CN104809593A - Distributed type parallel computation management method of power system - Google Patents

Abstract

A method for managing distributed parallel computing in a power system, the method comprising: A. When any user of any subsystem initiates any advanced application calculation at any time, the user sends a calculation request to the coordination management object through the user management object ; B. The coordination management object issues instructions to other subsystems to start computing; C. The subsystem management objects of other subsystems open up new computing processes in their respective subsystems, and create shadow users in the process; D. Shadow user users The management object establishes a connection with the coordination management object, and prepares to decompose the coordination calculation; E, each user and the coordination layer respectively establish calculation objects; F, the coordination management object assigns three-dimensional coordinates to all calculation objects, and all calculation objects are mutually positioned through the three-dimensional coordinates and communication; G. Coordinate computing objects to complete the decomposition and coordination between computing objects, and each computing object completes iterations according to the decomposition coordination mechanism and the subsystem master-slave mechanism.

Description

Distributed Parallel Computing Management Method for Power System

technical field

The invention relates to the technical field of power system analysis, in particular to the technical field of power system energy management.

Background technique

my country's power production management has formed a complete set of "level management, layered control, step-by-step processing" system. Dispatching centers at all levels establish a more detailed power system model for the power grids within their jurisdiction, but each sub-regional power grid does not know the operating parameters and status of other sub-regional power grids. From the perspective of physical structure, the power grids are interconnected, and each power grid cannot ignore other networks to calculate and manage the local power grid in isolation. The physical changes on the power components affect each other and restrict each other through the grid connection, making the power system gradually present the characteristics of complex models and wide-area distribution of data. On the other hand, the emergence of the electricity market puts forward many non-technical requirements for the operation of the power grid. The contradiction between the interconnection of the power grid and the hierarchical and partitioned management system of the power system determines the need for decomposition and coordination calculations among multiple control centers.

At the same time, the safety of the power system not only depends on the operation, analysis and control technology of the power system, but also is subject to the effectiveness and reliability of the information system. And other factors are closely related, and the strong grid structure may still be shaken by a series of successive faults and improper operations, and even a system-wide blackout. For the power system, it is extremely urgent to realize online security quantitative assessment, self-adaptive optimization and coordination of each line of defense, and construct a space-time coordinated security and stability defense framework. Reflecting on several major power outages, the new generation of energy management systems should establish a complete space-time coordination mechanism: from isolated defense to comprehensive defense, scheduling and management based on space-time coordination. This requires that the decomposition and coordination calculation of multi-control centers distributed in space can be implemented concurrently in time and application.

On the other hand, distributed dynamic power flow, distributed state estimation, static security analysis and other distributed application computing research in power systems have gradually matured and have been practically applied in power systems. For energy management systems (Energy Management System, EMS) to coordinate various The capability of parallel computing for distributed applications puts forward more requirements. Literature [1] (Zhang Boming, Zhang Haibo. Research on the decomposition and coordination mode between multiple control centers [J]. Chinese Journal of Electrical Engineering, 2006, 26(22): 1-5) From the perspective of practical application, the realization of The development rules that the decomposition and coordination calculation of multiple control centers should follow, and the specific implementation forms and algorithms of the power flow decomposition and coordination calculation of the interconnected system under the three distributed computing modes of synchronous iteration, asynchronous iteration and real-time equivalent are respectively proposed. Literature [2] (Zhang Haibo, Jiang Liangmin, Tao Wenwei. The design and implementation of a practical distributed dynamic power flow calculation system [J]. Electric Power System Automation, 2012,36(9):67-71) introduces registration, activation, and start-up The distributed management mechanism solves the problem of conflict resolution when multiple subsystems initiate calculations at the same time, and conducts corresponding practical tests. Literature [3] (Zhang Haibo, Yi Wenfei. Distributed state estimation method of power system based on asynchronous iterative mode [J]. Power System Automation, 2014,38(9):125-131) proposed a method based on equivalence network The distributed state estimation algorithm based on the idea of equivalent measurement correction and the bad data identification method considering the external network measurement information have studied and verified the practical implementation process of state estimation decomposition coordination calculation.

So far, the existing technology still lacks the practical design of multi-control center decomposition and coordination calculation mode, the lack of consideration of space-time application of three-dimensional coordinates for distributed parallel computing system management mechanism, and the technology of building parallel computing systems and applying them to distributed energy management systems This makes the specific implementation of distributed parallel computing in power systems difficult.

Contents of the invention

In view of this, the purpose of the present invention is to overcome the system management problems existing in the prior art that the multi-control center decomposition and coordination calculation mode is applied to practice, and to design a distributed parallel computing management method for the power system, which is distributed and multi- The application provides a suitable access platform to meet the requirements of the future grid for the new generation of energy management systems.

