Electric power operation risk early warning system

Abstract

The invention relates to the technical field of electric power operation safety, and discloses an electric power operation risk early warning system, which comprises a plan acquisition module used for inputting a work plan of pre-execution operation and obtaining a main line switch related to the work plan; the text processing module is used for performing word vector processing on the pre-execution work plan and the historical work implementation content to obtain a first word vector related to the historical work implementation content, a second word vector related to the pre-execution work implementation content and a third word vector related to the route and the section; the judgment module is used for analyzing and judging a corresponding high-risk point according to the second word vector; the matching module is used for matching corresponding prompt words, new energy user information and involved transformer area information in the violation library according to the high-risk points; and the communication module is used for sending the prompt words, the new energy user information and the transformer area information to a worker. According to the invention, high-efficiency management is realized under the condition that the burden of workers is not increased.

Description

Technical Field

[0001] The present invention relates to the technical field of power operation safety, and in particular to a power operation risk early warning system. Background Art

[0002] Power operation is the key foundation for ensuring the stable operation of the power system. The safety of power operation is related to the life safety of operators and the safety of national property. However, based on current observations, the number of serious violations investigated and punished by superiors remains high, and the pressure of investigating and punishing violations is relatively high. The main reasons are as follows:

[0003] 1. There are many safety regulations, and it is difficult for on-site engineers to learn them completely;

[0004] 2. Insufficient personnel quality makes it difficult to accurately identify potential danger points in on-site work;

[0005] 3. On-site staff are lazy and have a certain resistance to safety management. Summary of the invention

[0006] In order to better solve this problem, the present invention provides a power operation risk early warning system, which can reduce the rate of investigation and punishment of serious violations without increasing the workload.

[0007] The present invention is achieved through the following technical solutions:

[0008] An electric power operation risk early warning system, comprising:

[0009] A plan acquisition module is used to input the work plan of the pre-execution operation and obtain the main line switches involved in the work plan, wherein the work plan includes the lines and sections involved in the work and the implementation content of the pre-execution work;

[0010] A text processing module, used for performing word vector processing on the pre-execution work plan and the historical work implementation content, to obtain a first word vector about the historical work implementation content, a second word vector about the pre-execution work implementation content, and a third word vector about the route and section;

[0011] A judgment module, used to judge the corresponding high-risk points according to the second word vector analysis;

[0012] A matching module is used to match the corresponding prompt words in the violation database according to the high-risk points, and at the same time, match the new energy user information involved in the lines and sections involved in the work and the involved substation information through the third word vector combined with the information stored in the middle station database;

[0013] The communication module sends the prompt, new energy user information, and station area information to the staff.

[0014] As an optimization, along the data transmission direction, the text processing module includes a word segmentation unit and a word vector conversion unit, wherein the word segmentation unit performs word segmentation processing on the work plan based on a natural language processing algorithm to obtain a corresponding word segmentation result set; the word vector conversion unit is used to convert the word segmentation result set to obtain a corresponding number of second word vectors and third word vectors, or the word segmentation unit performs word segmentation processing on the historical work implementation content based on a natural language processing algorithm to obtain a corresponding word segmentation result set; the word vector conversion unit is used to convert the word segmentation result set to obtain a corresponding number of first word vectors.

[0015] As an optimization, along the data transmission direction, the word vector conversion unit includes a pre-trained BERT model and a similarity calculation unit, the BERT model converts the second word segmentation result set into a second initial word vector set composed of a plurality of second initial word vectors, and the BERT model converts the first word segmentation result set into a plurality of first word vectors, and the similarity calculation unit is used to calculate the similarity between the plurality of second initial word vectors and the first word vector, and classify the second initial word vector whose similarity with the first word vector is not less than the set similarity threshold as the second word vector, and classify the second initial word vector whose similarity with the first word vector is less than the set similarity threshold as the third word vector.

[0016] As an optimization, the judgment module includes a trained recurrent neural network RNN ​​model, and the RNN model training process is:

[0017] A1. Obtain historical work implementation content and historical high-risk data. The historical work implementation content is converted into a plurality of first word vectors through the text processing module. The plurality of first word vectors and the corresponding historical high-risk data together form a data set, and the data set is divided into a training set, a validation set, and a test set in proportion.

[0018] A2. Optimize the parameters of each layer of the RNN model based on the improved dung beetle algorithm and the training set to construct a judgment model, fine-tune the parameters of the judgment model through the verification set, and test the proxy model through the test set.

