Second-level response online complex event processing platform for power system safety assessment

Abstract

The invention discloses a second-level response online complex event processing platform for power system safety assessment, and relates to safety assessment and online analysis in a power system. According to the platform, efficient event processing and evaluation are achieved through a complex event processing CEP engine, a rule engine and a distributed memory data grid, and power grid online analysis of second-level response is supported. The CEP engine integrates multifunctional modules including a rule engine, a state machine and a memory data grid, and can analyze event modes in real time, manage system state changes and automatically execute and evaluate safe and stable operation rules. Through a D5000 data bus adapter, integration with a power grid regulation and control platform is realized, and real-time power grid state tracking and trend identification are supported. According to the method, the data processing speed and real-time performance are improved, manual intervention is reduced, the situation awareness capability of power grid operation is enhanced, and powerful auxiliary decision support is provided.

Description

Technical Field

[0001] The present invention relates to the field of safety assessment and online analysis in power systems, and in particular to an online complex event processing platform with a second-level response for power system safety assessment. Background Art

[0002] With the large-scale grid connection of high-proportion new energy power electronic equipment and the popularization of long-distance ultra-high voltage AC and DC transmission, the power structure has undergone major changes. At present, the power system has successively launched three batches of large-scale wind power and photovoltaic base projects, and plans to build a total installed capacity of about 455 million kilowatts of wind and solar bases by 2030. When the scale of new energy reaches tens of millions of kilowatts or even hundreds of millions of kilowatts, its output fluctuation will reach millions of kilowatts to tens of millions of kilowatts, and the fluctuation rate is fast, often completing an output fluctuation of about 10% of the installed capacity within a few minutes or even minutes. At the same time, the load and output fluctuations of the new power system are highly random, and the limited adjustable resources and poor controllability of new energy have led to a huge increase in the combination of power grid operation modes, frequent changes in modes, and complex and de-typical operation modes. At present, the daily fluctuation of new energy output in some provincial power grids has exceeded 30% of the maximum load. The existing online safety and stability analysis technology is insufficient in real-time safety evaluation of power grid operation under large-scale rapid fluctuations of new energy, formulation of operation plans considering uncertainty, and rapid rolling deduction of power grid preventive regulation, which poses severe challenges to the safe and stable operation of new power systems.

[0003] Grid dispatching is responsible for maintaining power balance and ensuring the safe and stable operation of large power grids. The safe and stable dispatching and operation of new power systems urgently need to solve the following problems: First, the "slow model". The operation mode of the power grid changes rapidly and becomes more complicated and de-typical. The construction of existing online analysis models for power grids involves many processes and long chains, resulting in a model update speed of minutes. It is impossible to provide the current power grid status for real-time, multi-period power grid stability analysis, reducing the real-time and accuracy of the analysis. Second, there are "many events". The event information received by the new power system dispatching has increased explosively. The current manual, periodic, and batch processing modes are difficult to meet the needs of rapid changes in the system and need to be processed in a massive amount of chaotic and complex data. Information can be used to quickly find meaningful events and make the next decision, but existing technologies make it difficult to identify complex events in power grid regulation and control and conduct event-driven situation assessment; third, the "methods are complicated", the number of stability regulations and the complexity of logic have increased dramatically, the stability regulations for power grid operation in a day have increased from dozens to hundreds or thousands, the operating rules change rapidly, and the existing operating rule monitoring models and methods cannot cope with the stability regulations with complex logic and rapid changes; "Planning is difficult", the formulation of power grid operation plans requires repeated switching between different platforms and telephone communication among multiple departments. There are problems of multiple levels and long chains. It takes tens of minutes or even hours to modify a plan, and it is impossible to quickly achieve a full range of safety deductions of the plan.

