, this invention provides a distributed control network deployment system based on industrial PON.

[0006] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0007] A distributed control network deployment system based on industrial PON includes an industrial network link deployment unit, a wireless communication unit, an intent-driven management unit, a blockchain storage unit, and a security isolation unit.

[0008] The industrial network link deployment unit communicates with the wireless communication unit to build industrial network links and equipment, and collect power production control signals and power production-related data.

[0009] The wireless communication unit communicates with the intent-driven management unit, including a hybrid network of 5G URLLC module and LoRa module. The 5G URLLC module encrypts and transmits real-time power production control signals, while the LoRa module transmits non-real-time power production related data.

[0010] The intent-driven management unit communicates with the blockchain evidence storage unit, including a semantic parsing module and a dynamic scheduling module. The semantic parsing module uses the BERT model to parse the natural language commands of power production control signals and power production-related data. The dynamic scheduling module collects the network status of industrial network links in real time through the SDN controller and calculates and distributes the link load through a dynamic load balancing algorithm.

[0011] The blockchain evidence storage unit communicates with the secure isolation unit and uses the Hyperledger Fabric consortium blockchain to realize the evidence storage of natural language commands for power production control signals and power production-related data.

[0012] The security isolation unit and firewall are based on the national cryptographic SM4 algorithm and deploy ALC policy. The one-way network gateway isolates the control area and management information area of ​​the industrial network link.

[0013] Furthermore, the BERT model uses the encoder part of the Transformer, which is composed of multiple Transformer blocks stacked together. Each Transformer block mainly contains two sub-layers: the Multi-Head Attention Sub-layer and the Feed-Forward Neural Network Sub-layer.

[0014] In the multi-head attention sublayer, the input is divided into two heads, including the input of power production control signals and power production related data; the attention weights of the two heads are calculated separately, and then the results are concatenated; this approach allows the model to capture semantic information in the input sequence from different perspectives, thereby improving the model's expressive power.

[0015] The feedforward neural network sublayer further processes the output of the multi-head attention sublayer. It consists of two fully connected layers with a ReLU activation function in between. The multi-head attention sublayer is mainly responsible for performing non-linear transformations on semantic information, enhancing the model's learning ability.

[0016] By stacking multiple Transformer blocks, BERT can continuously perform deep semantic encoding on the text of input power production control signals and power production-related data, thereby learning rich semantic representations.

[0017] Furthermore, the industrial network link deployment unit establishes industrial network links, including PON network links and AON network links;

[0018] The architecture of a PON network link includes an optical line terminal located at the service provider end, connected to multiple optical network units or optical network terminals via optical fibers, and distributed optical signals through a passive optical splitter.

[0019] The AON network link architecture is an optical switching network architecture, which includes hierarchical edge computing nodes and optical switching units. The optical access network connects user-side equipment and edge data center slave nodes (ECN-S). The optical switching network realizes the interconnection between edge computing nodes and connects edge data center master nodes (ECN-M) and cloud data centers through the optical transport network.

[0020] The optical switching unit consists of an arrayed waveguide grating (AWG), an optical splitter, a coupler, a head-up extraction module (HEM), a tunable filter (TF), a cyclic fiber delay line (FDL), and a high-speed optical switch (OS). The ECN-S output is connected to the input of the AWG in the OSU via TX. The AWG output is connected to port 1 of the HEM and the OSU via a 1×2 splitter. The output of the HEM is connected to the ECN-S. Port 2 of the OSU is connected to (m+1) OSs sequentially via a 1×(m+1) splitter. These OSs are connected to m ECN-Ss and one ECN-M within their respective slave node group via RX.

[0021] Furthermore, before calculating and allocating the link load, it is necessary to first determine whether the industrial network link of the industrial network deployment unit is a PON network link or an AON network link; at the same time, it is necessary to monitor whether the utilization rate of the main link is greater than the preset threshold. When the utilization rate of the main link is greater than the preset threshold, the dynamic load balancing algorithm is then used to calculate and allocate the link load to the backup link.

