Integrated operation system and method for rail transit power supply

Abstract

The invention provides an integrated operation system for rail transit power supply, which comprises primary equipment, a sensing unit, a protocol conversion unit, a data lake unit, a service center and an application function unit, and is characterized in that the primary equipment is used for supporting transmission, distribution and conversion of electric energy; the sensing unit acquires the operation data of the primary equipment through multiple types of measurement and control devices; the protocol conversion unit converts the collected operation data into a unified standard format and sends the unified standard format to the data lake unit; the service center provides service support for the application function unit; and the application function unit performs data transmission with the data lake unit through the message middleware. According to the invention, through cooperation of the data lake unit, the service center and the application function unit, a faster response can be made to a high-real-time event, and the real-time processing capability is improved; and a rule engine is combined with a dynamic programming algorithm, so that an optimal calling strategy of high-level service to low-level service is realized, and service scheduling and resource configuration are optimized.

Description

Technical Field

[0001] The present invention relates to the field of rail transit power supply, and in particular to an integrated operation system and method for rail transit power supply. Background Art

[0002] The power supply system is a core component of rail transit systems, directly impacting the stable operation of trains, operational efficiency, and passenger safety. Rail transit systems are powered by electricity, and traction, signaling, and communications systems, as well as station equipment like air conditioning and lighting, all require a continuous and reliable power supply. Therefore, the reliability of the power supply system is crucial to ensuring the safe and stable operation of trains.

[0003] Currently, the operational systems within the subway power supply system primarily include power monitoring systems, online monitoring systems, and power supply operation safety management systems. Most of these systems were built at different stages, developed independently, and designed to meet their respective application requirements. However, in actual operation, these systems have gradually exposed issues such as incompatibility and difficulties in collaborative management. Firstly, the lack of information sharing mechanisms between these systems has led to the formation of "information silos," preventing efficient data flow between systems and impacting overall management effectiveness. Secondly, the lack of a unified data management and analysis platform forces each system to independently report data to a higher-level platform, resulting in fragmented and difficult-to-integrate data and hindering comprehensive analysis and decision support. Furthermore, insufficient standardization across these systems, with inconsistent technical protocols and device interface standards, complicates system interoperability and hinders the effective integration of data and resources. Furthermore, the independent operation of multiple systems requires redundant hardware configuration and wiring assistance, increasing investment costs and system maintenance complexity.

[0004] Therefore, it is necessary to improve the existing rail transit operation and maintenance system to overcome the defects of the existing technology. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an integrated operation system and method for rail transit power supply that can respond faster to high real-time events, improve the real-time processing capability of the rail transit operation and maintenance system, and optimize service scheduling and resource allocation.

[0006] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0007] First, the integrated operation system for rail transit power supply includes: primary equipment, sensing unit, protocol conversion unit, data lake unit, service center and application function unit, among which:

[0008] The primary devices are interconnected to form a power network for supporting the transmission, distribution and conversion of electric energy;

[0009] The sensing unit collects the operating data of the primary equipment through multiple types of measurement and control devices, and uploads the collected operating data to the protocol conversion unit;

[0010] The protocol conversion unit converts the collected operating data into a unified standard format, and performs denoising on the standard format data through a filtering algorithm to obtain standard denoised operating data and sends it to the data lake unit for storage and call;

[0011] The service center is connected to the data lake unit and the application function unit respectively, and is used to access the data lake data, process requests from the application function unit, and provide service support for the application function unit;

[0012] The application function unit is provided with a stream processing engine based on an event-driven mechanism, and the stream processing engine transmits data with the data lake unit through a message middleware.

[0013] Preferably, the primary equipment includes at least a transformer, a circuit breaker and an isolating switch, and the primary equipment is connected via a high-voltage cable or a busbar.

[0014] Preferably, the measurement and control device includes at least a relay protection device, an infrared thermal imager, a distributed temperature sensor, a smoke detector, a current transformer, and a voltage transformer.

[0015] Preferably, the stream processing engine adopts Kafka Streams, and the stream processing engine is used to process data streams.

[0016] Preferably, the sensing unit uploads the collected operating data to the protocol conversion unit via IEC61850 protocol, Modbus, IEC60870-5-101 / 104, or SNTP protocol.