[0004] The D5000 smart grid dispatching and control system developed by State Grid integrates four major applications such as real-time monitoring and early warning based on an integrated platform. It has been promoted to the national provincial and above and many regional power grid dispatching organizations. However, the existing dispatching and control system has developed from the stage of weakly interconnected medium and small power grids. The dispatching business operation is basically carried out according to the internal balance mode of the provincial power grid. There are many levels, long chains, low business collaboration, and insufficient overall decision-making capabilities. The literature (Xu Hongqiang, Yao Jianguo, Yu Yijun, et al. Dispatching and control system architecture and key technologies supporting integrated large power grids [J]. Automation of Electric Power Systems, 2018, 42(6): 1-8.) proposed the construction ideas of a new generation of dispatching and control systems with "physical distribution and logical unification" and the design of the system functional framework. The literature (Guo Jiancheng, Nan Guilin, Xu Dan, et al. Connotation and key technologies of global monitoring of large power grids [J]. Automation of Electric Power Systems, 2018, 42(18): 1-7) points out that global monitoring of large power grids needs to have a wider monitoring range, more diverse monitoring content, more powerful real-time data processing capabilities, higher data real-time and consistency, coordinated control with a larger optimization space, and more friendly and practical auxiliary decision support. It can be seen that fast online analysis is one of the key supporting technologies for realizing a new generation of control systems, especially global monitoring of large power grids.

[0005] Based on the above analysis, the existing technologies have problems such as slow response speed, low data processing efficiency, much manual intervention and lack of real-time monitoring, which need to be solved urgently. Summary of the invention

[0006] In order to solve the problems of slow response speed, low data processing efficiency, frequent manual intervention and lack of real-time monitoring in the existing technology of security assessment and response processing of smart grid dispatching and control systems, the present invention provides an online complex event processing platform with a second-level response for power system security assessment.

[0007] The present invention optimizes the architecture of the power grid online analysis system by introducing complex event processing technology. (1) For the problem of slow response speed, the current power grid online analysis architecture adopts a periodic and batch processing computing mode with a response speed of minutes, which cannot meet the needs of fast (seconds) power grid online analysis (DSA). The present invention uses a self-developed complex event processing (CEP) engine to achieve real-time power grid stability assessment and provide second-level response capabilities. (2) For the problem of low data processing efficiency, the existing technology mainly relies on data storage and post-processing modes, which cannot achieve real-time and dynamic event processing and analysis. The present invention adopts an event-driven and real-time data processing mode, combined with memory data grid technology, to support the storage and memory computing of large-scale historical data, significantly improving data processing efficiency. (3) For the problem of much manual intervention, during the operation of the power grid, the dispatcher needs to remember and manually execute a large number of safe and stable operation procedures, which is labor-intensive and prone to errors. The present invention adopts automated rule evaluation, and through the rule engine function of the CEP engine, automatically executes and evaluates safe and stable operation rules, reduces manual intervention, and reduces the workload of dispatchers. (4) Lack of real-time monitoring and situational awareness: The existing system has insufficient real-time monitoring and situational awareness capabilities for the operation status of the power grid, and is unable to identify and respond to potential threats or opportunities in a timely manner. The present invention adopts a method to enhance real-time monitoring and situational awareness capabilities, that is, the CEP engine tracks the changes in the operation status of the power grid in real time, combines pattern matching rules, identifies complex event correlations, enhances the awareness of the operation status of the power grid, and timely warns and handles potential problems.

