A distributed power distribution cabinet and its control system

Abstract

The present invention relates to a power distribution cabinet, in particular to a distributed power distribution cabinet and its control system; it includes a plurality of power distribution cabinets with information interaction function, and the plurality of power distribution cabinets, a power distribution master station system and a load balancer form a local area network or a ZigBee network, wherein the power distribution cabinets are all equipped with a microcomputer, a power collection device, a remote signal collection device, a temperature and humidity collection device and a switch outlet control device. The power distribution cabinets within the management scope are incorporated into the local area network or the ZigBee network, and the intensive control mode of the bus is changed to a control system similar to a distributed computer system, so that each power distribution cabinet has a relatively strong self-control capability, and can use the excess computing power of other power distribution cabinets in the network to serve itself, reducing the load of some low-performance power distribution cabinet microcomputers.

Description

Technical Field

[0001] The invention relates to a power distribution cabinet, in particular to a distributed power distribution cabinet and a control system thereof. Background Art

[0002] The so-called distributed system refers to the interconnection of multiple computers with the ability to communicate with each other, dividing the tasks and issuing them to each computer, so that each computer completes several tasks. The benefits are obvious. Since the task is faced by a computer network with huge computing power instead of a single computer, tasks that were originally difficult for computers to quickly achieve can now be completed through a distributed system.

[0003] As for distribution cabinets, due to the power grid's requirements for telemetry, telesignaling, and remote control, many distribution cabinets are actually equipped with functional components such as microcomputers and communicators (for example, Yao Guang et al. published "Design of Intelligent Ring Main Cabinet Based on Distributed Distribution Terminals" in "Communication Power Supply Technology" on March 25, 2019). They have actually met the hardware requirements of distributed distribution systems, but currently still use bus control, and there has been no report on actually transforming them into distributed systems. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a distributed power distribution cabinet and a control system thereof.

[0005] The technical solution of the present invention is:

[0006] A distributed power distribution cabinet, which includes multiple power distribution cabinets with information exchange function. Multiple power distribution cabinets, power distribution master station system and load balancer form a local area network or ZigBee network. In theory, it is also feasible to connect to the Internet, but the risk control cost is very high;

[0007] Among them, the distribution cabinets are equipped with microcomputers, power collection devices, remote signal collection devices, temperature and humidity collection devices and switch outlet control devices.

[0008] Among them, the power collection device, remote signal collection, and temperature and humidity collection device collect analog voltage, analog current, switch state, temperature, and humidity respectively. The switch state is converted into digital information through photoelectric isolation to obtain status signals, position change information (COS), and event sequence records (SOE); then the digital voltage / current, power, frequency, and power direction are calculated through a microcomputer; finally, it is necessary to send fault judgment calculation task requests and working condition calculation task requests to the load balancer. The load balancer cuts the task into subtasks and sends them to the idle distribution cabinets, and connects and binds the distribution cabinet that issues the task and the distribution cabinet that receives the subtask. The microcomputer of each distribution cabinet calculates its own subtask, and the results are uploaded to the distribution master station system by the distribution cabinet that issues the task. It is understandable that the hardware configurations of the distribution cabinets at each site may be different, and their computing power may be qualitatively different, and not every distribution cabinet is in a computing state. In this case, the distribution cabinets with computing power can be integrated in a manner similar to a distributed computer system, and distributed support can be provided for distribution cabinets that cannot perform larger computing tasks on their own, thereby improving the feedback rate of the entire system without increasing hardware costs.

[0009] The subtask allocation method of the load balancer is as follows:

[0010] According to the different processing capabilities of the microcomputers in different distribution cabinets, different weights are assigned to each microcomputer so that it can accept service requests with corresponding weights. For example, the weight of distribution cabinet 1 is designed to be 1, the weight of distribution cabinet 2 is 3, and the weight of distribution cabinet 3 is 6. Then these three distribution cabinets accept 10%, 30% and 60% of service requests respectively. The advantage of this approach is that it ensures that microcomputers with better computing power take on more computing tasks and avoids overloading microcomputers with lower performance.

