Remote control state evaluation method and system based on alony forest

Abstract

The invention discloses a remote control state evaluation method and system based on an alony forest, and the method is characterized in that the method comprises the following steps: collecting and numbering connection channels between transformer substation RTU equipment, and constructing a channel information vector according to the network topology between transformer substations and the state data of the RTU equipment; main channels in all the connecting channels are extracted according to the difference between the channel information vectors of all the connecting channels; issuing a remote control command to each transformer substation, and receiving remote control return information generated by the screened RTU equipment on the main channel; and extracting effective information from corresponding fields of the remote control return information, constructing a multi-dimensional input vector of each main channel, and inputting the multi-dimensional input vectors into a pre-trained random forest model to obtain alarms of a channel normal state and various channel abnormal states.

Description

Technical Field

[0001] The present invention relates to the field, and more specifically, to a remote control status evaluation method and system based on isolation forest. Background Art

[0002] The substation equipment centralized monitoring system is an intelligent monitoring technical support system for substation equipment, providing technical support for services such as operation monitoring, operation and control, monitoring assistant, business management, and remote intelligent inspection of equipment. The substation equipment centralized monitoring system is a specific application of the computer monitoring system in the power industry.

[0003] Pre - processing data acquisition is an important part of the substation equipment centralized monitoring system. Almost all on - site real - time data enters the substation equipment centralized monitoring system through the pre - processing data acquisition subsystem. Therefore, the scalability and availability of the pre - processing data acquisition subsystem are very important. In addition, the number of substations accessed and monitored by the substation equipment centralized monitoring system is relatively large, and the access capacity of the pre - processing acquisition subsystem becomes an important technical indicator of the substation equipment centralized monitoring system.

[0004] The pre - processing acquisition server is responsible for communicating with the RTUs accessed by all substations, collecting and processing the uploaded real - time data, and is also responsible for interacting with the RTUs for control commands.

[0005] In order to increase the stability and reliability of collecting substation real - time data, the substation equipment centralized monitoring system uses multiple (usually 4) communication channels to connect two RTUs of the substation simultaneously and collect the real - time data of the substation. The existing technology selects a main channel from multiple communication channels, and the other communication channels are in hot standby. The real - time data collected by the main channel is used for monitoring substation equipment, and the interaction of control commands with the RTU is also carried out on the main channel. Only when the main channel is abnormal and cannot communicate normally, a channel is selected from the hot standby channels as the main channel.

[0006] The remote control function is one of the core functions of the substation equipment centralized monitoring system. Whether the remote control status is abnormal is one of the important indicators for measuring the substation equipment centralized monitoring system.

[0007] In view of the above problems, there is an urgent need for a remote control status evaluation method and system based on isolation forest. Summary of the Invention

[0008] To solve the deficiencies in the prior art, the present invention provides a remote control status evaluation method and system based on isolation forest. Using the single communication channel between RTUs as samples, a random forest of multiple samples is constructed and solved to obtain alarm information on abnormal status of different types of channels.

[0009] The present invention adopts the following technical solutions.

[0010] In the first aspect of the present invention, a remote control status evaluation method based on the isolation forest is involved. The method includes the following steps: collecting and numbering the connection channels between the substation RTU devices, and constructing a channel information vector according to the network topology between substations and the status data of the RTU devices; extracting the main channels among all connection channels according to the differences between the channel information vectors of each connection channel; sending remote control commands to each substation, and receiving the remote control return information generated by the RTU devices on the selected main channels; extracting valid information from the corresponding fields of the remote control return information, and constructing a multi-dimensional input vector for each main channel, and inputting the multi-dimensional input vector into a pre-trained random forest model to obtain the alarms of the normal channel status and various abnormal channel statuses.

[0011] Preferably, collecting and numbering the connection channels between the substation RTU devices, and constructing a channel information vector according to the network topology between substations and the status data of the RTU devices includes: continuously collecting the RTU communication messages of each connection channel within a preset time period, and analyzing the differences between the valid fields in the RTU messages between the previous moment and the next moment; if the similarity of the differences between the valid fields is higher than the preset threshold, it is determined that the current connection channel is in the "not refreshed" state; if the RTU communication message of the current channel cannot be collected, it is determined that the current connection channel is in the "not put into use" state.