[0008] The present invention is implemented by the following technical scheme: an online complex event processing platform with a second-level response for power system security assessment, comprising the following processing steps: S1: Design the architecture of the complex event processing (CEP) engine. The CEP engine uses the rule engine to analyze event patterns in real time, the memory data grid to store and update large-scale historical data, and the state machine to manage system state changes, to achieve efficient event processing and evaluation. S1-1: Design complex event processing function module: Based on the existing power grid online analysis system, a CEP engine based on event-driven architecture is constructed; the CEP engine includes multiple event processing units, each of which runs independently and processes events through local event queues and single event processing loops. It consists of three basic functional modules: ① rule engine; ② memory; ③ state machine; While processing real-time events or data, the current state / historical data of the monitored physical system can be saved in the engine memory in the form of objects, and the information saved in the engine memory can be updated in real time through event processing; when one or a group of events trigger event processing, the rule engine performs rule evaluation; S1-2: Design the overall architecture of the complex event processing engine: The overall architecture of the CEP engine works as follows: multiple input event queues and event topics subscribed by the enterprise service bus ESB are bound to the local event queue of a processing unit for processing, and an input event topic is bound to multiple event processing units through broadcasting; the message preprocessor is bound to the event processing unit; the event is processed and filtered by the message preprocessor before being inserted into the local event queue; the message preprocessor is used to implement streaming data processing application scenarios; the CEP engine event processing results are sent out through the output event ESB queue / topic for alarm or driving other applications based on the event-driven architecture; the event processing results are saved to the database; A CEP engine can carry multiple event processing units, each of which is completely independent, with a high degree of parallel processing and strong scalability. Multiple CEP engines run in parallel to ensure data consistency in the distributed data network; object data related to event processing is stored in a cache based on a distributed data grid, and object data is loaded and saved from the cache to the CEP engine node memory during the entire processing process; according to application scenario requirements, object data is stored in the memory of the event processing unit computing node.

[0009] S2: Design core functional modules. The core part of the CEP engine is the rule engine, and the rest is considered to provide corresponding services for the engine rule processing. The most important task of the event processing loop of the event processing unit is to drive the rule engine and perform event processing through rule evaluation. The implementation architecture of the CEP software system is divided into two layers: ① CEP engine framework; ② CEP application system; the CEP engine framework has common functions independent of the CEP application layer, including: ① rule engine; ② distributed cache; ③ state machine; ④ common and reusable functions in streaming data processing; S2-1: CEP engine design: The CEP engine development work is completed on the basis of expanding open source software technologies, including: Vert.x, Hazelcast, EasyRules, MVEL, YAML; the implementation of the CEP engine runtime environment is based on the event-driven, non-blocking Vert.x open source software framework; The main idea of ​​the rule engine is to consider efficient parallel processing and high scalability. The rule engine UML class: VertxCEPEngineApp and BaseEventProcessingUnitVerticle are a one-to-many combination relationship: a CEP engine can carry multiple event processing units, manage and start a group of event processing units; the processing unit is based on the Vertic class of the Vert.x framework, and each processing unit is completely independent; the combination relationship of BaseEventProcessingUnitVerticle and the IESBAdapter interface is: the processing unit is inserted into the ESB adapter for external binding events such as subscribing / publishing messages; each processing unit has a local event queue, which is the only entry point for the processing unit; each processing unit has a loop thread that processes the local event queue; the combination relationship of VertxCEPEngineApp and ICacheAdapter is: the CEP engine uses the memory data grid to process and store a large number of events; S2-2: Rule engine design: The evaluation of rules is triggered by events, and its input includes two parts: ① Facts; ② RuleSet; the definition information of facts and rule sets is loaded from the rule description file in YAML format when the CEP engine is started; the rule content is defined based on an expression language (MVEL); the developed rule engine is implemented by extending the Easy Rules open source software rule engine framework, and the rule evaluation output is sent to downstream applications in the form of events.

[0010] S3: Intelligent matching and trend identification of power grid operation modes based on CEP technology, using a system based on reactive architecture and real-time processing, running a power grid analysis model in a distributed memory data grid, subscribing to real-time power grid change measurement information released by a distributed SCADA system through the D5000 data bus, and coordinating the real-time update of the model by the CEP engine to achieve real-time second-level tracking of power grid operation status changes; during the operation of the power grid, when the system changes or is disturbed, the CEP engine subscribes to changes and disturbance events from the D5000 data bus, and updates the power grid analysis model stored in the data grid in real time.

