Cloud DCS platform computing resource adaptive allocation method and system based on control performance perception

Abstract

The invention discloses a cloud DCS platform computing resource adaptive allocation method and system based on control performance perception. An equipment access layer collects operation data of each subsystem of a whole plant in real time through an industrial protocol; the containerized platform layer is constructed based on Kubernetes, and pooling and elastic scheduling of calculation, storage and network resources are realized; the data middle platform layer is responsible for data storage, preprocessing, feature extraction and machine learning model training; a plurality of independent micro-services are deployed in the micro-service application layer, and intelligent analysis and optimization control are realized based on an autonomous learning model; the service grid layer ensures high availability and security of communication between micro-services; the unified configuration center realizes centralized management and operation and maintenance of whole system configuration, a traditional DCS monitoring system is reconstructed through a cloud native architecture, unification, intellectualization and operation and maintenance integration of whole plant level monitoring are realized, and the system resource utilization rate, the control precision and the operation reliability are improved.

Description

Technical Field

[0001] This application relates to the field of industrial control technology, and more specifically, to a method and system for adaptive allocation of computing resources on a cloud DCS platform based on control performance awareness. Background Technology

[0002] Traditional distributed control systems (DCS) in thermal power plants typically employ a hierarchical hardware and network architecture, with each unit, auxiliary system, and utility system operating independently, creating data silos. This architecture makes it difficult to achieve unified monitoring and coordinated control across the entire plant. When faced with complex operating environments and fluctuating load demands, it often relies on manual intervention and open-loop regulation, resulting in low operating efficiency, insufficient control precision, and limitations in scalability and resource utilization.

[0003] In recent years, with the advancement of the smart power plant concept, some research has attempted to introduce cloud computing technology into power plant control systems, building private cloud platforms to integrate computing and storage resources. However, existing solutions are mostly based on traditional virtualization or hyperconverged architectures, which, while improving resource utilization to some extent, still struggle to meet the stringent requirements of DCS for real-time performance, reliability, and security. Especially in Security Zone I environments, ensuring high availability, low latency, and data security for control operations remains a major challenge in the current construction of cloud-based DCS systems.

[0004] Cloud-native technologies, with their containerization, microservices, and continuous delivery features, offer new approaches to building elastic, scalable, and highly available distributed systems. However, directly applying cloud-native architectures to factory-level DCS unified monitoring systems still requires addressing key challenges such as real-time response to control operations, unified access to multi-source heterogeneous data, and cross-subsystem collaborative optimization. Therefore, there is an urgent need for a factory-level DCS unified monitoring system architecture and methodology that can both inherit the advantages of cloud-native technologies and adapt to the specific needs of industrial control environments. Summary of the Invention

[0005] The purpose of this application is to overcome the shortcomings of existing technologies and provide a method and system for adaptive allocation of computing resources on a cloud DCS platform based on control performance awareness. By constructing a cloud-native system architecture based on a Kubernetes containerized platform layer, a unified data middleware layer, and a microservice application layer, it solves the technical problems of data silos, resource rigidity, and inability to achieve unified intelligent monitoring and optimization control across the entire power plant DCS system.

[0006] The objective of this application is achieved through the following technical solution: Firstly, this application proposes a cloud DCS platform computing resource adaptive allocation method based on control performance awareness. The method is applied to a cloud DCS platform computing resource adaptive allocation system, which includes a device access layer, a containerized platform layer, a data middleware layer, a microservice application layer, a service mesh layer, and a unified configuration center. The method includes: The equipment access layer collects real-time operating data, equipment parameters, and environmental data from all unit generators, auxiliary systems, and utility systems throughout the plant. By using software-defined technology to pool computing, storage, and network resources through a containerized platform layer, a flexible and scalable resource pool is formed, supporting unified management and on-demand scheduling of resources. The collected data is transmitted and stored through the containerized database of the data middle platform layer. The stored data is preprocessed and its features are extracted through the big data platform of the data middle platform layer. Based on the preprocessed data, an autonomous learning model for intelligent analysis and optimization control is trained. Multiple independent monitoring and optimization microservices are deployed in the microservice application layer. Based on the self-learning model, optimization control strategies are generated and executed. The optimization control strategies are then distributed to relevant equipment or subsystems in the power plant for execution, thereby optimizing and controlling the operation of the power plant. The monitoring and optimization microservices include real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services. The service mesh layer manages communication, load balancing, and security policies between microservices. By centrally managing the configuration information, version control, and policy deployment of the entire plant's DCS monitoring system through a unified configuration center, integrated operation and maintenance can be achieved.