[0012] Preferably, collecting and numbering the connection channels between the substation RTU devices, and constructing a channel information vector according to the network topology between substations and the status data of the RTU devices includes: filtering the connection channels in the "not refreshed" state and the "not put into use" state, and marking the remaining connection channels as the "put into use" state; collecting the positions of each "put into use" connection channel in the network topology and the status data of the RTU devices corresponding to the connection channels to construct a channel information vector.

[0013] Preferably, sending remote control commands to each substation, and receiving the remote control return information generated by the RTU devices on the selected main channels includes: the remote control return information at least includes the following fields in the RTU message: remote control point number, substation ID, substation name, remote control type, remote control target value, RTU ID, current time stamp.

[0014] Preferably, extracting valid information from the corresponding fields of the remote control return information, and constructing a multi-dimensional input vector for each main channel, and inputting the multi-dimensional input vector into a pre-trained random forest model includes: in each node splitting process of the random forest model, a random sampling strategy without replacement is used to randomly extract the channel information vectors of some main channels from all main channels; based on the channel information vectors of the randomly extracted part of the main channels, node splitting is implemented.

[0015] Preferably, extract valid information from the corresponding fields of the remote control return information, and construct a multi-dimensional input vector for each main channel. Input the multi-dimensional input vector into a pre-trained random forest model, including: when the number of channels in the bottommost layer of nodes is less than a preset threshold, or when the growth of all isolated forest trees reaches a preset height, end the splitting process.

[0016] Preferably, obtain alarms for the normal state of the channel and various abnormal states of the channel, including: summarize the splitting results of each isolated forest tree to obtain the channel state type; the channel state type includes the normal state and various different abnormal states; for each abnormal state, send an abnormal alarm message, and the abnormal alarm message includes the remote control point number, substation, remote control type, remote control target value, remote control return time, and remote control abnormality.

[0017] In the second aspect of the present invention, it relates to a remote control state evaluation method system based on an isolated forest; the system is implemented by using a remote control state evaluation method in the first aspect of the present invention; the system includes a collection module, an extraction module, a screening module, and an alarm module; wherein, the collection module is used to collect and number the connection channels between substation RTU devices, and construct a channel information vector according to the network topology between substations and the status data of RTU devices; the extraction module is used to extract the main channels among all connection channels according to the differences between the channel information vectors of each connection channel; the screening module is used to send remote control commands to each substation and receive the remote control return information generated by the RTU devices on the selected main channels; the alarm module is used to extract valid information from the corresponding fields of the remote control return information, construct a multi-dimensional input vector for each main channel, and input the multi-dimensional input vector into a pre-trained random forest model to obtain alarms for the normal state of the channel and various abnormal states of the channel.

[0018] In the third aspect of the present invention, it relates to a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method in the first aspect of the present invention.

[0019] In the fourth aspect of the present invention, it relates to a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps of the method in the first aspect of the present invention.

[0020] The beneficial effects of the present invention are as follows. Compared with the prior art, a remote control status evaluation method and system based on the isolation forest in the present invention uses a single communication channel between RTUs as a sample to construct a multi-sample random forest solution, and obtains alarm information on abnormal statuses of different types of channels. The present invention screens out abnormal remote control operations from the remote control operation record dataset and performs alarm display, effectively improving the maintenance work efficiency and accuracy of the centralized control station, and providing an auxiliary decision for channel maintenance by diagnosing the communication quality of the main channel of the substation.

[0021] The beneficial effects of the present invention also include:

[0022] The present invention collects multi-channel real-time data, selects the main channel according to the statuses and priorities of multiple communication channels of the same substation, uses the selected main channel to interact with the substation for remote control commands, and saves the remote control operation information into the database. By screening out abnormal remote control operations from the remote control operation record dataset and performing alarm display, such as displaying the "remote control abnormal" identifier on the human-machine interface, the maintenance work efficiency and accuracy of the centralized control station are effectively improved, and an auxiliary decision is provided for channel maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a schematic flowchart of a remote control status evaluation method based on the isolation forest according to the present invention;

[0024] Figure 2 is a schematic diagram of a remote control status evaluation system based on the isolation forest according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] To make the objectives, technical solutions and advantages of the present invention clearer and more accurate, the technical solutions of the present invention will be described in detail below through multiple specific embodiments. The embodiments adopted by the present invention are only used to explain the present invention and do not limit the content of the present invention.