[0011] S3-1: Develop D5000 data bus adapter: The input and output of the CEP engine are presented in the form of events through the subscription and publication of messages on the data bus. In the online analysis application of the power grid, the CEP engine is integrated with the D5000 power grid control platform data bus. The developed CEP engine is based on the Vert.x operating environment and uses the Java programming language, while the D5000 data bus uses the C language. Therefore, a D5000 data bus adapter is developed to support message interaction between the C / C++ subsystem and the Java subsystem, realizing the integration of the CEP engine and the D5000 data bus. Bus message: Bus message is a general message type with byte as the message carrier. This message is transmitted from one node to another through the D5000 bus. Application message: Application message is a specific message type of application. Application messages can be sent and received by C / C++ application or Java application. Message adapter: The main function of the message adapter is to realize the conversion between general bus messages and specific application messages (Java, C / C++). For applications based on different programming languages, only one corresponding adapter needs to be maintained to realize interaction with all subsystems.

[0012] S3-2: Real-time analysis of operating status: Based on CEP engine driven implementation of fast search matching sequence diagram Figure 8 As shown in the figure, it is mainly divided into the following three steps: ① Historical simulation cases are arranged in the Hadoop data warehouse in advance, such as process 1, and the search indicators of historical cases are extracted, such as process 2, and the indicators are saved in the data grid, such as process 3; ② The CEP engine drives the reading of the current operating status, such as process 4, and generates the corresponding distributed power grid model in the data grid, such as process 5; ③ SCADA generates power change events of loads and generators, such as process 6, and the change events are sent to the subscribed CEP engine, such as process 7. The CEP engine updates the real-time model according to the event content, such as process 8. The CEP engine periodically triggers the implementation of search matching, such as process 9. The CEP engine sends a memory computing request to the memory data grid, such as process 10. The memory data grid calculates the search indicators of the current power grid model through the memory computing method, such as process 11. Finally, search and match to find the most similar historical computing case, such as process 12.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides an online complex event processing platform with a second-level response for power system security assessment: ① Efficient memory computing and caching mechanism: The CEP engine provides an efficient memory computing and caching mechanism, and selects the most appropriate caching mechanism through the cache service interface, thereby improving the efficiency of data processing and the response speed of the system. Multiple CEP engines can run in parallel based on the same distributed memory, while ensuring data consistency and processing efficiency, and improving the data processing speed and real-time performance of the system through a distributed architecture. ② Improved data processing speed and real-time performance: The CEP engine adopts an event-driven architecture to realize an online analysis system with a second-level response, which improves the efficiency of real-time dynamic security assessment of the power grid. ③ Reduced manual dependence and automated rule evaluation: The CEP engine quantifies the safe and stable operation regulations of the power grid, realizes automated rule evaluation, and reduces manual dependence on dispatchers. Dispatchers no longer need to remember and execute a large number of dispatching procedures, thereby reducing their mental workload and enabling them to focus on solving problems and providing decision support. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the CEP engine functional module proposed in the present invention.

[0015] Figure 2 This is the overall architecture of the CEP engine proposed in the present invention.

[0016] Figure 3 It is a diagram of the implementation framework of the CEP engine constructed by the present invention.

[0017] Figure 4 It is a UML class diagram of the rule engine framework design involved in the present invention.

[0018] Figure 5 It is the implementation technology roadmap of the rule engine involved in the present invention.

[0019] Figure 6 This is a diagram of the D5000 data bus adapter architecture involved in the present invention.

[0020] Figure 7 This is the implementation architecture of the fast online analysis system proposed by the present invention.

[0021] Figure 8 It is a fast search and matching sequence diagram based on CEP engine drive proposed by the present invention. DETAILED DESCRIPTION

[0022] The present invention is further described below in conjunction with specific embodiments.

[0023] An online complex event processing platform with a second-level response for power system security assessment includes the following processing steps: S1: Design the architecture of the complex event processing (CEP) engine. The CEP engine uses the rule engine to analyze event patterns in real time, the memory data grid to store and update large-scale historical data, and the state machine to manage system state changes, to achieve efficient event processing and evaluation. S1-1: Design complex event processing function module: The CEP engine based on event-driven architecture is built on the basis of the existing power grid online analysis system. The CEP engine includes multiple event processing units, each of which runs independently and processes events through local event queues and single event processing loops. The CEP engine consists of three basic functional modules: rule engine, memory data grid, and state machine. While processing real-time events, the current state / historical data of the monitored physical system can be saved in the engine memory and updated in real time through event processing. The rule engine performs rule evaluation when event processing is triggered. The basic principles of CEP engine event processing are as follows: Figure 1 shown.