[0007] In one possible implementation, the containerization platform layer is deployed in the power plant's security zone I, and the computing, storage, and network resources of the server cluster are managed in a unified pool.

[0008] In one possible implementation, the machine learning model training in the data platform layer uses a gradient boosting decision tree algorithm to construct an autonomous learning model. The input of the autonomous learning model is the power plant operating parameters, equipment status and environmental factor data characteristics, and the output is the predicted value of power generation efficiency, energy consumption index or equipment life.

[0009] In one possible implementation, the intelligent analysis service in the microservice application layer is used to call the self-learning model trained by the data platform layer to analyze real-time and historical data, and send the analysis results to the optimization control service.

[0010] In one possible implementation, the optimization control service generates an optimized control strategy for the power plant equipment based on the received analysis results, and issues control commands for execution to perform closed-loop control of the equipment operation.

[0011] In one possible implementation, the service mesh layer manages communication between microservices to ensure collaborative operation between real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services.

[0012] In one possible implementation, a unified configuration center centrally manages the configuration of the entire plant's DCS monitoring system, supporting version control and unified deployment of configuration policies.

[0013] Secondly, embodiments of this application also propose a method and system for adaptive allocation of computing resources on a cloud DCS platform based on control performance awareness. The system includes a device access layer, a containerized platform layer, a data middleware layer, a microservice application layer, a service mesh layer, and a unified configuration center. The method includes: The equipment access layer connects to the power plant's DCS subsystem via industrial protocols to collect real-time operating data, equipment parameters, and environmental data from all unit generators, auxiliary systems, and utility systems throughout the plant. The containerization platform layer is built on Kubernetes and uses software-defined technology to pool computing, storage and network resources to form an elastic and scalable resource pool, supporting unified management and on-demand scheduling of resources; The data middle platform layer includes containerized databases and big data platforms, which uniformly store, clean, preprocess, extract features, and train machine learning models for the collected data. The microservice application layer deploys multiple independent monitoring and optimization microservices, which generate and execute optimization control strategies based on a self-learning model. The monitoring and optimization microservices include real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services. The service mesh layer manages communication, load balancing, and security policies between microservices. The unified configuration center centrally manages the configuration information, version control, and policy deployment of the entire plant's DCS monitoring system, enabling integrated operation and maintenance.

[0014] The main solution and its various further alternatives described above can be freely combined to form multiple solutions, all of which are solutions that can be adopted and are claimed in this application; furthermore, the (non-conflicting alternatives) can also be freely combined with each other and with other alternatives. Those skilled in the art, after understanding the solution of this application, will realize from the prior art and common general knowledge that there are many combinations, all of which are technical solutions to be protected in this application, and will not be exhaustively listed here.

[0015] This application discloses a method and system for adaptive allocation of computing resources on a cloud DCS platform based on control performance awareness. The device access layer collects operational data of all subsystems in the plant in real time through industrial protocols; the containerized platform layer is built on Kubernetes to realize the pooling and elastic scheduling of computing, storage, and network resources; the data middle platform layer is responsible for data storage, preprocessing, feature extraction, and machine learning model training; the microservice application layer deploys multiple independent microservices and realizes intelligent analysis and optimization control based on a self-learning model; the service mesh layer ensures high availability and security of communication between microservices; and the unified configuration center realizes centralized management and operation and maintenance of the entire system configuration. By reconstructing the traditional DCS monitoring system through a cloud-native architecture, the unified, intelligent, and integrated operation and maintenance of plant-wide monitoring is achieved, improving system resource utilization, control accuracy, and operational reliability. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating an adaptive allocation method for computing resources on a cloud DCS platform based on control performance awareness is shown. Detailed Implementation

[0018] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0019] Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] In order to address the various drawbacks of traditional ICS platform architecture in the prior art, this application proposes a cloud DCS platform computing resource adaptive allocation method and system based on control performance awareness. By introducing security mechanisms and backup strategies, the system improves the stability of smart power plant operation management and equipment control while ensuring the security of ICS cloud platform data and the reliability of the system.