[0017] In a second aspect, an integrated operation method for rail transit power supply is provided, which applies the integrated operation system for rail transit power supply as described above, and the method includes the following steps:

[0018] S1 Collection of operation data: collecting the original operation data of the primary equipment through the perception layer;

[0019] S2 Operation Data Storage Division: Based on timeliness and security requirements, the collected raw operation data is divided into three storage areas: Security I, Security II, and Security III of the data lake unit;

[0020] S3 role-based access control service access: The role-based access control mechanism divides services into four types: Security Zone I Basic Services, Security Zone II Basic Services, Security Zone III Basic Services, and Advanced Services. Different service types have different permissions to use the operating data in the three storage areas of the data lake unit;

[0021] S4. Service call based on adaptive dynamic programming and rule engine:

[0022] Based on the fault type, business requirements, and device status of a primary device, a rule engine is used to generate multiple basic service call strategies for operation data processing. Then, based on the adaptive dynamic programming algorithm, the optimal service call strategy for operation data processing is obtained.

[0023] Furthermore, the data contents stored in the three storage areas are:

[0024] The safety I storage area stores high-real-time control operation data, including voltage, current, active power, reactive power, power factor, active energy, reactive energy, position status, status of non-homologous devices, alarm signals, action signals, and self-diagnosis information data;

[0025] The Safety II storage area stores non-real-time control auxiliary data, including stray current, power metering, online monitoring and power quality monitoring data;

[0026] The Security III storage area stores management data that assists in operation and maintenance, including inspection videos, scheduling plans, work tickets, and operation ticket data.

[0027] Furthermore, the usage permissions for the operational data in the three storage areas of the data lake unit corresponding to different service types are as follows:

[0028] Security Zone 1 basic services allow reading and writing of operating data in all three storage zones;

[0029] Security Zone 2 basic services allow reading and writing of data in Security Zones II and III, and allow reading of data in Security Zone I under authorization;

[0030] Security Zone 3 basic services allow reading and writing of data in Security Zone III, and allow reading of data in Security Zones I and II under authorization;

[0031] Advanced services allow reading of data in security zone I under authorization, and allow reading and writing of data in security zones II and III under authorization.

[0032] Furthermore, the specific steps of the service call based on adaptive dynamic programming and rule engine in S4 are as follows:

[0033] S4-1. Establish different calling strategies for advanced services to basic services based on fault types and business requirements.

[0034] S4-2. Obtain the optimal service call strategy based on the adaptive dynamic programming algorithm:

[0035] S4-21: Modeling service resource consumption and response time:

[0036] The resource consumption status of the integrated operation system at the current moment is defined as:

[0037] S(k)=[c1(k),m1(k),b1(k),…,c M (k),m M (k),b M (k),T(k)];

[0038] Among them, c j (k),j∈[1,M] is the CPU usage of the jth service at time k, b j (k),j∈[1,M] is the network bandwidth consumption of the jth service at time k, m j (k),j∈[1,M] is the memory usage of the jth service at time k, and T(k) is the total response time of the system running services;

[0039] The resource consumption of calling a service in the integrated operation system at the operation state S(k) at time k is defined as:

[0040]

[0041] Among them, α j ,β j ,ε j are the importance weights of the jth service to CPU usage, network bandwidth, and memory usage, respectively;

[0042] Define the total response time of the service as:

[0043]

[0044] Among them, t j is the processing time of the jth service, d j The amount of data to be transmitted for service calls, b j Network bandwidth for the service;

[0045] Define the action space:

[0046] A={a k ∣a k =(s1,…,s M ),s j ∈{0,1}};

[0047] where s j =1 means in strategy a k Call the jth basic service;

[0048] S4-22: Introducing policy generation rules:

[0049] a k =RuleEngine(F t ,B t );

[0050] F t is the current fault type, B t For business requirement labels, the rule engine outputs a candidate strategy set {a k};

[0051] Define the Bellman optimality equation:

[0052] V * (S(k))=min a∈A {R(k)+γE[V * (S(k+1))]};

[0053] in Then, the optimal value function is approximated through the Critic-Executor architecture based on neural networks to obtain the optimal strategy and execute it.

[0054] Furthermore, the weight is updated in real time according to the system operation status, and w is defined as j =[α j ,β j ,ε j ], then the weight update formula is:

[0055]

[0056] Where Proj(·) is the operation of projecting the vector onto the three-dimensional probability simplex, η is the learning rate, and its value range is [0.01, 0.2]. The actual resource allocation weight w j The gradient of , λ is the collaborative penalty coefficient, and satisfies M is the number of microservices.