In order to achieve this goal, the technical solution adopted by the present invention is as follows.

A distributed parallel computing management method for a power system, the method comprising:

A. When a user of a subsystem initiates an advanced application calculation, the user sends a calculation request to the coordination management object through its user management object;

B. Coordinate management objects to issue start calculation instructions to other subsystems;

C. The subsystem management objects of the other subsystems open up new computing processes in their respective subsystems, and create shadow users in the process, and cooperate with the computing users to perform distributed computing;

D. The user management object of the shadow user establishes a connection with the coordination management object, and prepares for decomposition and coordination calculation;

E. The user management objects of the user and the shadow user establish a master user computing object and a secondary user computing object, and coordinate management objects to establish a coordination computing object;

F. The coordination management object assigns three-dimensional coordinates to all calculation objects, and all calculation objects locate and communicate with each other through three-dimensional coordinates;

G. Coordinate calculation objects to complete the decomposition and coordination between each calculation object, and each calculation object completes iterative calculation according to the decomposition coordination mechanism and the subsystem master-slave mechanism;

H. After the calculation is completed, the main user calculation object outputs the calculation result;

I. Any user in the computing system initiates any kind of advanced application computing, and completes a computing process according to the above steps.

Each calculation instance in the method includes calculation objects formed by three-dimensional coordinates to form data layers. Different users belonging to the same or different subsystems can start the same or different advanced application calculations at the same time without affecting each other. The parallel completion, the three-dimensional coordinates are time, space and application three-dimensional three-dimensional coordinates, wherein,

The time dimension coordinates represent the different users participating in the computation,

The spatial dimension coordinates indicate the different subsystems to which the computing objects belong,

Application dimension coordinates represent different high-level application calculations.

In addition, the method extends the decomposition and coordination calculation mode between multiple control centers to multi-user and multi-application fields, ensuring the correctness of the distributed calculation of each subsystem when multiple distributed applications of the energy management system are connected and multiple users are started. , the advanced application computing can be connected to various power system distributed advanced applications of the energy management system, including state estimation calculation, power flow calculation and static security analysis.

In addition, before the step A, it also includes:

A01. Set up a coordination layer for distributed computing in the superior control center, and the coordination layer has coordination management functions and coordination computing functions;

A02. Use the control centers at the lower level to act as various subsystems. Each subsystem is connected to multiple users. The subsystems have subsystem management functions, and the users have user management functions and user computing functions.

The management and calculation functions of the coordination layer, subsystems and users are realized by objects, the coordination layer has coordination management objects and coordination calculation objects, the subsystem has subsystem management objects, and the user has user management objects and user Calculate objects, where,

Coordination management objects are used to manage the entire computing system, including the management of subsystem status and user status, and management of all computing processes. It is at the core and lasts for the entire computing system;

The coordination calculation object is used to act as the communication agent of each subsystem, and complete the boundary coordination process in the specific calculation process, and the duration is the duration of one calculation;

The subsystem management object is used to manage the status of the subsystem, manage the creation of user processes, the establishment of computing objects, and the process of extinction, and the duration is the duration of the subsystem;

The user management object is used to manage the status of the user and establish the communication process between the user and the coordination layer, and the duration is the duration of the user;

The user calculation object is used to complete the actual calculation process, communicate with the coordination calculation object through the interface, and the duration is the duration of one calculation.

Particularly, step G further comprises:

G1. Before the calculation starts, the coordination layer reads the connection line information between the subsystems, and divides the subsystems according to the subsystem master-slave mechanism. Each subsystem only has the network model of the subsystem, and the network structure is equivalent to the intranet before sending to the coordination layer;

G2. After the coordination layer obtains the equivalent model of the internal network of each subsystem, it combines the connection line information to establish the equivalent model of the external network of each subsystem, and sends it to the corresponding subsystem;

G3. The subsystem accepts the external network model to form the grid computing topology of the subsystem;

G4. After the calculation starts, each subsystem sends the boundary coordination information corresponding to the application calculation to the coordination layer during each iteration. Subsystems, each subsystem completes iterative calculations in combination with boundary information.

The buffer network is determined by using the decoupled linearization distribution factor of the power of the tie line or the voltage of the boundary node relative to the components of the external network from the boundary node or the boundary station outside the original tie line of the current subsystem, and is included in the detailed modeling of the internal network of the subsystem .

The boundary coordination information exchanged by different computing applications is different. The dynamic power flow calculation is the diagonal element of the inverse matrix of the boundary impedance matrix, and the state estimation calculation is the diagonal element of the inverse matrix of the information matrix.

In addition, the communication mode between the coordination layer and the subsystems and users is as follows:

Establish a communication connection between the coordination layer and each subsystem;

The coordination layer establishes a communication connection between each user and the shadow user;

No blocking interference between multiple communication connections in the coordination layer;

The communication connection between the user and the coordination layer is an asynchronous communication connection.