[0024] S1-2: Design the overall architecture of the complex event processing engine: The overall architecture of the CEP engine is as follows Figure 2 As shown, the working process is as follows: multiple input event queues and event topics subscribed by the enterprise service bus ESB are bound to the local event queue of a processing unit for processing, and an input event topic is bound to multiple event processing units through broadcasting; the message preprocessor is bound to the event processing unit; the event is processed and filtered by the message preprocessor before being inserted into the local event queue; the message preprocessor is used to implement the streaming data processing application scenario; the CEP engine event processing results are sent out through the output event ESB queue / topic for alarm or driving other applications based on the event-driven architecture; the event processing results are saved to the database; A CEP engine can carry multiple event processing units, each of which is completely independent, with a high degree of parallel processing and strong scalability. Multiple CEP engines run in parallel to ensure data consistency in the distributed data network (Cache); object data related to event processing is stored in a cache based on a distributed data grid, and object data is loaded and saved from the cache to the CEP engine node memory during the entire processing process; according to application scenario requirements, object data is stored in the memory of the event processing unit computing node.

[0025] The event processing unit has a single event processing loop thread (Single Event Processing Loop) that processes the local event queue (Local Event Queue). When an event enters the local event queue, the event processing is single-threaded, and the order of event processing can be guaranteed; Multiple input event ESB queues / topics can be bound to a local event queue of a processing unit for processing, and an input event ESB topic can be bound to multiple processing units; A message preprocessor can be bound to an event processing unit. Before an event is inserted into the local event queue, a message preprocessor is used to process all incoming events. A message preprocessor can be used to implement streaming data application scenarios; Object data related to event processing is usually stored in a cache based on a distributed data network. Object data can be loaded and saved from the cache to the memory of the CEP engine node during the entire processing process. According to the application scenario requirements, object data is stored in the memory of the event processing unit computing node; The event processing unit can define multiple state machines. The state machine is used to manage the state transition of the object, including monitoring the state transition timeout of the object. The state machine has a timer to monitor the state transition timeout phenomenon. If a state transition timeout occurs, the state machine will generate a timeout warning event (TimeoutEvent); Event processing results are usually sent out through the output event ESB queue / topic, which can be used for alarms or to drive other event-driven applications. Event processing results can also be connected and saved in its database; The event processing unit has a repeating timer, which is used to schedule periodic repetitive tasks to be triggered in the form of events. The repeating timer of the event processing unit can trigger multiple repetitive tasks with different periods through the timer event (TimerEvent); Chain Rule function: The execution of a rule may trigger another rule. If this rule will be processed by the same event processing unit, the triggering event (RuleEvent) of this rule can be directly put into the local event processing queue to form a chain rule execution process.

[0026] S2: Design core functional modules. The core part of the CEP engine is the rule engine, and the rest is considered to provide corresponding services for engine rule processing. The most important task of the event processing loop of the event processing unit is to drive the rule engine and perform event processing through rule evaluation. The implementation architecture of the CEP software system is divided into two layers, such as Figure 3 As shown: ① CEP engine framework; ② CEP application system; the CEP engine framework has common functions independent of the CEP application layer, including: ① rule engine; ② distributed cache; ③ state machine; ④ common and reusable functions in streaming data processing; S2-1: CEP engine design: The CEP engine development work is completed on the basis of expanding open source software technologies, including: Vert.x, Hazelcast, EasyRules, MVEL, YAML; the implementation of the CEP engine runtime environment is based on the event-driven, non-blocking Vert.x open source software framework; The main idea of ​​the rule engine is to consider efficient parallel processing and high scalability. The UML class diagram of the rule engine is shown in Figure 4: VertxCEPEngineApp and BaseEventProcessingUnitVerticle are a one-to-many combination relationship: a CEP engine can carry multiple event processing units, manage and start a group of event processing units; the processing unit is based on the Vertic class of the Vert.x framework, and each processing unit is completely independent; the combination relationship of BaseEventProcessingUnitVerticle and the IESBAdapter interface is: the processing unit is inserted into the ESB adapter for external binding events such as subscribing / publishing messages; each processing unit has a local event queue, which is the only entry point of the processing unit; each processing unit has a loop thread that processes the local event queue; the combination relationship of VertxCEPEngineApp and ICacheAdapter is: the CEP engine uses the memory data grid to process and store a large number of events; S2-2: Rule engine design: The implementation technology route of the rule engine is as follows: Figure 5 As shown in the figure, rule evaluation is triggered by events, and its input includes two parts: ① Facts; ② RuleSet; the definition information of facts and rule sets is loaded from the rule description file in YAML format when the CEP engine is started; the rule content is defined based on an expression language (MVEL), which extends the Easy Rules open source software rule engine framework to realize rule evaluation and event triggering; the developed rule engine is implemented by extending the Easy Rules open source software rule engine framework, and the rule evaluation output is sent to downstream applications in the form of events.