[0057] The above solution of the present invention includes at least the following beneficial effects:

[0058] The present invention ensures faster response to high-real-time operation data through the cooperation of data lake units, service centers and application function units, thereby improving the real-time processing capabilities of the overall system;

[0059] By combining the rule engine with the dynamic programming algorithm, the optimal calling strategy of high-level services to low-level services is implemented, service scheduling and resource allocation are optimized, and efficiency and quality are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a schematic diagram of the overall framework of an integrated operation system for rail transit power supply according to the present invention.

[0061] Figure 2 This is a flowchart of the steps of an integrated operation method for rail transit power supply of the present invention. DETAILED DESCRIPTION

[0062] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0063] like Figure 1 As shown, an embodiment of the present invention proposes an integrated operation system for rail transit power supply, which realizes the coordination of data collection, processing and application functions, and each layer and module of the system achieves flexible decoupling and efficient expansion through a microservice architecture.

[0064] Specifically, an integrated operation system for rail transit power supply according to an embodiment of the present invention includes: a primary device, a sensing unit, a protocol conversion unit, a data lake unit, a service center, and an application function unit, wherein:

[0065] The primary devices are interconnected to form a power network for supporting the transmission, distribution and conversion of electric energy.

[0066] Specifically, primary equipment, including key components like transformers, circuit breakers, and disconnectors, is responsible for the transmission, distribution, and conversion of electrical energy, providing fundamental support for rail transit power supply systems. These components are connected via high-voltage cables or busbars, forming the core architecture of the power network and ensuring stable power delivery to traction substations and other facilities along the line.

[0067] The sensing unit collects the operating data of the primary equipment through multiple types of measurement and control devices, and uploads the collected operating data to the protocol conversion unit through the IEC61850 protocol, Modbus, IEC60870-5-101 / 104, and SNTP protocol;

[0068] Specifically, the sensing unit is composed of multiple types of measurement and control devices, including: relay protection equipment that monitors power grid faults in real time and triggers protection actions; infrared thermal imagers that detect temperature anomalies at key nodes (such as cable joints and switch cabinets); distributed temperature sensors that collect local temperature rise data of equipment such as transformer windings; smoke detectors that prevent fire hazards; current transformers (CTs) and voltage transformers (PTs) that measure electrical quantities such as current and voltage, etc.

[0069] The protocol conversion unit converts the collected operating data into a unified standard format, and performs denoising on the standard format data through a filtering algorithm to obtain standard denoised operating data and sends it to the data lake unit for storage and call;

[0070] Specifically, the protocol conversion unit converts the data collected by the sensing unit into a unified standard format, ensuring that heterogeneous data from different devices can be uniformly processed and stored. It also uses filtering algorithms to denoise the collected data, eliminating power frequency interference and environmental noise. For example, missing CT samples are repaired using linear interpolation or Kalman filtering. Z-Score outlier detection is performed on infrared thermal imager temperature data to remove abnormal data. After protocol conversion and denoising, the data is sent to the data lake unit.

[0071] The data lake unit is mainly used to store data. It includes relational databases, real-time databases, and time series databases. It can store various structured, semi-structured, and unstructured data, and supports big data processing and fast query.

[0072] The service center is connected to the data lake unit and the application function unit respectively, and is used to access the data lake data, process requests from the application function unit, and provide service support for the application function unit;

[0073] In the service center, each microservice module communicates via a RESTful API. The service center performs permission authentication and role management for all services, ensuring that data access and operations are strictly controlled according to role permissions. The service center not only accesses the data lake unit but also processes requests from modules within the application functional units, providing diverse service support. Furthermore, to facilitate data exchange with other external systems, the permission authentication center is equipped with an external interface. Authorized systems can access relevant functions via HTTP, and relevant data can be uploaded to other external systems via MQTT.

[0074] The application function unit is provided with a stream processing engine based on an event-driven mechanism, and the stream processing engine transmits data with the data lake unit through a message middleware.

[0075] Specifically, an event-driven stream processing engine is introduced between the data lake unit and the application functional units. This stream processing engine transmits data to the data lake through a messaging middleware. The messaging middleware, Apache Kafka, implements a publish-subscribe mechanism for high-real-time data transmission, enabling high concurrency and low-latency data transmission, ensuring that relevant, high-real-time information can be uploaded quickly. The stream processing engine, Kafka Streams, processes the data stream. When the specified temperature rise or current anomaly conditions are met, it directly triggers the corresponding alarm function in the corresponding application functional layer according to preset rules, reducing latency.