The method also includes a user management step, the user management step comprising:

After the user starts, establish a communication link with the coordination layer;

Record the current user status, which includes registration and activation, representing the permissions of the current user;

After receiving the user state switching request, forward the request to the coordination layer and wait for the response;

Accept and record user state switching results,

Create a calculation object after accepting the start calculation command.

By adopting the power system distributed parallel computing management method of the present invention, the multi-control center decomposition and coordination computing mode is extended to multi-user and multi-application fields for practice, and the parallel computing problem of distributed computing in the future distributed energy management system is solved. It provides a suitable access platform for the distributed multi-application of the mature power system, and provides a solution for the new generation of three-dimensional coordinated energy management system.

Description of drawings

Fig. 1 is a schematic diagram of functional modules of a distributed parallel computing management method for a power system in a specific embodiment of the present invention.

Fig. 2 is an expanded schematic diagram of the tie-line of the main subsystem of the distributed parallel computing management system of the power system in the specific embodiment of the present invention.

Fig. 3 is a diagram showing the correspondence between users of each subsystem of the power system distributed parallel computing management system in a specific embodiment of the present invention.

Fig. 4 is a distributed parallel computing system object setting table of the power system distributed parallel computing management method in the specific embodiment of the present invention.

Fig. 5 is a logical relationship diagram of each object in the power system distributed parallel computing management method in the specific embodiment of the present invention.

Fig. 6 is a schematic diagram of three-dimensional coordinate setting of the distributed parallel computing management method of the power system in the specific embodiment of the present invention.

Fig. 7 is a schematic diagram of the data layering effect of the power system distributed parallel computing management method in the specific embodiment of the present invention.

Fig. 8 is a diagram of the deployment mode of the distributed parallel computing management system of the power system in the specific embodiment of the present invention.

Fig. 9 is a schematic diagram of the master-slave mechanism division of subsystems in the power system distributed parallel computing management method in the specific embodiment of the present invention.

Fig. 10 is a table of extension results of tie lines in the power system distributed parallel computing management method in the specific embodiment of the present invention.

Fig. 11 is a parallel computing effect test table of the power system distributed parallel computing management method in the specific embodiment of the present invention.

Fig. 12 is a multi-user parallel computing test result table of the power system distributed parallel computing management method in the specific embodiment of the present invention.

Fig. 13 is a multi-application parallel computing test result table of the power system distributed parallel computing management method in the specific embodiment of the present invention.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings.

Detailed exemplary embodiments are disclosed below. However, specific structural and functional details disclosed herein are merely for purposes of describing example embodiments.

It should be understood, however, that the invention is not limited to the particular exemplary embodiments disclosed, but covers all modifications, equivalents, and alternatives falling within the scope of the disclosure. Throughout the description of the figures, the same reference numerals denote the same elements.

Also, it should be understood that as used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items. Also it will be understood that when a component or unit is referred to as being “connected” or “coupled” to another component or unit, it can be directly connected or coupled to the other component or unit or intervening components or units may also be present. Also, other words used to describe the relationship between elements or elements should be interpreted in the same fashion (eg, "between" versus "directly between," "adjacent" versus "directly adjacent," etc.).

The purpose of the present invention is to design a construction method and management method for a distributed parallel computing system of a power system, aiming at the system management problem that the multi-control center decomposition and coordination computing mode is applied in practice, so as to provide a suitable interface for the distributed multi-application of the energy management system. into the platform.

Concrete technical scheme of the present invention is as follows.

1. Form the management framework of the distributed parallel computing system of the power system, the specific process is:

According to the multi-control center decomposition and coordination calculation mode, in order to ensure the correctness of the distributed parallel calculation results of the power system, when there are many subsystems involved in the calculation, a coordination layer is used to coordinate the calculation process of each subsystem. A distributed computing coordination layer is set up in the upper-level control center; each lower-level control center acts as each subsystem, each subsystem is connected to multiple users, and can initiate the control center to manage the subsystem; there is a communication link between the coordination layer and the subsystems. The construction method of the distributed parallel computing system of the power system is shown in Figure 1.

A coordination layer construction

In actual power system management, different users belonging to different subsystems may initiate different advanced application calculations for this subsystem at the same time. Therefore, the decomposition and coordination computing model should be extended to multi-user and multi-application fields, and the construction of the coordination layer needs to have the functions of coordination management and coordination computing.

(1) Coordination and management function

First of all, based on the requirements of system management for user rights, the coordination layer needs to have state management functions, including the management of subsystem states and user states.

The state of the subsystem is divided into join state and exit state: join means that the current subsystem normally participates in distributed computing; exit means that the current subsystem does not participate in distributed computing, which usually occurs when the current subsystem control center crashes.