[0026] Figure 1 is a schematic flowchart of a remote control status evaluation method based on the isolation forest according to the present invention. As Figure 1 shown, in the first aspect of the present invention, it relates to a remote control status evaluation method based on the isolation forest, and the method includes Step 1 to Step 4.

[0027] Step 1: Collect and number the connection channels between the RTU devices in the substation, and construct a channel information vector based on the network topology between substations and the status data of the RTU devices. Collect and number the connection channels between the RTU devices in the substation, and construct a channel information vector based on the network topology between substations and the status data of the RTU devices, including: continuously collect the RTU communication messages of each connection channel within a preset time period, and analyze the differences between the valid fields in the RTU messages between the previous moment and the next moment; if the similarity of the differences between the valid fields is higher than the preset threshold, then determine that the current connection channel is in the "not refreshed" state;

[0028] If the RTU communication message of the current channel cannot be collected, then determine that the current connection channel is in the "not put into use" state. Filter the connection channels in the "not refreshed" state and the "not put into use" state, and mark the remaining connection channels as the "put into use" state; collect the positions of each "put into use" connection channel in the network topology and the status data of the RTU devices corresponding to the connection channels to construct a channel information vector.

[0029] In one embodiment, the valid field is a field that changes following the message content in the RTU message. Such as the data, CRC checksum, etc. in the RTU message, while the device address, function code, data format, etc. involved in the message header actually usually change according to the different data sources or according to the different message types. These contents are not within the scope of consideration of the valid fields mentioned in this step.

[0030] It is easy to think that in one embodiment, after the valid fields of multiple consecutive messages are extracted, the differences between the messages of each field can be obtained by field splitting first, and then the total difference can be obtained according to the weights of multiple valid fields; the difference of each field can be calculated by a similarity algorithm, and the weights of multiple valid fields can be obtained according to a preset algorithm. In one embodiment, such as the length of the valid field, can be an important basis for the size of the weight. The similarity algorithm can be calculated by common methods such as Euclidean distance, Pearson correlation coefficient, etc.;

[0031] In another embodiment, the fields of multiple consecutive messages can be obtained according to the type of information in the fields. For example, all valid fields of the floating-point number type are extracted and concatenated, and a difference is calculated. Then, the valid fields of the text type are extracted and concatenated, and another difference is calculated. Finally, the weight is calculated according to the influence degree of multiple differences on the message;

[0032] Step 2: Extract the main channels from all the connection channels according to the differences between the channel information vectors of each connection channel.

[0033] Queue the communication channels with the filtered status of "put into use" in descending order of channel priority, with the communication channels with higher priority at the front of the queue. Then, queue the communication channels with the filtered status of "not refreshed" in descending order of channel priority at the back of the queue. Finally, queue the filtered channels in descending order of channel priority at the back of the queue. Select the first channel in the queue as the main channel, and the other channels as backup channels. The priority can be obtained based on the channel voltage level, whether it is located at the topological center or edge, etc.

[0034] Step 3: Send remote control commands to each substation and receive the remote control return information generated by the RTU devices on the selected main channel.

[0035] Send remote control commands to each substation and receive the remote control return information generated by the RTU devices on the selected main channel, including: The remote control return information at least includes the following fields in the RTU message: remote control point number, substation ID, substation name, remote control type, remote control target value, RTU ID, current time stamp. Analyze the remote control return information, extract the remote control return time stamp, and calculate the remote control return time based on the current time stamp in the remote control sending information. Receive and analyze the remote control return information from the selected main channel, extract the remote control return time stamp, and calculate the remote control return time based on the current time stamp in the remote control sending information; the calculation method of the return time rt is: rt = remote control sending current time stamp - remote control return time stamp. Save the remote control return information and the remote control return time rt to the database.

[0036] Step 4: Extract the valid information from the corresponding fields of the remote control return information and construct the multi-dimensional input vectors of each main channel. Input the multi-dimensional input vectors into the pre-trained random forest model to obtain the alarms for the normal channel status and various abnormal channel statuses.

[0037] Extract the valid information from the corresponding fields of the remote control return information and construct the multi-dimensional input vectors of each main channel. Input the multi-dimensional input vectors into the pre-trained random forest model, including: In each node splitting process of the random forest model, a random sampling without replacement strategy is used to randomly extract the channel information vectors of some main channels from all the main channels; based on the channel information vectors of the randomly extracted part of the main channels, node splitting is implemented.

