A power dispatch risk early warning method and system integrating power data

By extracting emotional information from social application chat records in power scheduling, using the Naive Bayes algorithm and emotional dictionary to build a data pipeline, the problem of insufficient user emotional information in power scheduling is solved, and an accurate warning of power scheduling risks is achieved and optimization strategies improve the responsiveness of the power system.

CN118396371BActive Publication Date: 2024-11-19GUONENG (GUANGDONG) INTEGRATED ENERGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410485459.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-22
Publication Date
2024-11-19
Estimated Expiration
2044-04-22

AI Technical Summary

Technical Problem

The lack of user emotional information in power scheduling in the prior art leads to insufficient accuracy of power scheduling risk warning and it is difficult to perceive peaks and fluctuations in time.

Method used

By extracting chat records in social applications, using naive Bayes algorithm and emotional dictionary, a data pipeline is constructed, the standard emotional value of short sentences is calculated, and the fluctuation curve is drawn, the difference between power data and real-time data is judged, and a Alarm message.

Benefits of technology

It broadens the data source, improves the efficiency of chat record processing, accurately analyzes the relationship between emotional values ​​and power demand, realizes early warning and optimization strategies for power scheduling risks, and improves the flexibility and response capabilities of the power system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118396371B_ABST
    Figure CN118396371B_ABST
Patent Text Reader

Abstract

The present invention is applicable to the technical field of risk warning, and in particular, to a method and system for early warning of power dispatching risks that integrate power data, the method comprising: extracting chat records in social applications, retrieving sensitive words in the chat records, integrating short sentences containing sensitive words, and creating a data set; creating a pipeline, inserting a classifier trained using a naive Bayes algorithm into the pipeline, and hanging at least one filter under the classifier. By constructing a fluctuation curve, the present invention can accurately understand the relationship between sentiment value and power demand, thereby achieving early warning of power dispatching risks, and can also optimize power dispatching strategies, improve the flexibility and responsiveness of the power system, and can more intuitively display power dispatching risks by calculating differences and generating alarm information, thereby improving the efficiency of risk warning.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of risk warning, and in particular to a method and system for early warning of electric power dispatching risks integrating electric power data. Background Art

[0002] my country's power grid lines are complex. In order to ensure the safe and stable operation of the power grid, it is necessary to collect power dispatching information through various channels, combine the power grid data parameters of each time period, and coordinate the power grid transmission tasks at each stage; power dispatching requires optimizing the allocation and adjustment of various power generation resources in the power system to ensure the stable operation of the power system. However, due to the complexity of the AC and DC lines in the power grid, the large number of power equipment, the constant changes in power grid information, and the frequent natural disasters, the factors affecting power grid dispatching continue to increase, and the difficulty of power dispatching is also gradually increasing. This requires the continued improvement and development of power dispatching technology, and the continuous introduction of new technologies and methods to adapt to the ever-changing power dispatching situation.

[0003] In real life, the peaks and fluctuations of future electricity demand are generally predicted by monitoring and analyzing real-time load data in combination with historical data and historical events. However, there are few sources of such data and a lack of user emotional information, which can easily lead to untimely perception of events and affect the accuracy of power dispatch risk warnings. Therefore, "how to use social networks to provide decision support for power dispatch risk warnings" is a technical problem that the present invention needs to solve. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for early warning of power dispatching risks that integrate power data, so as to solve the problem of "how to use social networks to provide decision support for early warning of power dispatching risks" raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A power dispatch risk early warning method integrating power data, the method comprising:

[0007] Extract chat records from social applications, retrieve sensitive words from the chat records, integrate short sentences containing sensitive words, and create a data set;

[0008] Building a pipeline, inserting a classifier trained by the Naive Bayes algorithm into the pipeline, and attaching at least one filter to the classifier, integrating the pipeline, the classifier and the filter, forming a data pipeline, inputting the data set into the data pipeline, and outputting the total score of the short sentence;

[0009] Define target power data, determine the corresponding relationship between the total score and the target power data, and draw a fluctuation curve with the total score as the abscissa and the actually collected target power data as the ordinate;

[0010] The acquired real-time total score is substituted into the fluctuation curve, the target power data is located, and the difference between the target power data and the real-time power data is determined. If the difference is greater than a preset threshold, an alarm message is generated using the difference, and the alarm message is sent to a preset terminal.

[0011] Furthermore, the steps of extracting chat records from social applications, retrieving sensitive words from the chat records, integrating short sentences containing sensitive words, and creating a data set include:

[0012] Export chat records from social applications and use the preset annotation tools to tag sensitive words;

[0013] The short sentences are classified into different data sets according to the parts of speech, and data set attributes are defined, where the attributes include: weather, temperature and humidity, and travel.