From the perspective of initiating computing permissions, the user status has registration status and activation status: registration status means that the user has joined the computing system and can observe the current status of the subsystem, but has no right to initiate advanced application computing; activation status means that the user has the ability to initiate the current subsystem Permissions for system advanced application computing. From the mode of initiating advanced application calculations, the user status has real-time status and research status: users in the real-time status are system users and can initiate real-time status estimation calculations; users in the research status can observe the real-time calculation results of the current subsystem, but have no authority Initiating real-time state estimation calculations can only initiate research state advanced application calculations.

The coordination layer needs to record the current status of each subsystem and user. At the same time, when the user needs to change the current user status, it needs to send a request to the coordination layer, and the coordination layer finally determines the user status change.

Secondly, for the complex computing process of distributed computing, the coordination layer also needs to have computing management functions.

The method of realizing distributed computing in the decomposition and coordination computing mode is as follows: on the basis of maintaining the correctness of independent computing of each subsystem, boundary coordination is carried out between subsystems by exchanging boundary information, and each subsystem continuously corrects and iterates to complete the calculation process of the entire network. Therefore, the calculation management function of the coordination layer is as follows: when a user of a certain subsystem initiates calculation, the coordination layer controls other subsystems to start the calculation synchronously, and cooperates with the initiating calculation subsystem to complete a distributed calculation. The coordination layer performs unified coordination calculation for the distributed calculation initiated by each subsystem.

(2) Coordination calculation function

Specific to each high-level application calculation of the system, each subsystem is independent of each other, and the completion of distributed calculation needs to exchange boundary coordination information through the coordination layer. Therefore, the coordination calculation function of the coordination layer is as follows:

Before the calculation, the coordination layer reads the connection line information between the subsystems, and divides the subsystems accordingly; each subsystem only has the network model of the subsystem, and sends the network structure to the coordination layer after the internal network equivalent; the coordination layer After obtaining the equivalent model of the internal network of each subsystem, the equivalent model of the external network of each subsystem is established in combination with the connection line information, and sent to the corresponding subsystem; the subsystem accepts the external network model to form the grid calculation topology of the subsystem.

At the beginning of the calculation, each subsystem sends the boundary coordination information corresponding to this advanced application calculation to the coordination layer during each iteration; the coordination layer obtains the merge parameters according to the boundary coordination information, and sends the boundary node state quantity to the corresponding subsystem ; Each subsystem completes the iteration with the boundary information. Among them, the boundary coordination information exchanged by different application calculations is different, for example, the dynamic power flow is the diagonal elements of the inverse matrix of the boundary impedance matrix, and the state estimation is the diagonal elements of the inverse matrix of the information matrix, etc., depending on the specific advanced application calculations. Special note: The iteration process of each subsystem in advanced application computing is asynchronous iteration.

B subsystem construction

Corresponding to the function setting of the coordination layer, the subsystem must also have management functions and computing functions. The construction of the subsystem architecture is carried out from two levels: the subsystem and the user: the subsystem completes the management of the entire subsystem, and the user manages the user and completes specific advanced application calculations.

(1) Master-slave subsystem division

According to the description of the calculation management part in the management function of the coordination layer, any kind of advanced application calculation is initiated by a subsystem, and other subsystems start the calculation synchronously and cooperate to complete it. In a calculation, the initiator must be more concerned about the calculation results of the internal network of the power grid in the area under the jurisdiction of the control center, rather than the status and calculation results of the entire power grid. Therefore, the power grid model is divided according to the master-slave mechanism, as shown in Figure 2. The division method is as follows: the subsystem that initiates calculation is set as the master subsystem, and the subsystem that cooperates with the calculation is set as the slave subsystem; The network model is extrapolated from the boundary nodes of the actual physical network model, and a buffer network is set up to expand its equivalent model; when any subsystem initiates calculation as a master subsystem, the coordination layer divides the current subsystems according to the master-slave mechanism. The method of obtaining the buffer network from the border nodes and calculating the equivalent model of the internal network of the main subsystem can effectively ensure the correctness of the calculation results of the main subsystem in an advanced application calculation.

(2) Subsystem management function construction

Multi-computing process management. According to the decomposition and coordination computing mode, when a subsystem user in the computing system initiates computing, all subsystems start computing synchronously to complete distributed computing together. When the decomposition and coordination calculation is extended to multi-user and multi-application fields, different users belonging to multiple subsystems may initiate calculations at the same time, and the computing system may have multiple distributed computing conflicts. At the same time, in order to maintain the robustness of the computing system, the two distributed computing results initiated by the same user cannot affect each other. Therefore, it is necessary to design a parallel computing mechanism to complete distributed computing in the form of multi-process computing. The specific construction method is: the computing system uses the user as the minimum realization unit of specific computing; process, creating a shadow user to establish a communication relationship with the coordination layer, and establishing a one-to-one correspondence between users and shadow users to form a computing instance, as shown in Figure 3; the coordination layer issues instructions, and users belonging to the same computing instance and shadow users start computing synchronously , start the iterative calculation. Among them, the management function of the subsystem is mainly reflected in: accepting the command of the coordination layer to open up the computing process, establishing an independent computing process and generating shadow users.