[0022] If the industrial network link is a PON network link, then the weighted least connection algorithm is used to calculate and allocate the link load;

[0023] If the industrial network link is an AON network link, then the objective function is established by estimating the edge processing cost, and the optimal solution for task allocation is obtained by solving the problem using a genetic algorithm.

[0024] Furthermore, when the link is a PON network, the dynamic load balancing algorithm is the weighted least connection algorithm, and the algorithm flow of the weighted least connection scheduling is as follows:

[0025] Suppose there is a set of servers S = {S0, S1, ..., S2} n-1}, W(S i ) indicates server S i The weights,

[0026] C(S i ) indicates server S i The current number of connections, the sum of the current number of connections for all servers is:

[0027] CSUM=ΣC(S i (i = 0, 1, ..., n-1);

[0028] The current new connection request will be sent to server S m If and only if server S m The following conditions must be met:

[0029] (C(S m ) / CSUM) / W(S m )=min{(C(Si) / CSUM) / W(Si)}(i=0, 1,···,n-1)}

[0030] Where W(S) i Since CSUM is not zero, it remains a constant in this round of search. The condition simplifies to:

[0031] C(S m ) / W(S m )=min{C(S i ) / W(S i (i = 0, 1, ..., n-1)

[0032] Where W(S) i ) is not zero;

[0033] Since division requires more CPU cycles than multiplication, and floating-point division is not allowed in the Linux kernel, and the weights of the server are all greater than zero, the condition c(S) is not met. m ) / W(S m )>C(S i ) / W(S i Further optimization to C(S) m )*W(S i )>C(S i )*W(S m At the same time, it ensures that the server is not scheduled when its weight is zero. Therefore, the algorithm executes the following process.

[0034]

[0035] Furthermore, in the AON network link, the objective function is established by estimating the edge processing cost, and the optimal solution for task allocation is obtained through a genetic algorithm. The detailed steps are as follows:

[0036] Read task information and edge data center information;

[0037] Calculate the longest waiting time T during the polling phase of each task. EN And the longest wait time T in FDL FDL ;

[0038] Express the objective functions: f1 delay and f2 total cost;

[0039] Solve using a genetic algorithm, and initialize the population;

[0040] Calculate the population fitness value;

[0041] Determine if the termination condition is met. If yes, output the optimal solution that minimizes the objective function and the corresponding objective function value. If no, return to the previous step to calculate the population fitness value through selection, crossover, and mutation.

[0042] Based on the optimal allocation solution, tasks are scheduled to the target edge data center in order of priority.

[0043] After obtaining the optimal solution for task allocation, the control unit of the AON network link generates control signals to control the switching state of OXC, and schedules the tasks to the corresponding edge data centers for processing and computation in sequence.

[0044] Furthermore, when it is an AON network link, the dynamic load balancing algorithm calculates and allocates the link load by estimating the edge processing cost. The objective functions for the edge processing cost are the delay f1(x) and the total cost f2(x).

[0045] The objective function time delay f1(x) is expressed as:

[0046] f1(x)=max(T P (R i )+T trans (R i ,x)+T cal (R i ,x)) (1≤x≤M)

[0047] T p (R i () represents the longest waiting time T for a task in the polling phase of the dynamic scheduling module. EN (R iThe average longest wait time T for a task in FDL FDL (R i The sum of the two; let the assigned edge data center number be x, 1≤x≤M, and the task transmission time T. trans (R i (x) represents the time it takes for the task to be transmitted from the FDL to the target edge data center via fiber optic cable; and is related to the task computational load d(R). i ) and the distance L of the edge data center x to be solved from the FDL fiber optic cable f (E x The direct proportion is expressed as:

[0048] T trans (R i ,x)=d(R i )*L f (E x )=d(R i )*L f (x(R i (1≤x≤M)

[0049] The computation time T of the task cal (R i x) represents the time required for the task to be completed in the target edge data center according to the solution allocation strategy, and is related to the task computation amount d(R). i It is directly proportional to the computing speed V of the edge data center to be solved. c (E x It is inversely proportional, expressed as:

[0050]

[0051] The total cost f2(x) of the objective function is expressed as:

[0052]

[0053] The total cost f2(x) represents the sum of the costs incurred by all tasks being computed in the edge data center. The cost of using each task in the edge data center can be expressed as the computation time T of the task in the target edge data center. cal (R i (x) and the unit time usage cost C of the edge data center p (E j The product of the two is represented as [the product of the two].