[0021] Please refer to Figure 1 , Figure 1 A flowchart illustrating a cloud DCS platform computing resource adaptive allocation method based on control performance awareness is shown. This method is applied to a cloud DCS platform computing resource adaptive allocation system, which includes a device access layer, a containerized platform layer, a data middleware layer, a microservice application layer, a service mesh layer, and a unified configuration center. The method includes: The equipment access layer collects real-time operating data, equipment parameters, and environmental data from all unit generators, auxiliary systems, and utility systems throughout the plant. By using software-defined technology to pool computing, storage, and network resources through a containerized platform layer, a flexible and scalable resource pool is formed, supporting unified management and on-demand scheduling of resources. The collected data is transmitted and stored through the containerized database of the data middle platform layer. The stored data is preprocessed and its features are extracted through the big data platform of the data middle platform layer. Based on the preprocessed data, an autonomous learning model for intelligent analysis and optimization control is trained. Multiple independent monitoring and optimization microservices are deployed in the microservice application layer. Based on the self-learning model, optimization control strategies are generated and executed. The optimization control strategies are then distributed to relevant equipment or subsystems in the power plant for execution, thereby optimizing and controlling the operation of the power plant. The monitoring and optimization microservices include real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services. The service mesh layer manages communication, load balancing, and security policies between microservices. By centrally managing the configuration information, version control, and policy deployment of the entire plant's DCS monitoring system through a unified configuration center, integrated operation and maintenance can be achieved.

[0022] The device access layer, serving as the interface between the system and the physical world, securely and reliably connects to various DCS subsystems within the power plant, including distributed unit generators, auxiliary workshops, and public systems, through standard industrial protocols such as OPC UA and Modbus TCP. It collects operational data, key equipment parameters, and environmental information in real time and comprehensively, providing a unified data source for upper-layer applications. The containerization platform layer is the cloud-based cornerstone of the entire system, built on the industry-leading Kubernetes container orchestration engine. Through software-defined technologies, it abstracts and pools the underlying heterogeneous physical resources, such as computing servers, storage devices, and network devices, forming an elastic and scalable converged resource pool.

[0023] The data platform layer integrates a containerized, high-performance real-time / historical database with a big data processing platform. It is responsible for aggregating, uniformly storing, cleaning, standardizing preprocessing, and extracting deep features from massive, multi-source collected data. This layer provides a model training environment that can continuously train and optimize autonomous learning models for prediction and diagnosis based on historical and real-time data using machine learning algorithms, transforming data into intelligent assets that can directly serve production.

[0024] The microservice application layer carries specific monitoring and optimization business functions, breaking down traditional, large monolithic monitoring software into a set of independent, function-focused microservices, such as real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services. These services are independently developed, deployed, and expanded based on models and data provided by the data platform. Among them, the intelligent analysis service calls pre-trained models for deep analysis, while the optimization control service automatically generates and executes optimal control strategies accordingly.

[0025] To ensure the stability, efficiency, and security of the microservice architecture, a service mesh layer is introduced. This layer takes over all communication between microservices in a transparent manner, providing service discovery, intelligent load balancing, traffic governance, circuit breaking and rate limiting, and end-to-end security policy enforcement, thus ensuring the high availability, reliability, and isolation of the complex microservice network.

[0026] Finally, the unified configuration center enables centralized and versioned management of all microservice configurations, application versions, and business policies in the plant-wide DCS monitoring system. Operations personnel can use this center to perform one-click configuration distribution, canary releases, and rapid rollbacks, greatly simplifying the operational complexity of large-scale distributed systems.

[0027] The containerized platform layer is deployed in the power plant's security zone I, and it manages the computing, storage, and network resources of the server cluster in a unified pool.

[0028] The containerized platform layer is deployed in the power plant's production control area (i.e., Security Zone I). Based on the Kubernetes container orchestration engine, it abstracts, integrates, and uniformly pools the computing (CPU / memory), storage (including block storage and file storage), and network resources contained in the server cluster deployed in this area through software-defined technology.

[0029] The machine learning model training in the data platform layer uses the gradient boosting decision tree algorithm to build an autonomous learning model. The input of the autonomous learning model is the power plant operating parameters, equipment status and environmental factor data features, and the output is the predicted value of power generation efficiency, energy consumption index or equipment life.

[0030] The data platform layer employs the gradient boosting decision tree algorithm as the core method for constructing the autonomous learning model. This algorithm, through an ensemble learning framework, sequentially constructs multiple decision trees and continuously corrects the prediction residuals of the previous tree in the direction of gradient descent, thereby significantly improving the overall prediction accuracy and generalization ability of the model. The input to the autonomous learning model is a multi-dimensional feature vector processed by the data platform layer through cleaning, standardization, and feature engineering. This vector includes power plant operating parameters (such as unit load, main steam pressure and temperature, reheat steam temperature, feedwater flow rate, etc.), key equipment status (such as vibration spectrum of fans and pumps, bearing temperature, motor current, valve position feedback, etc.), and external environmental factors (such as ambient temperature, atmospheric pressure, cooling water temperature, etc.). The model's output corresponds to predicted values ​​of high-level performance indicators directly related to the power plant's economy, safety, and reliability, mainly including power generation efficiency (such as thermal efficiency, plant power consumption rate), energy consumption indicators (such as coal consumption for power generation, coal consumption for power supply), or remaining life predictions of key rotating equipment.