(3) User structure construction

User management functions. There are multiple user states for users, and user states can be switched, and user states need to be managed. The specific management method is as follows: after the user starts, establish a communication link with the coordination layer; record the current user status; after accepting the user status switching request, forward the request to the coordination layer and wait for a response; accept and record the user status switching result.

User computing functions. As the smallest realization unit of distributed computing, the user completes the specific calculation process in the actual advanced application calculation. Both users and shadow users should have specific advanced application computing functions.

2. Object-based implementation of system management architecture functions, the specific process is:

Consider the practicality of the system management framework and implement it object-oriented.

A functional objectification

Based on the above-mentioned system management framework design, the functions of each part are abstracted: the coordination layer has coordination management functions and coordination calculation functions; subsystems have subsystem management functions; users have user management functions and user calculation functions. Therefore, five types of objects are designed to realize the functions of the above five modules: coordination management objects, coordination computing objects, subsystem management objects, user management objects and user computing objects. The specific functions of the five types of objects are shown in Figure 4.

Different objects have different functions, and their status and life cycle in the computing system are also different: the management object manages the system state and the computing process, and the duration is long; among them, the coordination management object occupies a core position and manages the entire computing system. The cycle runs through the entire computing system; the subsystem management object manages the computing process of the subsystem, the user management object manages the user status, and the two management objects cooperate with the coordination management object to complete system management. To complete a specific calculation process, the duration period is the same as that of a calculation process.

Arrange the coordination management objects and coordination calculation objects on the coordination layer, the subsystem management objects on the subsystem side, and the user management objects and user calculation objects on the user side. The five objects cooperate with each other to complete the management and calculation process of the system .

B three-dimensional coordinate labeling

From the data level, when the system runs according to the above parallel computing method, the data communication between the actual user and the shadow user of each computing instance must be completed through the coordination layer, so it is necessary to distinguish the data flows belonging to different computing instances. Since the design of the distributed parallel computing system management mechanism involves three dimensions of users, subsystems, and applications, the present invention utilizes time, space, and application three-dimensional coordinates to annotate different computing objects three-dimensionally to realize data flow belonging to different computing instances. layered isolation. Through the three-dimensional coordinate labeling method, each calculation object in the computing system will have a unique three-dimensional coordinates, and the calculation data will flow among the calculation objects in the same calculation instance to realize data layering.

C to establish a communication interface

Communication between the coordination layer and subsystems and users is required. The communication establishment method is: establish a communication connection between the coordination layer and each subsystem; establish a communication connection between the coordination layer and each user and the shadow user; the coordination layer has multiple The communication connection cannot be blocked and interfered, and the communication connection is a multi-threaded communication method; the distributed computing iteration process is an asynchronous iteration, and the communication established between the user and the coordination layer has an asynchronous communication function.

Therefore, the power system distributed parallel computing management method of the present invention includes steps:

A. When a user of a subsystem initiates calculation, the user sends a calculation request to the coordination management object through its user management object;

B. Coordinate management objects to issue start calculation instructions to other subsystems;

C. The subsystem management objects of the other subsystems open up new computing processes in their respective subsystems, and create shadow users in the processes;

D. The user management object of the shadow user establishes a connection with the coordination management object, and prepares for decomposition and coordination calculation;

E. The user management objects of the user and the shadow user establish a master user computing object and a secondary user computing object, and coordinate management objects to establish a coordination computing object;

F. The coordination management object assigns three-dimensional coordinates to all calculation objects, and all calculation objects locate and communicate with each other through three-dimensional coordinates;

G. Coordinate calculation objects to complete the decomposition and coordination between each calculation object, and each calculation object completes iterative calculation according to the decomposition coordination mechanism and the subsystem master-slave mechanism;

H. After the calculation is completed, the main user calculation object outputs the calculation result;

I. Any user in the computing system initiates any kind of advanced application computing, and completes a computing process according to the above steps.

As shown in Figure 6, the three-dimensional coordinates of the calculation object are time, space and application three-dimensional coordinates, wherein,

The time coordinate represents the users participating in the computation,

The spatial coordinates indicate the subsystem to which the calculation object belongs,

Application coordinates represent different advanced computing applications.