[0019] As an optimization, the judgment model takes the maximum similarity between the historical high-risk point data and the predicted high-risk point data as the objective function, wherein the historical high-risk point data is the result summarized from the execution of the historical work implementation content, and the predicted high-risk point data is the actual output result of the judgment model. A set of values ​​of the parameters of each layer of the RNN model is regarded as the location of a dung beetle. The specific process of optimizing the parameters of each layer of the RNN model based on the improved dung beetle algorithm and the training set is:

[0020] A2.1. Initialize the position of each dung beetle in the dung beetle population and the parameters of the dung beetle algorithm;

[0021] A2.2, sort the dung beetle population according to the size of fitness, so as to obtain the current best position and the current worst position;

[0022] A2.3. All the dung beetles are sorted according to their fitness, and the dung beetles are respectively subjected to rolling, dancing, breeding, foraging and stealing behaviors according to their sorted positions to update their positions;

[0023] A2.4, determine whether each updated dung beetle position exceeds the boundary;

[0024] A2.5. Re-rank and update the dung beetle population according to fitness;

[0025] A2.6. Determine whether the maximum number of iterations has been reached;

[0026] A2.7. If yes, output the currently updated dung beetle population to obtain the optimal parameters of each layer in the RNN model; otherwise, repeat A2.2 to A2.6.

[0027] As an optimization, the objective function is:

[0028] f=max∑(A·B) / (||A||||B||)

[0029] Among them, A is the word vector of historical high-risk point data, and B is the word vector of predicted high-risk point data;

[0030] The fitness is:

[0031] s=∑(A·B) / (||A||||B||).

[0032] As an optimization, the parameters of each layer of the RNN model include the dimension of the input features, the dimension of the hidden layer features, the number of RNN layers, the size of the bias, and the dropout probability.

[0033] As an optimization, in A2.3, the specific update formula for the theft behavior is:

[0034]

[0035] Where N is the number of dung beetles, represents the fitness ranking of the i-th dung beetle after the t-th update, X * is the current best position, that is, the current local optimal position, represents the position of the i-th dung beetle after the t-th update, represents the position of the i-th dung beetle after the t+1th update, g is a random vector of size 1*D that obeys the normal distribution, and Q is a constant.

[0036] As an optimization, the communication module includes one or more of wireless Bluetooth, wireless network and wired network.

[0037] As an optimization, the staff includes management departments, team leaders, work supervisors and safety supervisors.

[0038] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0039] Through the early warning system of the present invention, high-risk points are sent to staff members in a targeted manner before each mission is executed, so that the staff members can pay attention to them in a targeted manner, and there is no need to remember all high-risk points (including high-risk points that will not be involved in the current mission), thereby reducing the violation rate of the staff members.

[0040] Through the early warning system of the present invention, it is possible to proactively analyze the possible safety risk points and quality service complaint risks in on-site operations before the operation, and inform the person in charge of the work via text messages, thereby achieving efficient management of on-site safety without increasing the burden on the staff.

[0041] By training the recurrent neural network model with the improved dung beetle algorithm, it is possible to train a more accurate recurrent neural network model with fewer training sets, making subsequent risk warnings more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0043] Figure 1 This is a module connection diagram of a power operation risk early warning system of the present invention. DETAILED DESCRIPTION

[0044] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0045] This embodiment 1 provides a power operation risk early warning system, such as Figure 1 As shown, including:

[0046] 1. A plan collection module, which is used to input the work plan of the pre-execution operation and obtain the main line switches involved in the work plan. The work plan includes the lines and sections involved in the work and the implementation content of the pre-execution work;

[0047] The plan collection module mainly serves as a text input module. A large language model is set up in the module. The large language model is used to identify the lines and sections involved in the entered plan. After intelligent query, the main line switches that need to be opened are automatically given.

[0048] The early warning system is provided with a database, in which data mapping lines and main line switches is provided. Through this data, the corresponding main line switch can be directly obtained after knowing the line.

[0049] The platform will understand the work content of the operation plan, analyze the high-risk points (high-risk points) in this defect elimination, and automatically generate a list of on-site safety risk warnings after matching with the "Typical Violation Database". Then read the information of the middle-end database such as the homologous system and the acquisition system for analysis, and determine two risk points. One is that the operation scope involves new energy users and there is a risk of power supply back. The other is that the involved substation has had a power outage recently, and high-quality services need to be provided in advance. It mainly relies on the following text processing modules, judgment modules and matching modules.

[0050] Next, the working principles of these three modules are introduced in detail.