[0014] Furthermore, the steps of creating a pipeline, inserting a classifier trained using the Naive Bayes algorithm into the pipeline, and attaching at least one filter to the classifier include:

[0015] The classifier is trained using the acquired sentiment dictionary, and the short sentences are assigned values ​​according to the training results, and the standard sentiment value of the short sentences is calculated based on a preset formula;

[0016] Delimiting the score segments of the standard sentiment value, and superimposing the score segments into the filter;

[0017] Each score segment corresponds to a filter.

[0018] Furthermore, the step of calculating the standard sentiment value of the short sentence includes:

[0019] Based on the sentiment dictionary, determine the sentiment base value of the sensitive word and record the number of occurrences of the sensitive word;

[0020] The weighted average of the sentiment values ​​is calculated using the preset sentiment value formula;

[0021] The sentiment value formula is:

[0022]

[0023] where v i is the sentiment base value of the i-th sensitive word, c iis the number of occurrences of the i-th sensitive word, and V is the weighted average;

[0024] Substitute the calculated weighted average into the standardization formula to calculate the standard sentiment value;

[0025] The standardized formula is:

[0026]

[0027] Where V′ is the standard sentiment value, V is the weighted average, A is -1, B is 1, a is the minimum value among the sentiment base values, and b is the maximum value among the sentiment base values.

[0028] Furthermore, the steps of integrating the pipeline, the classifier and the filter, forming a data pipeline, inputting the data set into the data pipeline, and outputting the total score of the sensitive words include:

[0029] Using the data set as a source, connecting the classifier and the filter using the pipeline to form a data pipeline;

[0030] Input the data set into the data pipeline, determine whether the standard sentiment value of the short sentence satisfies the score segment of the filter, and if so, classify the short sentence into the filter corresponding to the score segment;

[0031] The standard sentiment values ​​of all short sentences in the filter are superimposed and the superimposed result is determined as the total score.

[0032] Furthermore, the steps of defining the target power data, determining the corresponding relationship between the total score and the target power data, and drawing the fluctuation curve with the total score as the horizontal coordinate and the target power data as the vertical coordinate include:

[0033] Based on the corresponding relationship, a fluctuation curve is drawn, and a label consisting of the attributes is inserted into the fluctuation curve;

[0034] A turning point in the fluctuation curve is selected, and the turning point is used as a trigger head to embed a trigger mechanism.

[0035] Further, the step of substituting the acquired real-time total score into the fluctuation curve, locating the target power data, determining the difference between the target power data and the real-time power data, and if the difference is greater than a preset threshold, generating an alarm message using the difference, and sending the alarm message to a preset terminal includes:

[0036] Using the real-time total score, traverse corresponding points in the fluctuation curve, and define the ordinates of the corresponding points as target power data;

[0037] The difference is divided into at least two difference segments, an alarm level is determined based on a preset alarm strategy, and the difference is integrated to generate alarm information.

[0038] Furthermore, the system comprises:

[0039] A retrieval module is used to extract chat records in social applications, retrieve sensitive words in the chat records, integrate short sentences containing sensitive words, and create a data set;

[0040] An output module is used to create a pipeline, insert a classifier trained by the naive Bayes algorithm into the pipeline, and hang at least one filter under the classifier, integrate the pipeline, classifier and filter, form a data pipeline, input the data set into the data pipeline, and output the total score of all short sentences;

[0041] A drawing module, used to define target power data, determine the corresponding relationship between the total score and the target power data, and draw a fluctuation curve with the total score as the horizontal coordinate and the actually collected target power data as the vertical coordinate;

[0042] The alarm module is used to substitute the acquired real-time total score into the fluctuation curve, locate the target power data, determine the difference between the target power data and the real-time power data, and if the difference is greater than a preset threshold, generate an alarm message using the difference, and send the alarm message to a preset terminal.

[0043] Furthermore, the retrieval module includes:

[0044] A tagging unit, which is used to export chat records from social applications and use a preset tagging tool to tag sensitive words with parts of speech;

[0045] A definition unit is used to classify the short sentences into different data sets according to the parts of speech and define data set attributes.

[0046] Furthermore, the output module includes:

[0047] An assignment unit, used to train the classifier using the acquired sentiment dictionary, assign values ​​to the short sentences according to the training results, and calculate the standard sentiment value of the short sentences based on a preset formula;

[0048] A recording unit, used to define the score segments of the standard sentiment value, and superimpose the score segments into the filter, determine the sentiment base value of the sensitive word based on the sentiment dictionary, and record the number of occurrences of the sensitive word, and calculate the weighted average value of the sentiment value using a preset sentiment value formula; substitute the calculated weighted average value into the standardization formula to calculate the standard sentiment value;

[0049] A building unit, used to use the data set as a source, connect the classifier and the filter using the pipeline, and build a data pipeline;