From the data level, when the system runs according to the above parallel computing method, the data communication between the actual user and the shadow user in each computing instance must be completed through the coordination layer, so it is necessary to distinguish the data flows belonging to different computing instances . Since the design of the management mechanism of the distributed parallel computing method involves three dimensions of users, subsystems, and applications, the present invention utilizes time, space, and three-dimensional coordinates of applications to three-dimensionally label different computing objects to realize data flow belonging to different computing instances. layered isolation.

Simply cite an example: Assuming that the regional power grid has three control centers, the power grid is divided into three subsystems according to the distributed computing method; for computing applications, the labels are: 1 for state estimation calculations, 2 for power flow calculations, and 3 for static security analysis calculations wait. Then, the system user 1 of subsystem 1 starts the real-time state estimation calculation in the real-time state, the actual three-dimensional coordinates of the master subsystem should be (1, +1, 1), and the corresponding shadow user coordinates of the slave subsystem are (1, -x, 1), the coordination layer calculation instance coordinates are (1, 0, 1), and x is the number of the subsystem to which the calculation object belongs.

Correspondingly, FIG. 7 is a schematic diagram of the data layering effect of the power system distributed parallel computing management method in the specific embodiment of the present invention. By means of three-dimensional coordinate labeling, each calculation object in the computing system will have a unique three-dimensional coordinate. Through Three-dimensional coordinate positioning, computing data flows between computing objects in the same computing instance, and data layering is realized.

In particular, the computing application is a variety of power system distributed applications that can be connected to the energy management system, including state estimation computing, power flow computing, and static security analysis.

In addition, the steps before step A further include:

A01. Set up a coordination layer for distributed computing in the superior control center, and the coordination layer has coordination management functions and coordination computing functions;

A02. Use the control centers at the lower level to act as various subsystems. Each subsystem is connected to multiple users. The subsystems have subsystem management functions, and the users have user management functions and user computing functions.

As shown in Figure 5, the management and calculation functions of the coordination layer, subsystems, and users are realized by objects, with coordination management objects and coordination calculation objects, the subsystems have subsystem management objects, and the users have user Admin Objects and User Computing Objects, where,

Coordination management objects are used to manage the entire computing system, including the management of subsystem status and user status, and management of all computing processes. It is at the core and lasts for the entire computing system;

The coordination calculation object is used to act as the communication agent of each subsystem, and complete the boundary coordination process in the specific calculation process, and the duration is the duration of one calculation;

The subsystem management object is used to manage the status of the subsystem, manage the creation of user processes, the establishment of computing objects, and the process of extinction, and the duration is the duration of the subsystem;

The user management object is used to manage the status of the user and establish the communication process between the user and the coordination layer, and the duration is the duration of the user;

The user calculation object is used to complete the actual calculation process, communicate with the coordination calculation object through the interface, and the duration is the duration of one calculation.

Among them, the buffer network is determined by using the decoupling linearization distribution factor of the power of the tie line or the voltage of the boundary node relative to the components of the external network by the boundary node or the boundary station outside the original tie line of the current subsystem, and included in the details of the internal network of the subsystem modeling.

Specifically, step G further includes:

G1. Before the calculation starts, the coordination layer reads the connection line information between the subsystems, and divides the subsystems according to the subsystem master-slave mechanism. Each subsystem only has the network model of the subsystem, and the network structure is equivalent to the intranet before sending to the coordination layer;

G2. After the coordination layer obtains the equivalent model of the internal network of each subsystem, it combines the connection line information to establish the equivalent model of the external network of each subsystem, and sends it to the corresponding subsystem;

G3. The subsystem accepts the external network model to form the grid computing topology of the subsystem;

G4. After the calculation starts, each subsystem sends the boundary coordination information corresponding to this advanced application calculation to the coordination layer in each iteration process. The coordination layer obtains the merged parameters according to the boundary coordination information, and sends the boundary node state quantity to Corresponding to the subsystems, each subsystem completes iterative calculations in combination with boundary information.

Among them, the boundary coordination information exchanged by different computing applications is different. The dynamic power flow calculation is the diagonal element of the inverse matrix of the boundary impedance matrix, and the state estimation calculation is the diagonal element of the inverse matrix of the information matrix.

The communication mode between the coordination layer and the subsystems and users is as follows:

Establish a communication connection between the coordination layer and each subsystem;

The coordination layer establishes a communication connection between each user and the shadow user;

No blocking interference between multiple communication connections in the coordination layer;

The communication connection between the user and the coordination layer is an asynchronous communication connection.

The method also includes a user management step, the user management step comprising:

After the user starts, establish a communication link with the coordination layer;

Record the current user status;

After receiving the user state switching request, forward the request to the coordination layer and wait for the response;

Accept and record user state switching results;

Create a calculation object after accepting the start calculation command.

The technical effects of the present invention will be described below through more specific calculation examples, and those skilled in the art know that the topology and data in the examples are only illustrative.