[0076] This embodiment further provides an integrated operation method for rail transit power supply, which applies the integrated operation system for rail transit power supply described above. The method includes the following steps:

[0077] S1 Collection of operation data: collecting the original operation data of the primary equipment through the perception layer;

[0078] S2 Operation Data Storage Division: Based on timeliness and security requirements, the collected raw operation data is divided into three storage areas: Security I, Security II, and Security III of the data lake unit;

[0079] Specifically, based on the system architecture established in step 1, data is divided into safety zones I, II, and III according to timeliness and security requirements. Safety zone I stores high-real-time control data, such as voltage, current, active power, reactive power, power factor, active energy, reactive energy, position status, status of non-homologous devices, alarm signals, action signals, self-diagnosis information, opening and closing of circuit breakers (electric load switches, electric disconnectors, etc.), protection action tripping, and other required remote reset information, as well as protection and automatic device activation and deactivation information. Safety zone II stores non-real-time control auxiliary data, including stray current, power metering, online monitoring, and power quality monitoring data. Safety zone III stores management data for auxiliary operations and maintenance, including inspection videos, dispatch plans, work tickets, operation tickets, etc.

[0080] The data transmission requirements for Security Zone I are as follows:

[0081] Table 1

[0082]

[0083]

[0084]

[0085] S3 role-based access control service access: The role-based access control mechanism divides services into four types: Security Zone I Basic Services, Security Zone II Basic Services, Security Zone III Basic Services, and Advanced Services. Different service types have different permissions to use the operating data in the three storage areas of the data lake unit;

[0086] The permissions for using the operating data in the three storage areas of the data lake unit corresponding to different service types are as follows:

[0087] Security Zone 1 basic services allow reading and writing of operating data in all three storage zones;

[0088] Security Zone 2 basic services allow reading and writing of data in Security Zones II and III, and allow reading of data in Security Zone I under authorization;

[0089] Security Zone 3 basic services allow reading and writing of data in Security Zone III, and allow reading of data in Security Zones I and II under authorization;

[0090] Advanced services allow reading of data in security zone I under authorization, and allow reading and writing of data in security zones II and III under authorization.

[0091] The details are shown in Table 2 below:

[0092] Table 2

[0093]

[0094]

[0095] S4. Service call based on adaptive dynamic programming and rule engine:

[0096] Based on the fault type, business requirements, and device status of a primary device and combined with the rule engine, a variety of basic service call strategies for operation data processing are generated. Then, based on the adaptive dynamic programming algorithm, the optimal service call strategy for operation data processing is obtained.

[0097] Furthermore, the specific steps of the service call based on adaptive dynamic programming and rule engine in S4 are as follows:

[0098] S4-1. Establish different calling strategies for advanced services to basic services based on the type of primary equipment failure and business needs.

[0099] The relationship between advanced services and basic services is as follows: when the system is running, multiple applications are executed, and the applications will call multiple advanced and basic services. Advanced services call different basic services according to business needs.

[0100] For example, advanced fault recovery services use different invocation strategies depending on the fault location. For a DC bus fault, the fault recovery service implements a large bilateral power supply strategy. First, the remote control service is invoked to trip the corresponding switch at the adjacent station. The telesignaling service is then invoked to verify the switch is closed. Next, the remote control service is invoked to trip the local on-line disconnector. The telesignaling service is then invoked to verify that the inter-station disconnector meets closing requirements. Finally, the remote control service is invoked to close the local inter-station disconnector and the adjacent station's switch. Finally, the telemetry service is invoked to monitor the voltage and current recovery in the de-energized section.

[0101] For section ring network faults, the fault recovery service executes the backup automatic switching strategy, calls the remote control service to cut off the third-level load within the power supply range of the station, then calls the remote control service to close the 33kV bus tie switch, and finally calls the telemetry service to monitor the voltage and current recovery of the power-lost section.

[0102] For AC 33kV busbar faults, the fault recovery program implements the low-voltage standby automatic re-closing strategy. First, the telesignaling service is called to ensure that the feeder switches and 33kV bus tie switches of the fault section bus are all in the open position. Then the remote control service is called to cut off the three-level load switch of the low-voltage system. Then the remote control service is called to close the low-voltage bus tie switch. Finally, the telemetry service is called to monitor the recovery of the power-off area.