[0051] 2. A text processing module, configured to perform word vector processing on the pre-execution work plan and the historical work implementation content, to obtain a first word vector for the historical work implementation content, a second word vector for the pre-execution work implementation content, and a third word vector for the route and section;

[0052] The historical work implementation content mainly refers to the work implementation content that has been executed. The work implementation content mainly includes the actions to be performed, such as connecting to a new ring main unit, disconnecting, cable operations, etc.

[0053] In some embodiments, along the data transmission direction, the text processing module includes a word segmentation unit and a word vector conversion unit, wherein the word segmentation unit performs word segmentation processing on the work plan based on a natural language processing algorithm to obtain a corresponding word segmentation result set; the word vector conversion unit is used to convert the word segmentation result set to obtain a corresponding number of second word vectors and third word vectors, or the word segmentation unit performs word segmentation processing on the historical work implementation content based on a natural language processing algorithm to obtain a corresponding word segmentation result set; the word vector conversion unit is used to convert the word segmentation result set to obtain a corresponding number of first word vectors.

[0054] The main function of the text processing module is to perform word segmentation and word vector conversion on the text. After the word segmentation and word vector conversion of the historical work implementation content, several first word vectors are obtained, and the main function of the first word vector is to train the subsequent recurrent neural network model; after the word segmentation and word vector conversion of the work plan, several second word vectors and third word vectors are obtained, and the second word vector is set to have a similarity with the first word vector that is not less than the set threshold, and the third word vector is set to have a similarity with the first vector that is less than the set threshold. The main function of the second word vector is to input it into the trained recurrent neural network model for high-risk point prediction. The third word vector mainly corresponds to lines and sections, and its function is to match with the subsequent middle-end database to determine the new energy users and the areas involved in the operation line. In this way, there is no need to manually distinguish which are the work implementation content and which are the lines and sections in the work plan at the front end to avoid omissions or errors.

[0055] In some embodiments, along the data transmission direction, the word vector conversion unit includes a pre-trained BERT model and a similarity calculation unit, the BERT model converts the second word segmentation result set into a second initial word vector set composed of a plurality of second initial word vectors, and the BERT model converts the first word segmentation result set into a plurality of first word vectors, the similarity calculation unit is used to calculate the similarity between the plurality of second initial word vectors and the first word vector, and classify the second initial word vector whose similarity with the first word vector is not less than the set similarity threshold as the second word vector, and classify the second initial word vector whose similarity with the first word vector is less than the set similarity threshold as the third word vector.

[0056] The similarity calculation unit is mainly used for similarity calculation between word vectors. Therefore, the similarity calculation unit can be performed by the following similarity calculation method:

[0057] Cosine similarity, Euclidean distance, Manhattan distance, Minkowski distance, Chebyshev distance, Jaccard similarity coefficient, Pearson correlation coefficient, Spearman correlation coefficient, relative entropy.

[0058] 3. A judgment module, used to judge the corresponding high-risk points according to the second word vector analysis;

[0059] In some embodiments, the judgment module includes a trained recurrent neural network (RNN) model, and the second word vector is input into the RNN model to obtain the corresponding high-risk point.

[0060] The RNN model training process is:

[0061] A1. Obtain historical work implementation content and historical high-risk data. The historical work implementation content is converted into a plurality of first word vectors through the text processing module. The plurality of first word vectors and the corresponding historical high-risk data together form a data set, and the data set is divided into a training set, a validation set, and a test set in proportion.

[0062] A2. Optimize the parameters of each layer of the RNN model based on the improved dung beetle algorithm and the training set to construct a judgment model, fine-tune the parameters of the judgment model through the verification set, and test the proxy model through the test set.

[0063] In some embodiments, the parameters of each layer of the RNN model include the dimension of the input features, the dimension of the hidden layer features, the number of RNN layers, the size of the bias, and the dropout probability.

[0064] That is, the dimension of the input features, the dimension of the hidden layer features, the number of RNN layers, the size of the bias, and the dropout probability are used as the input of the judgment model, and the output of the judgment model is the word vector of the high-risk point.

[0065] In some embodiments, the judgment model takes the maximum similarity between the historical high-risk point data and the predicted high-risk point data as the objective function, that is, the objective function is:

[0066] f=maxΣ(A·B) / (||A||||B||)

[0067] Among them, A is the word vector of historical high-risk point data, and B is the word vector of predicted high-risk point data;

[0068] The historical high-risk point data is the result summarized from the execution of the historical work implementation content, the predicted high-risk point data is the actual output result of the judgment model, and a set of values ​​of the parameters of each layer of the RNN model is regarded as the location of a dung beetle. The specific process of optimizing the parameters of each layer of the RNN model based on the improved dung beetle algorithm and the training set is as follows:

[0069] A2.1. Initialize the position of each dung beetle in the dung beetle population and the parameters of the dung beetle algorithm;

[0070] A2.2, sort the dung beetle population according to the size of fitness, so as to obtain the current best position and the current worst position;

[0071] The fitness is:

[0072] s = ∑(A·B) / (||A||||B||);

[0073] A2.3. All the dung beetles are sorted according to their fitness, and the dung beetles are respectively subjected to rolling, dancing, breeding, foraging and stealing behaviors according to their sorted positions to update their positions;

[0074] In some embodiments, the specific update formula of the stealing behavior is:

[0075]

[0076] Where N is the number of dung beetles, represents the fitness ranking of the i-th dung beetle after the t-th update. The fitness ranking is in descending order. * is the current best position, that is, the current local optimal position, that is, in the current iteration, the corresponding objective function is the largest. represents the position of the i-th dung beetle after the t-th update, represents the position of the i-th dung beetle after the t+1th update, g is a random vector of size 1*D that obeys the normal distribution, and Q is a constant.

[0077] By improving the update formula of the original stealing behavior, the stealing dung beetle can converge faster.

[0078] A2.4, determine whether each updated dung beetle position exceeds the boundary;

[0079] A2.5. Re-rank and update the dung beetle population according to fitness;

[0080] A2.6. Determine whether the maximum number of iterations has been reached;

[0081] A2.7. If yes, output the currently updated dung beetle population to obtain the optimal parameters of each layer in the RNN model; otherwise, repeat A2.2 to A2.6.

[0082] 4. A matching module, which is used to match the corresponding prompt words in the violation database according to the high-risk points, and at the same time, match the new energy user information involved in the lines and sections of the work design and the involved substation information through the third word vector combined with the information stored in the middle station database;

[0083] A mapping table is stored in the middle station database, which stores the third vectors of lines and sections and the corresponding new energy users and substation information. Then, the new energy users and substation information are found through the third word vector.

[0084] 5. The communication module sends the prompt, new energy user information, and area information to the staff. Specifically, the staff includes the management department, team leader, work supervisor, and safety guardian.

[0085] In some embodiments, the communication module includes one or more of wireless Bluetooth, a wireless network, and a wired network.

[0086] After being integrated by the expert system, comprehensive risk warning text messages are sent to management departments, team leaders, work supervisors and safety supervisors.

[0087] In actual use, the early warning system of the present invention has reduced the number of violations to 16 times through 478 text message reminders, a year-on-year decrease of 63 percentage points. Moreover, through gradual promotion and application, the monthly violation rate of a power station has also decreased synchronously, and the comprehensive serious violation rate has decreased by nearly 50 percentage points.

[0088] A power supply station in a certain city plans to carry out work on the 24-3# pole of the Donglai line. The work supervisor xxx enters "10kV Donglai line 24-3# pole operation" in the "Intelligent Operation Risk Warning Platform". The "Intelligent Operation Risk Warning Platform" reads the PMS, collection, power supply, risk control, and marketing sg186 middle-end database information based on the work point entered by the work supervisor, and informs the work supervisor through text messages, voice, and other methods of the risk matters that need to be paid attention to in advance for the operation.

[0089] For example: When carrying out work on the # pole of the hole line, the person in charge, You Hongbing, please pay attention to the following matters:

[0090] Please disconnect the 10kV tunnel line Pingqiao #1 switch and the 10kV tunnel line Liushuitan #2 switch.

[0091] The work involves a photovoltaic user under the 10kV Donglai line Qiaotou public shed area: xxx (0003994xxx), contact number: 15298xxxxxx, please take safety measures to prevent reverse power supply.

[0092] The operation here involves the outage substations: the 10kV Donglai line Qiaotougongpengzi substation and the 10kV Donglai line Shanmenkou substation, involving 63 and 60 users respectively. The last outage time was 2023-05-18. Please remember to inform users in advance and provide quality service.

[0093] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A power operation risk early warning system, characterized in that: include: A plan acquisition module is used to input the work plan of the pre-execution operation and obtain the main line switches involved in the work plan, wherein the work plan includes the lines and sections involved in the work and the implementation content of the pre-execution work; A text processing module, used for performing word vector processing on the pre-execution work plan and the historical work implementation content, to obtain a first word vector for the historical work implementation content, a second word vector for the pre-execution work implementation content, and a third word vector for the route and section; A judgment module, used to judge the corresponding high-risk points according to the second word vector analysis; A matching module is used to match the corresponding prompt words in the violation database according to the high-risk points, and at the same time, match the new energy user information involved in the lines and sections involved in the work and the involved substation information through the third word vector combined with the information stored in the middle station database; The communication module sends the prompt, new energy user information, and station area information to the staff.