Taking the IEEE118 node system data as an example, the calculation example is analyzed. Through C++ programming, supplemented by CORBA technology to solve the communication process, access to distributed dynamic power flow and distributed state estimation to verify the parallel computing effect of the distributed parallel computing system.

The specific implementation method is: package the specific dynamic power flow calculation and state estimation calculation application program into a library for calling; and design the corresponding communication interface to complete the exchange process of boundary coordination information; based on the object logic relationship and cooperation mode in Figure 5, the The five types of objects in the distributed parallel computing system are realized by computer with C++ programming; they are connected to dynamic power flow and state estimation applications, and tested with IEEE118 node data.

Fig. 8 is a diagram of the deployment mode of the distributed parallel computing management system of the power system in the specific embodiment of the present invention. In the test, the computing system constructed under the distributed parallel computing system management mechanism is deployed as four PCs, one of which runs the coordination layer program to simulate the control center server acting as the coordination layer in the actual power system; the other three PCs simulate Three subsystem control centers. Each subsystem PC runs a subsystem server process as the subsystem management object; at the same time, each subsystem PC runs several user processes, which means that each subsystem has user access at this time, and can send to the coordination layer Calculate the request.

Fig. 9 is a schematic diagram of the master-slave mechanism division of subsystems in the power system distributed parallel computing management method in the specific embodiment of the present invention. The distributed parallel computing method is tested on the IEEE118 node model, and the data of the IEEE118 node model must be divided. For the subsystem master-slave mechanism, the division of subsystems is carried out according to the extrapolation rules of the master-subsystem connection line as shown in Figure 9. The specific connection lines are shown in the table in Figure 10.

For the distributed parallel computing system programmed and deployed in the above manner, the parallel computing effect is tested according to the scheme designed in the table in Fig. 11 . In order to have a better test for parallel computing, in the test, user 4 of subsystem 1 is in an uninterrupted cyclic computing state, deliberately creating a possible conflict with user 5 of subsystem 2 when the calculation is initiated. The test results are as follows.

The effect of multi-user parallel computing is shown in the table in Figure 12: Comparing the calculation results of user 4 and user 5, it can be seen that when each user performs calculations, all three subsystems participate in a calculation, and the calculation process of each user follows the multi-subsystem It is carried out in the decomposition and coordination calculation mode between users 4 and 5; the number of iterations of the calculation results of user 4 and user 5 is different, and the number of nodes included in the calculation results is different. The extrapolation of the connection line is carried out according to the master-slave mechanism, and the calculation results of the main subsystem users are correct , which proves the realization of the master-slave mechanism in the computing system; the calculations are carried out based on different models, and the results are all correct, and the calculations can be performed concurrently under the condition that calculation conflicts can be created, which proves that the distributed parallelism of the power system of the present invention Calculate the effectiveness of management methods.

The test results of multi-application parallel computing are shown in the table in Figure 13: user 4 of subsystem 1 starts state estimation calculation, and user 5 of subsystem 2 simultaneously calculates dynamic power flow. The state estimation calculation and the dynamic power flow calculation converge separately, and there is no waiting in line. It shows that the management mechanism of the distributed parallel computing system can realize the parallel computing of multiple applications without mutual influence, and the test results are similar to the multi-user test.

Therefore, the power system distributed parallel computing management method of the present invention extends the multi-control center decomposition and coordination computing mode to multi-user and multi-application fields for practice, and solves the parallel computing problem of distributed computing in future distributed energy management systems. It provides a suitable access platform for the distributed multi-application of the mature power system, and provides a solution for the new generation of three-dimensional coordinated energy management system.

It should be noted that the above-mentioned embodiments are only preferred embodiments of the present invention, and should not be understood as limiting the protection scope of the present invention. Any minor changes and modifications made to the present invention should not depart from the concept of the present invention. All belong to the distance protection scope of the present invention.

Claims (10)