[0011] A control system for a distributed power distribution cabinet, which uses the distributed power distribution cabinet to identify power distribution faults, includes the following steps:

[0012] 1) The power distribution master station system regularly collects fault information from each distribution cabinet and polls whether each distribution cabinet is online;

[0013] 2) If the distribution cabinet is online at this time, when the distribution cabinet receives the request to collect fault information, it sends a fault judgment calculation task request and a working condition calculation task request to the load balancer, and records this request and original data in the form of a log message in the load balancer database, and uploads the result to the distribution master station system after obtaining the calculation result; if the distribution cabinet is not online at this time, the load balancer is searched for the original data of the previous fault judgment calculation task requests and working condition calculation tasks of the distribution cabinet;

[0014] 3) Analyze the original data obtained in step 2 through the fault prediction model to obtain the fault type;

[0015] 4) Trip to cut off the corresponding fault area and deal with the faulty distribution cabinet according to the plan.

[0016] Specifically, the fault prediction model is:

[0017] Step S1: Classify and annotate log events based on log messages.

[0018] Step S1 can be implemented in the following ways:

[0019] Step S1.1: Use regular expressions to match variables in log messages, and use preset symbols to replace successfully matched variables;

[0020] Step S1.2: Log messages after variable replacement are grouped based on edit distance and event IDs are marked, and log events with different event IDs are different types of log events. Specifically, preset symbols can be selected as needed, such as *, &, etc. In the specific implementation process, log messages usually include two types of information, one is a free-format text string used to describe the semantic information of the event; the other is a variable or parameter in a system event, which is used to represent some important system attributes. This implementation uses regular expressions to match variables (such as numbers, IP addresses, directories, etc.) in log messages. After the match is successful, "*" is used for replacement. Then the log messages after variable replacement are grouped using edit distance. Edit distance refers to the minimum number of character editing operations (insertion, deletion, and replacement) required to convert one string into another between two strings. This implementation calculates the edit distance based on the words of the log message, not the characters. For log messages after variable replacement, if the edit distance between two log messages is less than a predetermined threshold, they are divided into the same group. For log messages in the same group, the same log event ID is used for marking. In order to improve efficiency, after classifying and marking log events based on log messages, the method

[0021] The step S2 is also included:

[0022] Filters out redundant log events.

[0023] Step S2 specifically includes: filtering redundant log events based on the threshold of the time interval between log events. Specifically, filtering is performed based on the threshold of the time interval between events, that is, when an event is reported, first find the event with the same log event ID that is closest to the event and has not been filtered (rather than the event with the same ID that is closest to it in the original log), and then calculate the time interval between them. If the time interval is less than the preset threshold, the event is deleted, otherwise the event is retained.

[0024] Step S3: using an improved hierarchical clustering algorithm to mine frequent event sequences related to various preset fault events from the classified and labeled log events.

[0025] Step S3 can be implemented in the following way:

[0026] Step S3.1: Based on the unified time window, generate a log event sequence set related to the preset fault event;

[0027] Step S3.2: Select the longest common subsequence between log event sequences as the similarity measure between sequences, and use the improved hierarchical clustering algorithm to mine frequent event sequences.

[0028] The beneficial effects of the present invention are: incorporating the distribution cabinets within the management scope into the local area network or ZigBee network, changing the intensive control mode of the bus to a control system similar to a distributed computer system, so that each distribution cabinet unit has a relatively strong self-management ability, and can use the excess computing power of other distribution cabinets in the network to serve itself, thereby reducing the load on some low-performance distribution cabinet microcomputers. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a structural schematic diagram of the present invention. DETAILED DESCRIPTION

[0030] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings:

[0031] like Figure 1 As shown, a distributed power distribution cabinet includes multiple power distribution cabinets with information interaction function. Multiple power distribution cabinets, power distribution master station system and load balancer form a local area network or ZigBee network. In theory, it is also feasible to connect to the Internet, but the risk control cost is very high.

[0032] Among them, the distribution cabinets are equipped with microcomputers, power collection devices, remote signal collection devices, temperature and humidity collection devices and switch outlet control devices.