[0027] S2-3: State machine design: The state machine of this embodiment has the following three basic functions: Error reporting: If the current state is A, if the system executes state transition ②, the state machine will issue an error alarm; State change listener (TransitionListener): Add a listener to any state transition (transition). Once the state changes, triggering the listener to execute can trigger the execution of a section of code; State transition timeout (Transition Timeout): If the system state is B, and no state transition (transition) occurs within a given time ②, the state machine can be set to issue a time-limited alarm.

[0028] S3: Intelligent matching and trend identification of power grid operation mode based on CEP technology. The implementation architecture of intelligent matching and trend identification of operation mode based on CEP is as follows: Figure 7 As shown in the figure, a typical system based on reactive architecture and real-time processing is adopted. The power grid analysis model running in the distributed memory data grid subscribes to the real-time change measurement information of the power grid released by the distributed SCADA system through the D5000 data bus. The CEP engine coordinates the real-time update of the model to achieve real-time second-level tracking of power grid operation status changes. During the operation of the power grid, when the system changes or is disturbed, the CEP engine subscribes to the change and disturbance events from the D5000 data bus and updates the power grid analysis model stored in the data grid in real time.

[0029] S3-1: Develop D5000 data bus adapter: The input and output of the CEP engine are presented in the form of events through the subscription and publication of data bus messages. In the online analysis application of power grid, the CEP engine is integrated with the D5000 power grid control platform data bus. The developed CEP engine is based on the Vert.x operating environment and uses the Java programming language, while the D5000 data bus uses the C language. Therefore, a D5000 data bus adapter is developed to support the message interaction between the C / C++ subsystem and the Java subsystem, and realize the integration of the CEP engine and the D5000 data bus. The architecture of the D5000 data bus adapter is as follows: Figure 6 shown.

[0030] Bus message: Bus message is a general message type with byte as the message carrier. This message is transmitted from one node to another through the D5000 bus. Application message: Application message is a specific message type of application. Application messages can be sent and received by C / C++ application or Java application. Message adapter: The main function of the message adapter is to realize the conversion between general bus messages and specific application messages (Java, C / C++). For applications based on different programming languages, only one corresponding adapter needs to be maintained to realize interaction with all subsystems.

[0031] S3-2: Real-time analysis of operating status: Based on CEP engine driven implementation of fast search matching sequence diagram Figure 8 As shown in the figure, it is mainly divided into the following three steps: ① Historical simulation cases are arranged in the Hadoop data warehouse in advance, such as process 1, and the search indicators of historical cases are extracted, such as process 2, and the indicators are saved in the data grid, such as process 3; ② The CEP engine drives the reading of the current operating status, such as process 4, and generates the corresponding distributed power grid model in the data grid, such as process 5; ③ SCADA generates power change events of loads and generators, such as process 6, and the change events are sent to the subscribed CEP engine, such as process 7. The CEP engine updates the real-time model according to the event content, such as process 8. The CEP engine periodically triggers the implementation of search matching, such as process 9. The CEP engine sends a memory computing request to the memory data grid, such as process 10. The memory data grid calculates the search indicators of the current power grid model through the memory computing method, such as process 11. Finally, search and match to find the most similar historical computing case, such as process 12.