[0054] Furthermore, it also includes a protocol conversion unit, which is connected to the security isolation unit and enables Windows / Linux heterogeneous device collaboration and multiple protocol conversions through ROS2 middleware and OPC UA protocol.

[0055] Furthermore, the detailed process of using the Hyperledger Fabric consortium blockchain for natural language command notation is as follows:

[0056] Bind the currently collected power production control signals and power production-related data into a block;

[0057] Blocks are linked sequentially by storing the hash value of the previous block;

[0058] The linked data structure ensures that the data cannot be tampered with.

[0059] Compared with the prior art, the present invention has the following advantages:

[0060] By collecting and semantically parsing power production control signals and related data from different industrial network link architectures, natural language command parsing results are obtained and stored on the blockchain; this ensures accurate judgment of subsequent power faults. Secondly, by collecting and analyzing the network status of industrial network links in real time, results are obtained regarding whether uneven load distribution or overload occurs. Then, based on a dynamic load balancing algorithm, load data calculation and dynamic allocation are implemented for different industrial network link architectures; adapting to the processing and scheduling of different industrial network architecture environments. A unidirectional network gateway isolates the control area and management information area of ​​the industrial network link, ensuring the security of data transmission in and out of the entire distributed control network deployment system. Attached Figure Description

[0061] Figure 1 This is an overall structural block diagram of a distributed control network deployment system based on industrial PON in an embodiment of the present invention;

[0062] Figure 2 This is a system architecture diagram of the AON network link in an embodiment of the present invention;

[0063] Figure 3 This is a schematic diagram of the overall optical switching network architecture in an embodiment of the present invention;

[0064] Figure 4 This is a detailed flowchart illustrating the process of establishing an objective function for estimating edge processing costs and using a genetic algorithm to obtain the optimal solution for task allocation in this embodiment of the invention. Detailed Implementation

[0065] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0066] like Figure 1As shown, this embodiment provides a distributed control network deployment system based on industrial PON, including an industrial network link deployment unit, a wireless communication unit, an intent-driven management unit, a blockchain evidence storage unit, and a security isolation unit;

[0067] The industrial network link deployment unit communicates with the wireless communication unit to build industrial network links and equipment, and collect power production control signals and power production-related data.

[0068] The wireless communication unit communicates with the intent-driven management unit, including a hybrid network of 5G URLLC module and LoRa module. The 5G URLLC module encrypts and transmits real-time power production control signals, while the LoRa module transmits non-real-time power production related data.

[0069] The intent-driven management unit communicates with the blockchain evidence storage unit, including a semantic parsing module and a dynamic scheduling module. The semantic parsing module uses the BERT model to parse the natural language commands of power production control signals and power production-related data. The dynamic scheduling module collects the network status of industrial network links in real time through the SDN controller and calculates and distributes the link load through a dynamic load balancing algorithm.

[0070] The blockchain evidence storage unit communicates with the secure isolation unit and uses the Hyperledger Fabric consortium blockchain to realize the evidence storage of natural language commands for power production control signals and power production-related data.

[0071] The security isolation unit and firewall are based on the national cryptographic SM4 algorithm and deploy ALC policy. The one-way network gateway isolates the control area and management information area of ​​the industrial network link.

[0072] The BERT model uses the encoder part of the Transformer, which is composed of multiple Transformer blocks stacked together. Each Transformer block mainly contains two sub-layers: the Multi-Head Attention Sub-layer and the Feed-Forward Neural Network Sub-layer.