[0033] Among them, the power collection device, remote signal collection, and temperature and humidity collection device collect analog voltage, analog current, switch state, temperature, and humidity respectively. The switch state is converted into digital information through photoelectric isolation to obtain status signals, position change information (COS), and event sequence records (SOE); then the digital voltage / current, power, frequency, and power direction are calculated through a microcomputer; finally, it is necessary to send fault judgment calculation task requests and working condition calculation task requests to the load balancer. The load balancer cuts the task into subtasks and sends them to the idle distribution cabinets, and connects and binds the distribution cabinet that issues the task and the distribution cabinet that receives the subtask. The microcomputer of each distribution cabinet calculates its own subtask, and the results are uploaded to the distribution master station system by the distribution cabinet that issues the task. It is understandable that the hardware configurations of the distribution cabinets at each site may be different, and their computing power may be qualitatively different, and not every distribution cabinet is in a computing state. In this case, the distribution cabinets with computing power can be integrated in a manner similar to a distributed computer system, and distributed support can be provided for distribution cabinets that cannot perform larger computing tasks on their own, thereby improving the feedback rate of the entire system without increasing hardware costs.

[0034] The subtask allocation method of the load balancer is as follows:

[0035] According to the different processing capabilities of the microcomputers in different distribution cabinets, different weights are assigned to each microcomputer so that it can accept service requests with corresponding weights. For example, the weight of distribution cabinet 1 is designed to be 1, the weight of distribution cabinet 2 is 3, and the weight of distribution cabinet 3 is 6. Then these three distribution cabinets accept 10%, 30% and 60% of service requests respectively. The advantage of this approach is that it ensures that microcomputers with better computing power take on more computing tasks and avoids overloading microcomputers with lower performance.

[0036] A control system for a distributed power distribution cabinet, which uses the distributed power distribution cabinet to identify power distribution faults, includes the following steps:

[0037] 1) The power distribution master station system regularly collects fault information from each distribution cabinet and polls whether each distribution cabinet is online;

[0038] 2) If the distribution cabinet is online at this time, when the distribution cabinet receives the request to collect fault information, it sends a fault judgment calculation task request and a working condition calculation task request to the load balancer, and records this request and original data in the form of a log message in the load balancer database, and uploads the result to the distribution master station system after obtaining the calculation result; if the distribution cabinet is not online at this time, the load balancer is searched for the original data of the previous fault judgment calculation task requests and working condition calculation tasks of the distribution cabinet;

[0039] 3) Analyze the original data obtained in step 2 through the fault prediction model to obtain the fault type;

[0040] 4) Trip to cut off the corresponding fault area and deal with the faulty distribution cabinet according to the plan.

[0041] Specifically, the fault prediction model is:

[0042] Step S1: Classify and annotate log events based on log messages.

[0043] Step S1 can be implemented in the following ways:

[0044] Step S1.1: Use regular expressions to match variables in log messages, and use preset symbols to replace successfully matched variables;

[0045] Step S1.2: Log messages after variable replacement are grouped based on edit distance and event IDs are marked, and log events with different event IDs are different types of log events. Specifically, preset symbols can be selected as needed, such as *, &, etc. In the specific implementation process, log messages usually include two types of information, one is a free-format text string used to describe the semantic information of the event; the other is a variable or parameter in a system event, which is used to represent some important system attributes. This implementation uses regular expressions to match variables (such as numbers, IP addresses, directories, etc.) in log messages. After the match is successful, "*" is used for replacement. Then the log messages after variable replacement are grouped using edit distance. Edit distance refers to the minimum number of character editing operations (insertion, deletion, and replacement) required to convert one string into another between two strings. This implementation calculates the edit distance based on the words of the log message, not the characters. For log messages after variable replacement, if the edit distance between two log messages is less than a predetermined threshold, they are divided into the same group. For log messages in the same group, the same log event ID is used for marking. In order to improve efficiency, after classifying and marking log events based on log messages, the method

[0046] The step S2 is also included:

[0047] Filters out redundant log events.

[0048] Step S2 specifically includes: filtering redundant log events based on the threshold of the time interval between log events. Specifically, filtering is performed based on the threshold of the time interval between events, that is, when an event is reported, first find the event with the same log event ID that is closest to the event and has not been filtered (rather than the event with the same ID that is closest to it in the original log), and then calculate the time interval between them. If the time interval is less than the preset threshold, the event is deleted, otherwise the event is retained.

[0049] Step S3: using an improved hierarchical clustering algorithm to mine frequent event sequences related to various preset fault events from the classified and labeled log events.

[0050] Step S3 can be implemented in the following way:

[0051] Step S3.1: Based on the unified time window, generate a log event sequence set related to the preset fault event;

[0052] Step S3.2: Select the longest common subsequence between log event sequences as the similarity measure between sequences, and use the improved hierarchical clustering algorithm to mine frequent event sequences.