[0103] Different high-level services may have the same low-level service call requirements. For example, in the high-level service fault recovery and sequence control call strategy based on the requirements, the remote control service control switch may be called at the same time.

[0104] S4-2. Based on the adaptive dynamic programming algorithm, the optimal service call strategy for running data processing is obtained:

[0105] S4-21: Modeling the CPU usage, network bandwidth and memory resource consumption of the integrated operation system and the response time of operation data processing:

[0106] The resource consumption state of the integrated operation system at the current k moment is defined as:

[0107] S(k)=[c1(k),m1(k),b1(k),…,c M (k),m M (k),b M (k),T(k)], where c j (k),j∈[1,M] is the CPU usage of the jth service at time k, b j (k),j∈[1,M] is the network bandwidth consumption of the jth service at time k, m j (k),j∈[1,M] is the memory usage of the jth service at time k, and T(k) is the total response time of the system running services.

[0108] The resource consumption of calling a service in the integrated operation system at the operation state S(k) at time k is defined as:

[0109]

[0110] Among them, α j ,β j ,ε j are the importance weights of the jth service to CPU usage, network bandwidth, and memory usage, respectively;

[0111] Define the total response time of the service running data processing as:

[0112]

[0113] Among them, t j is the processing time of the jth service, d j The amount of data to be transmitted for service calls, b j Network bandwidth for the service;

[0114] Define the action space:

[0115] A={a k ∣a k =(s1,…,s M ),s j ∈{0,1}};

[0116] where s j =1 means in strategy a k Call the jth basic service;

[0117] S4-22: Introducing policy generation rules:

[0118] a k =RuleEngine(F t ,B t );

[0119] F t is the current fault type, B t For business requirement labels, the rule engine outputs a candidate strategy set {a k The policy set contains one or more policies for calling low-level services from high-level services that need to be run;

[0120] Define the Bellman optimality equation:

[0121] V * (S(k))=min a∈A {R(k)+γE[V * (S(k+1))]};

[0122] in The optimal value function of the Bellman optimal equation is approximated by the Critic-Executor architecture based on a neural network, corresponding to the optimal action a * =argmin a∈A {R(k)+γE[V * (S(k+1))]} is the optimal strategy, which is one or more basic services that should be executed first according to the current resource consumption status of the integrated operation system.

[0123] Furthermore, the weight is updated in real time according to the system operation status, and w is defined as j =[α j ,β j ,ε j ], then the weight update formula is:

[0124]

[0125] Where Proj(·) is the operation of projecting the vector onto the three-dimensional probability simplex, η is the learning rate, and its value range is [0.01, 0.2]. The actual resource allocation weight w j The gradient of , λ is the collaborative penalty coefficient, and satisfies M is the number of microservices.

[0126] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An integrated operation system for rail transit power supply, characterized in that: include: Primary equipment, perception unit, protocol conversion unit, data lake unit, service center, and application function unit, including: The primary devices cooperate with each other to form a power network, which is used to support the transmission, distribution and conversion of electric energy; The sensing unit collects the operating data of the primary equipment through multiple types of measurement and control devices, and uploads the collected operating data to the protocol conversion unit; The protocol conversion unit converts the collected operating data into a unified standard format, and performs denoising on the standard format data through a filtering algorithm to obtain standard denoised operating data and sends it to the data lake unit for storage and call; The service center is connected to the data lake unit and the application function unit respectively, and is used to access the data lake data, process requests from the application function unit, and provide service support for the application function unit; The application function unit is provided with a stream processing engine based on an event-driven mechanism, and the stream processing engine transmits data with the data lake unit through a message middleware.

2. The integrated operation system for rail transit power supply according to claim 1, characterized in that: The primary equipment at least includes a transformer, a circuit breaker and an isolating switch, and the primary equipment is connected via a high-voltage cable or a busbar.

3. The integrated operation system for rail transit power supply according to claim 1, characterized in that: The measurement and control device at least includes a relay protection device, an infrared thermal imager, a distributed temperature sensor, a smoke detector, a current transformer and a voltage transformer.

4. The integrated operation system for rail transit power supply according to claim 1, characterized in that: The stream processing engine adopts Kafka Streams, and the stream processing engine is used to process data streams.

5. The integrated operation system for rail transit power supply according to claim 1, characterized in that: The sensing unit uploads the collected operating data to the protocol conversion unit through the IEC61850 protocol, Modbus, IEC60870-5-101 / 104, and SNTP protocol.