[0073] In the multi-head attention sublayer, the input is divided into two heads, including the input of power production control signals and power production related data; the attention weights of the two heads are calculated separately, and then the results are concatenated; this approach allows the model to capture semantic information in the input sequence from different perspectives, thereby improving the model's expressive power.

[0074] The feedforward neural network sublayer further processes the output of the multi-head attention sublayer. It consists of two fully connected layers with a ReLU activation function in between. The multi-head attention sublayer is mainly responsible for performing non-linear transformations on semantic information, enhancing the model's learning ability.

[0075] By stacking multiple Transformer blocks, BERT can continuously perform deep semantic encoding on the text of input power production control signals and power production-related data, thereby learning rich semantic representations.

[0076] The industrial network link deployment unit establishes industrial network links, including PON network links and AON network links;

[0077] The architecture of a PON network link includes an optical line terminal located at the service provider end, connected to multiple optical network units or optical network terminals via optical fibers, and distributed optical signals through a passive optical splitter.

[0078] like Figure 2 and 3 As shown, the AON network link architecture is an optical switching network architecture, which includes hierarchical edge computing nodes and optical switching units. The optical access network connects user-side equipment and edge data center slave nodes (ECN-S). The optical switching network realizes the interconnection between edge computing nodes, and the optical transport network connects the edge data center master node (ECN-M) and the cloud data center.

[0079] The optical switching unit consists of an arrayed waveguide grating (AWG), an optical splitter, a coupler, a head-up extraction module (HEM), a tunable filter (TF), a cyclic fiber delay line (FDL), and a high-speed optical switch (OS). The ECN-S output is connected to the input of the AWG in the OSU via TX. The AWG output is connected to port 1 of the HEM and the OSU via a 1×2 splitter. The output of the HEM is connected to the ECN-S. Port 2 of the OSU is connected to (m+1) OSs sequentially via a 1×(m+1) splitter. These OSs are connected to m ECN-Ss and one ECN-M within their respective slave node group via RX.

[0080] Edge data center nodes include multi-protocol interfaces, storage modules, analysis and computing modules, data assembly modules, optical transceiver modules, and wireless transceiver modules.

[0081] Before calculating and allocating the link load, it is necessary to first determine whether the industrial network link of the industrial network deployment unit is a PON network link or an AON network link; at the same time, it is necessary to monitor whether the utilization rate of the main link is greater than the preset threshold. If the utilization rate of the main link is greater than the preset threshold, the dynamic load balancing algorithm is then used to calculate and allocate the link load to the backup link.

[0082] If the industrial network link is a PON network link, then the weighted least connection algorithm is used to calculate and allocate the link load;

[0083] If the industrial network link is an AON network link, then the objective function is established by estimating the edge processing cost, and the optimal solution for task allocation is obtained by solving the problem using a genetic algorithm.

[0084] When using a PON network link, the dynamic load balancing algorithm is the weighted least-connection algorithm. Weighted Least-Connection Scheduling (WLS) is a superset of Least-Connection Scheduling, where each server is assigned a weight to represent its processing performance. The default weight for a server is 1, and system administrators can dynamically set the server's weight. WLS aims to proportionally allocate new connections to a server's established connections and its weight when scheduling new connections. The algorithm flow for WLS is as follows:

[0085] Suppose there is a set of servers S = {S0, S1, ..., S2} n-1}, W(S i ) indicates server S i The weights,

[0086] C(S i ) indicates server S i The current number of connections, the sum of the current number of connections for all servers is:

[0087] CSUM = ΣC(Si) (i = 0, 1, ..., n-1). The current new connection request will be sent to server S. m ,

[0088] Server Sm satisfies the following condition if and only if:

[0089] (C(S m ) / CSUM) / W(S m )=min{(C(Si) / CSUM) / W(Si)}(i=0, 1,···,n-1)}

[0090] Where W(S) i Since CSUM is not zero, it remains a constant in this round of search. The condition simplifies to:

[0091] C(S m) / W(S m )=min{C(S i ) / W(S i (i = 0, 1, ..., n-1)

[0092] Where W(S) i ) is not zero;

[0093] Since division requires more CPU cycles than multiplication, and floating-point division is not allowed in the Linux kernel, and the weights of the server are all greater than zero, the condition c(S) is not met. m ) / W(S m )>C(S i ) / W(S i Further optimization to C(S) m )*W(S i )>C(S i )*W(S m At the same time, it ensures that the server is not scheduled when its weight is zero. Therefore, the algorithm executes the following process.

[0094] for(m=0;m <n;m++){

[0095] if(W(S m )>0){

[0096] for(i=m+1;i <n;i++){

[0097] if(C(S m )*W(S i )>C(S i )*W(S m ))

[0098]

[0099] In an AON network link, after establishing the objective function by estimating the edge processing cost, the detailed steps for obtaining the optimal solution for task allocation using a genetic algorithm are as follows:

[0100] Read task information and edge data center information;

[0101] Calculate the longest waiting time T during the polling phase of each task. EN And the longest wait time T in FDL FDL ;

[0102] Express the objective functions: f1 delay and f2 total cost;

[0103] Solve using a genetic algorithm, and initialize the population;

[0104] Calculate the population fitness value;

[0105] Determine if the termination condition is met. If yes, output the optimal solution that minimizes the objective function and the corresponding objective function value. If no, return to the previous step to calculate the population fitness value through selection, crossover, and mutation.

[0106] Based on the optimal allocation solution, tasks are scheduled to the target edge data center in order of priority.

[0107] After obtaining the optimal solution for task allocation, the control unit of the AON network link generates control signals to control the switching state of OXC, and schedules the tasks to the corresponding edge data centers for processing and computation in sequence.

[0108] When it is an AON network link, the dynamic load balancing algorithm calculates and distributes the link load by estimating the edge processing cost, with the edge processing cost as the objective function, which is the delay f1(x) and the total cost f2(x).

[0109] The objective function time delay f1(x) is expressed as:

[0110] f1(x)=max(T P (R i )+T trans (R i ,x)+T cal (R i ,x)) (1≤x≤M)

[0111] T p (R i () represents the longest waiting time T for a task in the polling phase of the dynamic scheduling module. EN (R i The average longest wait time T for a task in FDL FDL (R i The sum of the two; let the assigned edge data center number be x, 1≤x≤M, and the task transmission time T. trans (R i (x) represents the time it takes for the task to be transmitted from the FDL to the target edge data center via fiber optic cable; and is related to the task computational load d(R). i ) and the distance L of the edge data center x to be solved from the FDL fiber optic cable f (E x The direct proportion is expressed as:

[0112] T trans (R i ,x)=d(R i )*L f (E x )=d(R i )*L f (x(R i (1≤x≤M)

[0113] The computation time T of the task cal (R i x) represents the time required for the task to be completed in the target edge data center according to the solution allocation strategy, and is related to the task computation amount d(R). i It is directly proportional to the computing speed V of the edge data center to be solved. c (E x It is inversely proportional, expressed as:

[0114]

[0115] The total cost f2(x) of the objective function is expressed as:

[0116]

[0117] The total cost f2(x) represents the sum of the costs incurred by all tasks being computed in the edge data center. The cost of using each task in the edge data center can be expressed as the computation time T of the task in the target edge data center. cal (R i (x) and the unit time usage cost C of the edge data center p (E j The product of the two is represented as [the product of the two].

[0118] It also includes a protocol conversion unit, which communicates with the security isolation unit and enables Windows / Linux heterogeneous device collaboration and multiple protocol conversions through ROS2 middleware and OPC UA protocol.