[0038] Randomly extract m samples from the database using the random sampling without replacement strategy and use them as the training samples for the isolation forest tree, that is, put the training samples into the root node of an isolation tree.

[0039] Select the return time eigenvalue from the training samples, and randomly select a threshold between the maximum and minimum values of the remote control return time eigenvalue to cut the training samples; place the samples in the training samples that are less than the threshold on the left branch of the current node, and place the samples in the training samples that are greater than or equal to the threshold on the right branch of the current node; repeat the above cutting steps on the data sets of the left branch node and the right branch node of the current node until the termination condition is reached.

[0040] There is only one training sample on the node, or all the training samples are the same, or the growth of the isolation forest tree reaches the set height log(m).

[0041] Calculate the score s, comprehensively calculate the results of each isolation forest tree for each sample x, and calculate the score through the following formula:

[0042]

[0043] Among them, h(x) is the path length passed by sample x during screening in the isolation forest tree, and c(m) is the average value of the path lengths of the isolation forest trees for the given number of samples m, which is used to standardize the path length h(x) of sample x.

[0044] The calculation formula of c(m) is as follows:

[0045]

[0046] Among them, H(i) is the harmonic number, which can be approximated as ln(i)+0.5772156649, where m is the number of samples. For a given data set size m, the expected value of the average path length is a constant. This formula provides a standardized benchmark for standardizing the path length.

[0047] E(h(x)) is the average path length of sample x, that is, the average value of the path lengths of sample x in all trees in the isolation forest.

[0048] Extract the valid information from the corresponding fields of the remote control return information, and construct the multi-dimensional input vectors of each main channel. Input the multi-dimensional input vectors into the pre-trained random forest model, including: when the number of channels in the bottom layer node is less than the preset threshold, or when the growth of all isolation forest trees reaches the preset height, the splitting process ends.

[0049] According to the calculated score, if the score is closer to 1, the higher the possibility that the corresponding sample is abnormal, and send a remote control abnormal alarm message; if the score is less than 0.5, the corresponding sample is a normal sample; if all scores are around 0.5, there are no obvious abnormal samples in the data set.

[0050] The alarm module receives and processes messages. The processing logic is to mark "remote control anomaly" on the basis of the regular processing logic. The processing results are written into the historical database, including the "remote control anomaly" identifier and the corresponding real-time alarm information generated, and are displayed on the real-time alarm window.

[0051] Obtain alarms for the normal state of the channel and various abnormal states of the channel, including: summarizing the splitting results of each lonely forest tree to obtain the channel status type; the channel status type includes the normal state and various different abnormal states; for each abnormal state, send out abnormal alarm information, and the abnormal alarm information includes the remote control point number, substation, remote control type, remote control target value, remote control return time, and remote control anomaly.

[0052] The channel status is a non-negative integer value, and each valid non-negative integer value corresponds to a unique channel status; the channel priority is a non-negative integer value, and the smaller the value, the higher the priority it represents. The current time stamp extracted from the sent remote control information is the current system time stamp, and the unit is: second.

[0053] In the second aspect of the present invention, it relates to a remote control status evaluation method system based on a lonely forest; it is implemented by using a remote control status evaluation method in the first aspect of the present invention; the system includes an acquisition module, an extraction module, a screening module, and an alarm module; wherein, the acquisition module is used to acquire and number the connection channels between substation RTU devices, and construct a channel information vector according to the network topology between substations and the status data of RTU devices; the extraction module is used to extract the main channels among all connection channels according to the differences between the channel information vectors of each connection channel; the screening module is used to send remote control commands to each substation and receive the remote control return information generated by the RTU devices on the selected main channels; the alarm module is used to extract valid information from the corresponding fields of the remote control return information, construct a multi-dimensional input vector for each main channel, and input the multi-dimensional input vector into a pre-trained random forest model to obtain alarms for the normal state of the channel and various abnormal states of the channel.

[0054] In the third aspect of the present invention, it relates to a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method in the first aspect of the present invention.