[0032] Taking the online data model of the current national DSA system (directly exported from the D5000 system, about 40,000 nodes) as an example, this embodiment implements and tests six basic and simple CEP rules. On a laptop (Intel Core i7-68202.70 GHz), the rule execution performance test results are shown in Table 1. The execution efficiency of these six basic CEP rules is high, all in milliseconds. In actual applications, CEP rules will be more complex. Complex (compound) rules are implemented by combining simple basic rules, and their execution time will be longer, but it can be expected that it will still be in seconds.

[0033] Table 1

[0034] S3-3: Analysis process: The CEP engine subscribes to update events of the power grid digital twin analysis model and performs situation awareness analysis. If it senses that the power grid situation meets the CEP rules set therein, the CEP engine will start the relevant analysis function modules and perform online simulation analysis to provide auxiliary decision support for the power grid.

[0035] This embodiment takes the CEP engine-driven artificial intelligence online application as an example. The hardware environment for training and testing: IntelXeon CPU E3-1230 v5 @ 3.40 GHz / 16.00 GB RAM; software environment: Python 3.6.3 / TensorFlow 1.4. Based on the actual online data of a power grid from February to July 2017, the effectiveness of the model applied to transient stability is verified. The online data model includes all power grid equipment of 220kV and above; a total of 50,173 section data from February to July are collected to form a sample library; the critical clearing time (CCT) is selected as the target of GHNet training and prediction, and different faults need to train the model separately. 41,000 samples are randomly selected from the sample library as the training set, and the remaining 9,173 samples are used as the test set. Taking the CCT results of Fengxu Line as an example, the performance of the model on the test set is shown in Table 2. The average absolute error compared with the simulation results is small, and the overall performance of the model is good, but there are also cases where the error of individual samples is too large.

[0036] Table 2

[0037] A single model converges after 200 generations of training, and the training time is about 56.6 minutes. In the future, with the application of sample generation technology, the sample data will increase greatly, and the training time will also increase superlinearly. It is necessary to introduce parallel computing for model training.

[0038] The scope of protection claimed by the present invention is not limited to the above specific embodiments. For those skilled in the art, the present invention may have various deformations and changes. Any modifications, improvements and equivalent substitutions made within the concept and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An online complex event processing platform with a second-level response for power system security assessment, characterized by: The processing steps include: S1: Design the architecture of the complex event processing (CEP) engine. The CEP engine uses the rule engine to analyze event patterns in real time, the memory data grid to store and update large-scale historical data, and the state machine to manage system state changes, to achieve efficient event processing and evaluation. S2: Design core functional modules. The core part of the CEP engine is the rule engine, and the rest is considered to provide corresponding services for engine rule processing; The most important task of the event processing loop of the event processing unit is to drive the rule engine to perform event processing through rule evaluation; The implementation architecture of the CEP software system is divided into two layers: ① CEP engine framework; ②CEP application system; The CEP engine framework has common functions independent of the CEP application layer, including: ① rule engine; ② distributed cache; ③ state machine; ④ common and reusable functions for streaming data processing; S3: Intelligent matching and trend identification of power grid operation modes based on CEP technology, using a system based on reactive architecture and real-time processing, running a power grid analysis model in a distributed memory data grid, subscribing to real-time power grid change measurement information released by a distributed SCADA system through the D5000 data bus, and coordinating the real-time update of the model by the CEP engine to achieve real-time second-level tracking of power grid operation status changes; during the operation of the power grid, when the system changes or is disturbed, the CEP engine subscribes to changes and disturbance events from the D5000 data bus, and updates the power grid analysis model stored in the data grid in real time.