1. A power system distributed parallel computing management method, the method comprising: A. When a user of a subsystem initiates an advanced application calculation, the user sends a calculation request to the coordination management object through its user management object; B. Coordinate management objects to issue start calculation instructions to other subsystems; C. The subsystem management objects of the other subsystems open up new computing processes in their respective subsystems, and create shadow users in the process, and cooperate with the computing users to perform distributed computing; D. The user management object of the shadow user establishes a connection with the coordination management object, and prepares for decomposition and coordination calculation; E. The user management objects of the user and the shadow user establish a master user computing object and a secondary user computing object, and coordinate management objects to establish a coordination computing object; F. The coordination management object assigns three-dimensional coordinates to all calculation objects, and all calculation objects locate and communicate with each other through three-dimensional coordinates; G. Coordinate calculation objects to complete the decomposition and coordination between each calculation object, and each calculation object completes iterative calculation according to the decomposition coordination mechanism and the subsystem master-slave mechanism; H. After the calculation is completed, the main user calculation object outputs the calculation result; I. Any user in the computing system initiates any kind of advanced application computing at any time, and completes a computing process according to the above steps A-Step H. 2. The distributed parallel computing management method of power system according to claim 1, characterized in that, each computing instance in the method includes computing objects formed by three-dimensional coordinate labeling to form data layers, and different sub-systems belonging to the same or different subsystems Users can start the same or different advanced application calculations at the same time without affecting each other, and the calculation processes of each user are completed in parallel. The three-dimensional coordinates are time, space and application three-dimensional coordinates, wherein, The time dimension coordinates represent the different users participating in the computation, The spatial dimension coordinates indicate the different subsystems to which the computing objects belong, Application dimension coordinates represent different high-level application calculations. 3. The distributed parallel computing management method of power system according to claim 1, characterized in that, the method extends the decomposition and coordination computing mode between multiple control centers to multi-user and multi-application fields, ensuring that the energy management system When multi-distributed applications are connected and multi-users are started, the correctness of the distributed calculations of each subsystem. The advanced application calculations can be connected to various power system distributed advanced applications of the energy management system, including state estimation calculations, power flow calculations and static security analysis. 4. The distributed parallel computing management method of a power system according to claim 1, characterized in that before the step A, it also includes: A01. Set up a coordination layer for distributed computing in the superior control center, and the coordination layer has coordination management functions and coordination calculation functions; A02. Use the control centers at the lower level to act as various subsystems. Each subsystem is connected to multiple users. The subsystems have subsystem management functions, and the users have user management functions and user computing functions. 5. according to the power system distributed parallel computing management method described in claim 4, it is characterized in that, the management function and computing function of described coordination layer, subsystem and user are realized by object, and coordination layer has coordination management object and a coordinating computing object, the subsystem has a subsystem management object, and the user has a user management object and a user computing object, wherein, Coordination management objects are used to manage the entire computing system, including the management of subsystem status and user status, and management of all computing processes. It is at the core and lasts for the entire computing system; The coordination calculation object is used to act as the communication agent of each subsystem, and complete the boundary coordination process in the specific calculation process, and the duration is the duration of one calculation; The subsystem management object is used to manage the status of the subsystem, manage the creation of user processes, the establishment of computing objects, and the process of extinction, and the duration is the duration of the subsystem; The user management object is used to manage the status of the user and establish the communication process between the user and the coordination layer, and the duration is the duration of the user; The user calculation object is used to complete the actual calculation process, communicate with the coordination calculation object through the interface, and the duration is the duration of one calculation. 6. The power system distributed parallel computing management method according to claim 1, characterized in that step G further comprises: G1. Before the calculation starts, the coordination layer reads the connection line information between the subsystems, and divides the subsystems according to the subsystem master-slave mechanism. Each subsystem only has the network model of the subsystem, and the network structure is equivalent to the intranet before sending to the coordination layer; G2. After the coordination layer obtains the equivalent model of the internal network of each subsystem, it combines the connection line information to establish the equivalent model of the external network of each subsystem, and sends it to the corresponding subsystem; G3. The subsystem accepts the external network model to form the computing topology of the subsystem; G4. After the calculation starts, each subsystem sends the boundary coordination information corresponding to the application calculation to the coordination layer during each iteration. Subsystems, each subsystem completes iterative calculations in combination with boundary information. 7. The distributed parallel computing management method of power system according to claim 6, characterized in that, the boundary node or the boundary plant station outside the original tie line of the current subsystem uses the power of the tie line or the voltage of the boundary node relative to the external network The decoupling linearization distribution factor of components determines the buffer network and takes into account the detailed modeling of the internal network of this subsystem. 8. The distributed parallel computing management method of power system according to claim 1, wherein the boundary coordination information exchanged by different computing applications is different, wherein the dynamic power flow calculation is the diagonal element of the inverse matrix of the boundary impedance matrix, and the state estimation calculation is the diagonal element of the inverse matrix of the information matrix. 9. The distributed parallel computing management method for power system according to claim 1, characterized in that, the communication mode between the coordination layer, subsystems and users is: Establish a communication connection between the coordination layer and each subsystem; The coordination layer establishes a communication connection between each user and the shadow user; No blocking interference between multiple communication connections in the coordination layer; The communication connection between the user and the coordination layer is an asynchronous communication connection. 10. The method for managing distributed parallel computing in a power system according to claim 1, wherein the method further includes a user management step, and the user management step includes: After the user starts, establish a communication link with the coordination layer; Record the current user status, which includes registration and activation, representing the permissions of the current user; After receiving the user state switching request, forward the request to the coordination layer and wait for the response; Accept and record user state switching results, Create a calculation object after accepting the start calculation command.