6. An integrated operation method for rail transit power supply, applying an integrated operation system for rail transit power supply according to any one of claims 1 to 5, characterized in that: The method comprises the following steps: S1 Collection of operation data: collecting the original operation data of the primary equipment through the perception layer; S2 Operation Data Storage Division: The collected raw operation data is divided and stored according to timeliness and security requirements. The operation data is stored in the three storage areas of Security I, Security II, and Security III of the data lake unit according to the division type; S3 role-based access control service access: The role-based access control mechanism divides services into four types: Security Zone I Basic Services, Security Zone II Basic Services, Security Zone III Basic Services, and Advanced Services. Different service types have different permissions to use the operating data in the three storage areas of the data lake unit; S4. Service call processing based on adaptive dynamic programming and rule engine for operation data: Based on the fault type, business requirements and equipment status of a primary device, combined with the rule engine, multiple basic service call strategies for operation data processing are generated. Then, based on the adaptive dynamic programming algorithm, the optimal service call strategy for operation data processing is obtained.

7. The integrated operation method for rail transit power supply according to claim 6, characterized in that: The data contents stored in the three storage areas are: The safety I storage area stores high-real-time control operation data, including voltage, current, active power, reactive power, power factor, active energy, reactive energy, position status, status of non-homologous devices, alarm signals, action signals, and self-diagnosis information data; The Safety II storage area stores non-real-time control auxiliary data, including stray current, power metering, online monitoring and power quality monitoring data; The Security III storage area stores management data that assists in operation and maintenance, including inspection videos, scheduling plans, work tickets, and operation ticket data.

8. The integrated operation method for rail transit power supply according to claim 6, characterized in that: The permissions for using the operating data in the three storage areas of the data lake unit corresponding to different service types are as follows: Security Zone 1 basic services allow reading and writing of operating data in all three storage zones; Security Zone 2 basic services allow reading and writing of data in Security Zones II and III, and allow reading of data in Security Zone I under authorization; Security Zone 3 basic services allow reading and writing of data in Security Zone III, and allow reading of data in Security Zones I and II under authorization; Advanced services allow reading of data in security zone I under authorization, and allow reading and writing of data in security zones II and III under authorization.

9. The integrated operation method for rail transit power supply according to claim 6, characterized in that: The specific steps of the service call based on adaptive dynamic programming and rule engine of S4 are as follows: S4-1. Establish different call strategies for advanced services to basic services based on the type of primary equipment failure and business needs; S4-2. Obtain the optimal service call strategy based on the adaptive dynamic programming algorithm: S4-21: Modeling service resource consumption and response time: The resource consumption status of the integrated operation system at the current moment is defined as: S(k)=[c1(k),m1(k),b1(k),…,c M (k),m M (k),b M (k),T(k)]; Among them, c j (k),j∈[1,M] is the CPU usage of the jth service at time k, b j (k),j∈[1,M] is the network bandwidth consumption of the jth service at time k, m j (k),j∈[1,M] is the memory usage of the jth service at time k, and T(k) is the total response time of the system running services; The resource consumption of calling a service in the integrated operation system at the operation state S(k) at time k is defined as: Among them, α j ,β j ,ε j are the importance weights of the jth service to CPU usage, network bandwidth, and memory usage, respectively; Define the total response time of the service as: Among them, t j is the processing time of the jth service, d j The amount of data to be transmitted for service calls, b j Network bandwidth for the service; Define the action space: A={a k ∣a k =(s1,…,s M ),s j ∈{0,1}}; where s j =1 means in strategy a k Call the jth basic service; S4-22: Introducing policy generation rules: a k =RuleEngine(F t ,B t ); F t is the current fault type, B t For business requirement labels, the rule engine outputs a candidate strategy set {a k }; Define the Bellman optimality equation: V * (S(k))=min a∈A {R(k)+γE[V * (S(k+1))]}; in Then, the optimal value function is approximated through the Critic-Executor architecture based on neural networks to obtain the optimal strategy and execute it.

10. The integrated operation method for rail transit power supply according to claim 9, characterized in that: The weight is updated in real time according to the system operation status, and w is defined j =[α j ,β j ,ε j ], then the weight update formula is: Where Proj(·) is the operation of projecting the vector onto the three-dimensional probability simplex, η is the learning rate, and its value range is [0.01, 0.2]. The actual resource allocation weight w j The gradient of , λ is the collaborative penalty coefficient, and satisfies M is the number of microservices.

Images