2. According to claim 1, a second-level response online complex event processing platform for power system security assessment is characterized by: Step S1 is specifically as follows: S1-1: Design complex event processing function module: Based on the existing power grid online analysis system, a CEP engine based on event-driven architecture is constructed; the CEP engine includes multiple event processing units, each of which runs independently and processes events through local event queues and single event processing loops. It consists of three basic functional modules: ① rule engine; ② memory; ③ state machine; While processing real-time events or data, the current state / historical data of the monitored physical system can be saved in the engine memory in the form of objects, and the information saved in the engine memory can be updated in real time through event processing; when one or a group of events trigger event processing, the rule engine performs rule evaluation; S1-2: Design the overall architecture of the complex event processing engine: The overall architecture of the CEP engine works as follows: multiple input event queues and event topics subscribed by the enterprise service bus ESB are bound to the local event queue of a processing unit for processing, and an input event topic is bound to multiple event processing units through broadcasting; the message preprocessor is bound to the event processing unit; the event is processed and filtered by the message preprocessor before being inserted into the local event queue; the message preprocessor is used to implement streaming data processing application scenarios; the CEP engine event processing results are sent out through the output event ESB queue / topic for alarm or driving other applications based on the event-driven architecture; the event processing results are saved to the database; Multiple CEP engines run in parallel to ensure data consistency in the distributed data network; object data related to event processing is stored in a cache based on a distributed data grid, and object data is loaded and saved from the cache to the CEP engine node memory during the entire processing process; according to application scenario requirements, object data is stored in the memory of the event processing unit computing node.

3. According to claim 1, a second-level response online complex event processing platform for power system security assessment is characterized by: Step S2 is specifically as follows: S2-1: CEP engine design: The CEP engine development work is completed on the basis of expanding open source software technologies, including: Vert.x, Hazelcast, EasyRules, MVEL, YAML; the implementation of the CEP engine runtime environment is based on the event-driven, non-blocking Vert.x open source software framework; Rule engine UML class: VertxCEPEngineApp and BaseEventProcessingUnitVerticle are in a one-to-many combination relationship: a CEP engine can carry multiple event processing units, manage and start a group of event processing units; the processing unit is based on the Vertic class of the Vert.x framework, and each processing unit is completely independent; the combination relationship of BaseEventProcessingUnitVerticle and IESBAdapter interface is: the processing unit is inserted into the ESB adapter for external binding events such as subscribing / publishing messages; each processing unit has a local event queue, which is the only entry point for the processing unit; each processing unit has a loop thread that processes the local event queue; the combination relationship of VertxCEPEngineApp and ICacheAdapter is: the CEP engine uses the memory data grid to process and store a large number of events; S2-2: Rule engine design: The evaluation of rules is triggered by events, and its input includes two parts: ① Facts; ② RuleSet; the definition information of facts and rule sets is loaded from the rule description file in YAML format when the CEP engine is started; the rule content is defined based on an expression language MVEL; the developed rule engine is implemented by extending the Easy Rules open source software rule engine framework, and the rule evaluation output is sent to downstream applications in the form of events.

4. According to claim 1, a second-level response online complex event processing platform for power system security assessment is characterized by: Step S3 is as follows: S3-1: Develop D5000 data bus adapter: The input and output of the CEP engine are presented in the form of events through the subscription and publication of messages on the data bus. In the online analysis application of the power grid, the CEP engine is integrated with the data bus of the D5000 power grid control platform. The developed CEP engine is based on the Vert.x runtime environment and uses the Java programming language, while the D5000 data bus uses the C language; Therefore, a D5000 data bus adapter is developed to support message interaction between the C / C++ subsystem and the Java subsystem, and to achieve integration of the CEP engine and the D5000 data bus; S3-2: Real-time analysis of operating status: The process of realizing fast search and matching sequence through CEP engine includes: 1) placing historical simulation cases in Hadoop data warehouse; 2) extracting search indexes of historical cases and saving them in data grid; 3) CEP engine reads the current operation status and generates corresponding distributed power grid model; 4) SCADA system generates load and generator power change events and sends them to CEP engine; 5) CEP engine updates the model in real time according to the event content and triggers the search and matching process periodically; 6) calculating the search index of the current power grid model through memory calculation method, and finally matching to find the closest historical calculation case.

Images