[0055] In the fourth aspect of the present invention, it relates to a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps of the method in the first aspect of the present invention.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that there are still contents in the technical solutions of the present invention that can modify the specific implementation manners of the present invention or make equivalent replacements. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A remote control state evaluation method based on lonely forest, characterized in that: The method comprises the following steps: Collect and number the connection channels between substation RTU devices, and construct channel information vectors based on the network topology between substations and the status data of RTU devices; Extracting the main channel among all the connection channels according to the difference between the channel information vectors of each connection channel; Send remote control commands to each substation and receive remote control return information generated by the RTU devices on the selected main channel; Valid information is extracted from the corresponding fields of the remote control return information, and a multi-dimensional input vector of each main channel is constructed. The multi-dimensional input vector is input into the pre-trained random forest model to obtain alarms for normal channel status and various channel abnormal status.

2. The remote control state evaluation method based on lonely forest according to claim 1, characterized in that: The collecting and numbering of connection channels between RTU devices in substations, and constructing channel information vectors according to the network topology between substations and the status data of RTU devices, include: Continuously collect RTU communication messages of each connection channel within a preset time period, and analyze the differences between the valid fields in the RTU messages between the previous moment and the next moment; If the similarity of the differences between the valid fields is higher than the preset threshold, the current connection channel is judged to be in the "not refresh" state; If the RTU communication message of the current channel cannot be collected, the current connection channel is judged to be in the "off" state.

3. The remote control state evaluation method based on lonely forest according to claim 2 is characterized by: The collecting and numbering of connection channels between RTU devices in substations, and constructing channel information vectors according to the network topology between substations and the status data of RTU devices, include: Filter the connection channels in the "not refreshed" and "not engaged" states, and mark the remaining connection channels as "engaged" states; The position of each connection channel in the "on" state in the network topology and the state data of the RTU device corresponding to the connection channel are collected to construct a channel information vector.

4. The remote control state evaluation method based on lonely forest according to claim 3 is characterized by: The sending of remote control commands to each substation and receiving remote control return information generated by the RTU device on the screened main channel includes: The remote control return information includes at least the following fields in the RTU message: remote control point number, plant station ID, plant station name, remote control type, remote control target value, RTUID, and current time stamp.

5. The remote control state evaluation method based on lonely forest according to claim 4 is characterized by: The extracting of valid information from the corresponding fields of the remote control return information, constructing a multi-dimensional input vector for each main channel, and inputting the multi-dimensional input vector into a pre-trained random forest model includes: The random forest model randomly extracts channel information vectors of some main channels from all main channels using a random sampling strategy without replacement during each node splitting process; Node splitting is implemented based on the channel information vectors of some randomly extracted main channels.

6. The remote control state evaluation method based on lonely forest according to claim 5 is characterized by: The extracting of valid information from the corresponding fields of the remote control return information, constructing a multi-dimensional input vector for each main channel, and inputting the multi-dimensional input vector into a pre-trained random forest model includes: When the number of channels in the lowest layer of nodes is less than the preset threshold, or when the growth of all lonely forest trees reaches the preset height, the splitting process ends.

7. The remote control state evaluation method based on lonely forest according to claim 6 is characterized by: The obtaining of the normal state of the channel and the alarm of the abnormal state of multiple channels includes: Summarize the splitting results of each lonely forest tree to obtain the channel state type; The channel status types include normal status and various abnormal statuses; For each abnormal state, an abnormal alarm message is issued, and the abnormal alarm message includes the remote control point number, plant station, remote control type, remote control target value, remote control return time, and remote control abnormality.

8. A remote control state evaluation method system based on lonely forest; characterized in that: Implemented by using a remote control state evaluation method based on lonely forest as described in any one of claims 1 to 7; The system includes a collection module, an extraction module, a screening module and an alarm module; wherein, The acquisition module is used to collect and number the connection channels between the RTU devices in the substation, and to construct a channel information vector according to the network topology between the substations and the status data of the RTU devices; The extraction module is used to extract the main channel among all the connection channels according to the difference between the channel information vectors of each connection channel; The screening module is used to send remote control commands to each substation and receive remote control return information generated by the RTU devices on the screened main channel; The alarm module is used to extract valid information from the corresponding fields of the remote control return information, construct a multi-dimensional input vector of each main channel, and input the multi-dimensional input vector into a pre-trained random forest model to obtain alarms of normal channel status and multiple channel abnormal status.

9. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images