[0031] The intelligent analysis service in the microservice application layer is used to call the self-learning model trained by the data platform layer to analyze real-time and historical data, and send the analysis results to the optimization and control service.

[0032] The intelligent analytics service invokes a self-learning model deployed in the data platform layer, which has already undergone training and version management, to integrate and analyze real-time operational data from the device access layer and historical data stored in the data platform. Its workflow is as follows: After receiving a data request, the service retrieves the corresponding standardized feature data from the data platform and inputs it into a machine learning model such as Gradient Boosting Decision Tree (GBDT) to perform forward inference calculations. After completing the analysis, the intelligent analytics service sends the structured analysis results to the optimization and control service in real time and accurately through a predefined interface or asynchronous messaging mechanism.

[0033] Based on the received analysis results, the optimization control service generates optimized control strategies for power plant equipment and issues control commands for execution, thereby achieving closed-loop control of equipment operation.

[0034] The optimized control service receives deep analysis results (equipment sub-health warnings, optimal energy efficiency operating points, load allocation suggestions, or equipment life predictions, etc.) calculated based on a self-learning model from the intelligent analysis service. Based on preset optimization objectives (highest economic efficiency, strongest reliability, or optimal environmental performance) and constraints (equipment safe operating boundaries), it dynamically generates refined optimized control strategies for specific power plant equipment (steam turbine regulating valves, forced draft and induced draft fans, feedwater pumps, etc.) using a built-in optimization algorithm or strategy engine. Subsequently, the service parses the strategies into specific, executable control commands via a secure and reliable communication link and distributes them to the corresponding underlying DCS subsystems or field actuators.

[0035] The service mesh layer manages communication between microservices, ensuring collaborative operation between real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services.

[0036] The service mesh layer, by injecting a lightweight network proxy (Sidecar), takes over all communication traffic between the aforementioned services in a non-intrusive manner to the application. It provides a service discovery mechanism, enabling the intelligent analysis service to locate and invoke the fault warning service at any time; implements intelligent routing and load balancing to ensure that massive amounts of real-time data are stably and efficiently delivered from the monitoring service to the analysis service; executes circuit breaking, rate limiting, and retry policies to prevent the entire link from collapsing due to excessive instantaneous pressure on the optimization and control service; and enforces a secure communication policy based on authentication and encryption to ensure the confidentiality and integrity of analysis results and control commands during transmission.

[0037] The unified configuration center centrally manages the configuration of the entire plant's DCS monitoring system, and supports version control and unified deployment of configuration policies.

[0038] The unified configuration center provides unified configuration storage, access interfaces and management interface, supports versioned storage, change tracking and difference comparison of various configurations, and can deploy approved configuration policies to the corresponding containerized instances in the production environment in a one-click, batch and secure manner.

[0039] Compared with the prior art, the embodiments of this application have the following beneficial effects: First, by using a Kubernetes-based containerized platform layer, computing, storage, and network resources are pooled and uniformly scheduled, breaking down the resource silos of traditional DCS and significantly improving hardware resource utilization and investment efficiency.

[0040] Secondly, the data platform layer uses algorithms such as gradient boosting decision trees to train an autonomous learning model. The microservice application layer uses the model to realize intelligent analysis and optimization control, forming a closed-loop control, reducing manual intervention, thereby improving power generation efficiency, reducing energy consumption and optimizing equipment operation.

[0041] Third, the microservice architecture and service mesh enable each functional module to be developed, deployed and expanded independently; the unified configuration center realizes centralized and versioned management of configurations and one-click release, which greatly simplifies the complexity of operation and maintenance and supports rapid iteration and agile response of business.

[0042] The following describes a possible implementation of a cloud DCS platform computing resource adaptive allocation system based on control performance awareness. This system includes a device access layer, a containerized platform layer, a data middleware layer, a microservice application layer, a service mesh layer, and a unified configuration center. The method includes: The equipment access layer connects to the power plant's DCS subsystem via industrial protocols to collect real-time operating data, equipment parameters, and environmental data from all unit generators, auxiliary systems, and utility systems throughout the plant. The containerization platform layer is built on Kubernetes and uses software-defined technology to pool computing, storage and network resources to form an elastic and scalable resource pool, supporting unified management and on-demand scheduling of resources; The data middle platform layer includes containerized databases and big data platforms, which uniformly store, clean, preprocess, extract features, and train machine learning models for the collected data. The microservice application layer deploys multiple independent monitoring and optimization microservices, which generate and execute optimization control strategies based on a self-learning model. The monitoring and optimization microservices include real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services. The service mesh layer manages communication, load balancing, and security policies between microservices. The unified configuration center centrally manages the configuration information, version control, and policy deployment of the entire plant's DCS monitoring system, enabling integrated operation and maintenance.