[0119] Detailed process of using Hyperledger Fabric consortium blockchain for natural language command notation:

[0120] Bind the currently collected power production control signals and power production-related data into a block;

[0121] Blocks are linked sequentially by storing the hash value of the previous block;

[0122] The linked data structure ensures that the data cannot be tampered with.

[0123] Compared with the prior art, the present invention has the following advantages:

[0124] By collecting and semantically parsing power production control signals and related data from different industrial network link architectures, natural language command parsing results are obtained and stored on the blockchain; this ensures accurate judgment of subsequent power faults. Secondly, by collecting and analyzing the network status of industrial network links in real time, results are obtained regarding whether uneven load distribution or overload occurs. Then, based on a dynamic load balancing algorithm, load data calculation and dynamic allocation are implemented for different industrial network link architectures; adapting to the processing and scheduling of different industrial network architecture environments. A unidirectional network gateway isolates the control area and management information area of ​​the industrial network link, ensuring the security of data transmission in and out of the entire distributed control network deployment system.

[0125] The above provides a detailed description of a distributed control network deployment system based on industrial PON provided in this application. The specific embodiments are described only to aid in understanding the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. An industrial PON-based distributed control network deployment system, characterized by, It includes an industrial network link deployment unit, a wireless communication unit, an intent-driven management unit, a blockchain evidence storage unit, and a security isolation unit; The industrial network link deployment unit communicates with the wireless communication unit to build industrial network links and equipment, and collect power production control signals and power production-related data. The wireless communication unit communicates with the intent-driven management unit, including a hybrid network of 5G URLLC module and LoRa module. The 5G URLLC module encrypts and transmits real-time power production control signals, while the LoRa module transmits non-real-time power production related data. The intent-driven management unit communicates with the blockchain evidence storage unit, including a semantic parsing module and a dynamic scheduling module. The semantic parsing module uses the BERT model to parse the natural language commands of power production control signals and power production-related data. The dynamic scheduling module collects the network status of industrial network links in real time through the SDN controller and calculates and distributes the link load through a dynamic load balancing algorithm. The blockchain evidence storage unit communicates with the secure isolation unit and uses the Hyperledger Fabric consortium blockchain to realize the natural language command evidence storage of power production control signals and power production-related data. The security isolation unit deploys ALC policy based on the national cryptographic SM4 algorithm through a firewall, and uses a one-way network gateway to isolate the control area and management information area of ​​the industrial network link. The industrial network link deployment unit establishes industrial network links, including PON network links and AON network links; The architecture of a PON network link includes an optical line terminal located at the service provider end, connected to multiple optical network units or optical network terminals via optical fibers, and distributed optical signals through a passive optical splitter. The AON network link architecture is an optical switching network architecture, which includes hierarchical edge computing nodes and optical switching units. The optical access network connects user-side equipment and edge data center slave nodes. The optical switching network realizes the interconnection between various edge computing nodes and connects edge data center master nodes and cloud data centers through the optical transport network. The optical switching unit consists of an arrayed waveguide grating, an optical splitter, a coupler, an all-optical packet header extraction module, a tunable filter, a cyclic fiber delay line, and a high-speed optical switch. Before calculating and allocating the link load, it is necessary to first determine whether the industrial network link of the industrial network deployment unit is a PON network link or an AON network link; at the same time, it is necessary to monitor whether the utilization rate of the main link is greater than the preset threshold. If the utilization rate of the main link is greater than the preset threshold, the dynamic load balancing algorithm is then used to calculate and allocate the link load to the backup link. If the industrial network link is a PON network link, then the weighted least connection algorithm is used to calculate and allocate the link load; If the industrial network link is an AON network link, then the objective function is established by estimating the edge processing cost, and the optimal solution for task allocation is obtained by solving the problem using a genetic algorithm.