2. The power operation risk early warning system according to claim 1 is characterized in that: Along the data transmission direction, the text processing module includes a word segmentation unit and a word vector conversion unit, wherein the word segmentation unit performs word segmentation processing on the work plan based on a natural language processing algorithm to obtain a corresponding second word segmentation result set; the word vector conversion unit is used to convert the second word segmentation result set to obtain a corresponding number of second word vectors and third word vectors, or the word segmentation unit performs word segmentation processing on the historical work implementation content based on a natural language processing algorithm to obtain a corresponding first word segmentation result set; the word vector conversion unit is used to convert the first word segmentation result set to obtain a corresponding number of first word vectors.

3. The power operation risk early warning system according to claim 2 is characterized in that: Along the data transmission direction, the word vector conversion unit includes a pre-trained BERT model and a similarity calculation unit. The BERT model converts the second word segmentation result set into a second initial word vector set composed of a plurality of second initial word vectors, and the BERT model converts the first word segmentation result set into a plurality of first word vectors. The similarity calculation unit is used to perform similarity calculation on the plurality of second initial word vectors and the first word vector, and classify the second initial word vector whose similarity with the first word vector is not less than the set similarity threshold as the second word vector, and classify the second initial word vector whose similarity with the first word vector is less than the set similarity threshold as the third word vector.

4. The power operation risk early warning system according to claim 1 is characterized in that: The judgment module includes a trained recurrent neural network RNN ​​model, and the RNN model training process is: A1. Obtain historical work implementation content and historical high-risk data. The historical work implementation content is converted into a plurality of first word vectors through the text processing module. The plurality of first word vectors and the corresponding historical high-risk data together form a data set, and the data set is divided into a training set, a validation set, and a test set in proportion. A2. Optimize the parameters of each layer of the RNN model based on the improved dung beetle algorithm and the training set to construct a judgment model, fine-tune the parameters of the judgment model through the verification set, and test the proxy model through the test set.

5. The power operation risk early warning system according to claim 4 is characterized in that: The judgment model takes the maximum similarity between the historical high-risk point data and the predicted high-risk point data as the objective function, wherein the historical high-risk point data is the result summarized from the execution of the historical work implementation content, and the predicted high-risk point data is the actual output result of the judgment model. A set of values ​​of the parameters of each layer of the RNN model is regarded as the location of a dung beetle. The specific process of optimizing the parameters of each layer of the RNN model based on the improved dung beetle algorithm and the training set is as follows: A2.

1. Initialize the position of each dung beetle in the dung beetle population and the parameters of the dung beetle algorithm; A2.2, sort the dung beetle population according to the size of fitness, so as to obtain the current best position and the current worst position; A2.

3. All the dung beetles are sorted according to their fitness, and the dung beetles are respectively subjected to rolling, dancing, breeding, foraging and stealing behaviors according to their sorted positions to update their positions; A2.4, determine whether each updated dung beetle position exceeds the boundary; A2.

5. Re-rank and update the dung beetle population according to fitness; A2.

6. Determine whether the maximum number of iterations has been reached; A2.

7. If yes, output the currently updated dung beetle population to obtain the optimal parameters of each layer in the RNN model; otherwise, repeat A2.2 to A2.

6.

6. The power operation risk early warning system according to claim 5, characterized in that: The objective function is: f=maxΣ(A·B) / (||A||||B||) Among them, A is the word vector of historical high-risk point data, and B is the word vector of predicted high-risk point data; The fitness is: s=∑(A·B) / (||A||||B||).

7. An electric power operation risk warning system according to claim 4, 5 or 6, characterized in that: The parameters of each layer of the RNN model include the dimension of the input features, the dimension of the hidden layer features, the number of RNN layers, the size of the bias, and the dropout probability.

8. The power operation risk early warning system according to claim 5, characterized in that: In A2.3, the specific update formula for the theft behavior is: Where N is the number of dung beetles, represents the fitness ranking of the i-th dung beetle after the t-th update, X * is the current best position, that is, the current local optimal position, represents the position of the i-th dung beetle after the t-th update, represents the position of the i-th dung beetle after the t+1th update, g is a random vector of size 1*D that obeys the normal distribution, and Q is a constant.

9. The power operation risk early warning system according to claim 1, characterized in that: The communication module includes one or more of wireless Bluetooth, wireless network and wired network.

10. The power operation risk early warning system according to claim 1, characterized in that: The staff members include management departments, team leaders, work supervisors and safety supervisors.

Images