[0043] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A cloud DCS platform computing resource adaptive allocation method based on control performance awareness, characterized in that, The method is applied to a cloud DCS platform adaptive allocation system for computing resources. The system includes a device access layer, a containerized platform layer, a data middleware layer, a microservice application layer, a service mesh layer, and a unified configuration center. The method includes: The equipment access layer collects real-time operating data, equipment parameters, and environmental data from all unit generators, auxiliary systems, and utility systems throughout the plant. By using software-defined technology to pool computing, storage, and network resources through a containerized platform layer, a flexible and scalable resource pool is formed, supporting unified management and on-demand scheduling of resources. The collected data is transmitted and stored through the containerized database of the data middle platform layer. The stored data is preprocessed and its features are extracted through the big data platform of the data middle platform layer. Based on the preprocessed data, an autonomous learning model for intelligent analysis and optimization control is trained. Multiple independent monitoring and optimization microservices are deployed in the microservice application layer. Based on the self-learning model, optimization control strategies are generated and executed. The optimization control strategies are then distributed to relevant equipment or subsystems in the power plant for execution, thereby optimizing and controlling the operation of the power plant. The monitoring and optimization microservices include real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services. The service mesh layer manages communication, load balancing, and security policies between microservices. By centrally managing the configuration information, version control, and policy deployment of the entire plant's DCS monitoring system through a unified configuration center, integrated operation and maintenance can be achieved.

2. The cloud DCS platform computing resource adaptive allocation method as described in claim 1, characterized in that, The containerized platform layer is deployed in the power plant's security zone I, and it manages the computing, storage, and network resources of the server cluster in a unified pool.

3. The cloud DCS platform computing resource adaptive allocation method as described in claim 1, characterized in that, The machine learning model training in the data platform layer uses the gradient boosting decision tree algorithm to build an autonomous learning model. The input of the autonomous learning model is the power plant operating parameters, equipment status and environmental factor data features, and the output is the predicted value of power generation efficiency, energy consumption index or equipment life.

4. The cloud DCS platform computing resource adaptive allocation method as described in claim 1, characterized in that, The intelligent analysis service in the microservice application layer is used to call the self-learning model trained by the data platform layer to analyze real-time and historical data, and send the analysis results to the optimization and control service.

5. The cloud DCS platform computing resource adaptive allocation method as described in claim 1, characterized in that, Based on the received analysis results, the optimization control service generates optimized control strategies for power plant equipment and issues control commands for execution, thereby achieving closed-loop control of equipment operation.

6. The cloud DCS platform computing resource adaptive allocation method as described in claim 1, characterized in that, The service mesh layer manages communication between microservices, ensuring collaborative operation between real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services.

7. The cloud DCS platform computing resource adaptive allocation method as described in claim 1, characterized in that, The unified configuration center centrally manages the configuration of the entire plant's DCS monitoring system, and supports version control and unified deployment of configuration policies.

8. A cloud DCS platform computing resource adaptive allocation system based on control performance awareness, characterized in that, The system includes a device access layer, a containerization platform layer, a data middleware layer, a microservice application layer, a service mesh layer, and a unified configuration center. The method includes: The equipment access layer connects to the power plant's DCS subsystem via industrial protocols to collect real-time operating data, equipment parameters, and environmental data from all unit generators, auxiliary systems, and utility systems throughout the plant. The containerization platform layer is built on Kubernetes and uses software-defined technology to pool computing, storage and network resources to form an elastic and scalable resource pool, supporting unified management and on-demand scheduling of resources; The data middle platform layer includes containerized databases and big data platforms, which uniformly store, clean, preprocess, extract features, and train machine learning models for the collected data. The microservice application layer deploys multiple independent monitoring and optimization microservices, which generate and execute optimization control strategies based on a self-learning model. The monitoring and optimization microservices include real-time monitoring services, intelligent analysis services, fault early warning services, and optimization control services. The service mesh layer manages communication, load balancing, and security policies between microservices. The unified configuration center centrally manages the configuration information, version control, and policy deployment of the entire plant's DCS monitoring system, enabling integrated operation and maintenance.

Images