2. The industrial PON-based distributed control network deployment system of claim 1, wherein, The BERT model uses the encoder part of the Transformer, which is composed of multiple Transformer blocks stacked together. Each Transformer block mainly contains two sub-layers: a multi-head attention sub-layer and a feedforward neural network sub-layer. In the multi-head attention sublayer, the input is divided into two heads, including the input of power production control signals and power production related data; the attention weights are calculated for each head, and then the results are concatenated. The feedforward neural network sublayer further processes the output of the multi-head attention sublayer. The multi-head attention sublayer is used to perform non-linear transformations on semantic information and consists of two fully connected layers with the ReLU activation function in between. By stacking multiple Transformer blocks, the entire BERT model performs deep semantic encoding on the text of the input power production control signals and power production-related data.

3. The industrial PON based distributed control network deployment system of claim 2, wherein, When the network is PON, the dynamic load balancing algorithm is the weighted least connection algorithm. The algorithm flow of weighted least connection scheduling is as follows: Let there be a set of servers S = {S0, S1, ···, S n-1 n} and let W(S i ) denote the weight of server S i . C(S i ) denotes the current connection number of server S i , and the total of all servers' current connection numbers is: CSUM =∑C(S i )(i=0,1,···,n-1). The current new connection request is sent to the server S m , the server S m satisfies the following conditions: (C(S m ) / CSUM) / W(S m )= min{(C(Si) / CSUM) / W(S i )} (i=0,1,···,n-1)} where W(S i ) is not zero, CSUM is constant in this round of lookup.

4. The industrial PON-based distributed control network deployment system of claim 3, wherein, In an AON network link, the objective function is established by estimating the edge processing cost, and the optimal solution for task allocation is obtained through a genetic algorithm. The detailed steps are as follows: Read task information and edge data center information; calculating the longest latency T for each task polling phase EN and the longest latency T in the FDL FDL ; Express the objective functions: f1(x) time delay and f2(x) total cost; Solve using a genetic algorithm, and initialize the population; Calculate the population fitness value; Determine whether the termination condition is met. If yes, output the optimal solution that minimizes the objective function and the corresponding objective function value. If not, then select, crossover, or mutate back to the previous step to calculate the population fitness value; Based on the optimal allocation solution, tasks are scheduled to the target edge data center in order of priority.

5. The industrial PON-based distributed control network deployment system of claim 4, wherein, When it is an AON network link, the dynamic load balancing algorithm calculates and distributes the link load by estimating the edge processing cost. The objective functions for the edge processing cost are the delay f1(x) and the total cost f2(x). The objective function time delay f1(x) is expressed as: T p (R i ) represents the longest waiting time T EN (R i ) of the task of the dynamic scheduling module in the polling stage; T FDL (R i ) is the sum of the two of the average longest waiting time T trans (R i ,x) is the transmission time of the task from the circulating fiber delay line through the optical fiber to the target edge data center. The task computation amount d(R i ) and the edge data center x distance circulating fiber delay line fiber length L f (E x ) are expressed in proportion to each other as follows: The computation time Tcal(Ri,x) of the task is the time it takes for the task to complete computation in the target edge data center according to the solution allocation strategy. It is directly proportional to the computational amount d(Ri) of the task and inversely proportional to the computational rate Vc(Ex) of the edge data center to be solved, expressed as: The total cost f2(x) of the objective function is expressed as: The total cost f2(x) represents the sum of the costs incurred by all tasks in the edge data center. The cost of each task in the edge data center can be represented by the product of the computation time Tcal(Ri,x) of the task in the target edge data center and the cost of using the edge data center per unit time Cp(Ej).

6. The industrial PON based distributed control network deployment system of claim 1, wherein, It also includes a protocol conversion unit, which communicates with the security isolation unit and enables Windows / Linux heterogeneous device collaboration and multiple protocol conversions through ROS 2 middleware and OPC UA protocol.

7. The industrial PON based distributed control network deployment system of claim 1, wherein, Detailed process of using Hyperledger Fabric consortium blockchain for natural language command notation: Bind the currently collected power production control signals and power production-related data into a block; Blocks are linked sequentially by storing the hash value of the previous block; The linked data structure ensures that the data cannot be tampered with.

Images