[0053] The above embodiments and descriptions are only for illustrating the principles and best embodiments of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, all of which fall within the scope of the present invention to be protected.

Claims

1. A distributed power distribution cabinet, characterized in that: It includes multiple distribution cabinets with information interaction function, multiple distribution cabinets and distribution master station system and load balancer form a local area network or ZigBee network; Among them, the power distribution cabinets are equipped with microcomputers, power collection devices, remote signal collection devices, temperature and humidity collection devices and switch outlet control devices; The power collection device, remote signal collection device, and temperature and humidity collection device collect analog voltage, analog current, switch state, temperature, and humidity respectively. The switch state is converted into digital information through photoelectric isolation to obtain state signals, position change information (COS), and event sequence records (SOE); then the digital voltage / current, power, frequency, and power direction are calculated through a microcomputer; finally, it is necessary to send a fault judgment calculation task request and a working condition calculation task request to the load balancer. The load balancer cuts the task into subtasks and sends them to the idle distribution cabinets, and connects and binds the distribution cabinet that issues the task and the distribution cabinet that receives the subtask. The microcomputer of each distribution cabinet calculates its own subtask, and after obtaining the result, the distribution cabinet that issues the task uploads it to the distribution master station system; The subtask allocation method of the load balancer is: According to the different processing capabilities of the microcomputers in different distribution cabinets, different weights are assigned to each microcomputer so that it can accept service requests with corresponding weight numbers. For example, the weight of distribution cabinet 1 is designed to be 1, the weight of distribution cabinet 2 is 3, and the weight of distribution cabinet 3 is 6. Then these three distribution cabinets accept 10%, 30% and 60% of the service requests respectively.

2. A control system for a distributed power distribution cabinet, which uses the distributed power distribution cabinet to identify power distribution faults, characterized in that: The following steps are involved: 1) The power distribution master station system regularly collects fault information from each distribution cabinet and polls whether each distribution cabinet is online; 2) If the distribution cabinet is online at this time, when the distribution cabinet receives the request to collect fault information, it sends a fault judgment calculation task request and a working condition calculation task request to the load balancer, and records this request and original data in the form of log messages in the load balancer database. After obtaining the calculation results, the results are uploaded to the distribution master station system; If the power distribution cabinet is not online at this time, the original data of the previous fault judgment calculation task requests and working condition calculation tasks of the power distribution cabinet are searched from the load balancer; 3) Analyze the original data obtained in step 2 through the fault prediction model to obtain the fault type; 4) Trip to cut off the corresponding fault area and deal with the faulty distribution cabinet according to the plan.

3. A control system for a distributed power distribution cabinet according to claim 2, characterized in that: The fault prediction model is: Step S1: classify and annotate log events based on log messages; Step S1 can be implemented in the following ways: Step S1.1: Use regular expressions to match variables in log messages, and use preset symbols to replace successfully matched variables; Step S1.2: Group the log messages after variable replacement based on the edit distance and mark the event ID, and log events with different event IDs are different types of log events; use regular expressions to match the variables in the log messages, and after the match is successful, use "*" to replace; then use the edit distance to group the log messages after variable replacement; the edit distance refers to the minimum number of character editing operations required to convert one string into another; the edit distance is calculated based on the words of the log message, not the characters; For log messages after variable replacement, if the edit distance between two log messages is less than a predetermined threshold, they are classified into the same group; for log messages in the same group, the same log event ID is used for labeling; wherein, in order to improve efficiency, after classifying and labeling log events based on log messages, step S2 is further included: filtering redundant log events; Step S2 specifically includes: filtering redundant log events based on a threshold of the time interval between log events; specifically, filtering based on the threshold of the time interval between events, that is, when an event is reported, first find an event with the same log event ID that is closest to the event and has not been filtered, and then calculate the time interval between them. If the time interval is less than a preset threshold, delete the event, otherwise retain the event; Step S3: mining frequent event sequences related to various preset fault events from the classified and labeled log events through an improved hierarchical clustering algorithm; Step S3 can be implemented in the following way: Step S3.1: Generate a log event sequence set related to a preset fault event based on a unified time window; Step S3.2: Select the longest common subsequence between log event sequences as the similarity measure between sequences, and use the improved hierarchical clustering algorithm to mine frequent event sequences.

Images