[0050] The superposition unit is used to input the data set into the data pipeline, determine whether the standard sentiment value of the short sentence meets the score segment of the filter, and if so, classify the short sentence into the filter corresponding to the score segment, superimpose the standard sentiment values ​​of all short sentences in the filter, and determine the superposition result as the total score.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] By extracting chat records, we can broaden the data sources and provide better decision-making support for risk warning of power dispatching. By building a data pipeline, the processing efficiency of chat records is greatly improved, so that the sentiment value in the chat records can be analyzed more accurately. By constructing a fluctuation curve, we can accurately understand the relationship between sentiment value and power demand, so as to achieve early warning of risks in power dispatching. At the same time, we can also optimize the power dispatching strategy and improve the flexibility and responsiveness of the power system. By calculating the difference and generating alarm information, we can more intuitively display the risk of power dispatching and improve the work efficiency of risk warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0054] Figure 1 A flowchart of a power dispatch risk warning method integrating power data provided by an embodiment of the present invention;

[0055] Figure 2 A first sub-flow chart of the electric power dispatch risk warning method integrating electric power data provided in an embodiment of the present invention;

[0056] Figure 3 A second sub-flow chart of the electric power dispatch risk warning method integrating electric power data provided in an embodiment of the present invention;

[0057] Figure 4 A third sub-flow chart of the electric power dispatch risk warning method integrating electric power data provided in an embodiment of the present invention;

[0058] Figure 5 A fourth sub-flow chart of the electric power dispatch risk early warning method integrating electric power data provided in an embodiment of the present invention;

[0059] Figure 6A block diagram of the composition of a power dispatch risk early warning system integrating power data provided by an embodiment of the present invention;

[0060] Figure 7 A block diagram of the composition of a retrieval module in a power dispatch risk early warning system integrating power data provided by an embodiment of the present invention;

[0061] Figure 8 A block diagram of the composition of an output module in a power dispatch risk early warning system integrating power data provided by an embodiment of the present invention;

[0062] Fig. 9 A block diagram of the composition of a drawing module in the electric power dispatch risk warning system integrating electric power data provided by an embodiment of the present invention;

[0063] Fig.10 A block diagram of the composition of an alarm module in a power dispatch risk warning system integrating power data provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0065] In Example 1, Figure 1 The implementation process of the power dispatch risk warning method integrating power data provided by an embodiment of the present invention is shown, and is described in detail below:

[0066] S100: extracting chat records in social applications, retrieving sensitive words in the chat records, integrating short sentences containing sensitive words, and creating a data set.

[0067] Extract chat records from social applications. The number of social applications here is at least one. Of course, such operations require user authorization. Retrieve sensitive words from the chat records, extract short sentences containing sensitive words, and then summarize all short sentences to create a data set.

[0068] In this embodiment, multiple categories of data sets may be generated, and the number of data sets is not limited.

[0069] S200: Create a pipeline, insert a classifier trained by the Naive Bayes algorithm into the pipeline, and attach at least one filter to the classifier, integrate the pipeline, classifier and filter, form a data pipeline, input the data set into the data pipeline, and output the total score of all short sentences.

[0070] Create a pipeline, where a pipeline is a special channel for transferring data between processes, which allows one process to continue to the next process after output, and insert a classifier into the pipeline, where the classifier should be trained in advance using the naive Bayes algorithm, and the naive Bayes algorithm is mainly used to predict the probability that a sample belongs to different categories; in this embodiment, the naive Bayes algorithm can assign values ​​to sensitive words, and multiple filters are hung in the classifier, and the number of filters is mainly determined by the range of sensitive word assignments. Use pipelines to connect classifiers and filters, form a data pipeline, and input the data set into the data pipeline.

[0071] S300: define target power data, determine the corresponding relationship between the total score and the target power data, and draw a fluctuation curve with the total score as the horizontal coordinate and the actually collected target power data as the vertical coordinate.

[0072] A test time period is selected, and then the total score of the short sentences in the test time period is calculated. According to the development trend of the power data, the change of the power data in the expected time period is judged, and the power data in the expected time period is determined as the target power data.

[0073] For example, short sentences such as "The weather will be good tomorrow, suitable for camping" and "Let's go on a picnic tomorrow" are retrieved from the chat records, and the total score of the short sentences is calculated. At this time, the power dispatch data of the next day is observed and determined as the target power data, the relationship between the total score and the target power data is found, and the fluctuation curve is drawn.

[0074] By observing the fluctuation curve, the changing trend of power dispatch can be predicted, thus providing decision support for power dispatch risk warning.

[0075] S400: Substitute the acquired real-time total score into the fluctuation curve, locate the target power data, determine the difference between the target power data and the real-time power data, and if the difference is greater than a preset threshold, generate an alarm message using the difference, and send the alarm message to a preset terminal.

[0076] When a power dispatch risk warning is needed, the total score of short sentences in the chat records within a certain period of time is calculated. This total score is called the real-time total score. The real-time total score is substituted into the fluctuation curve to find the corresponding target power data, which can characterize the development trend of power dispatch.

[0077] It should be noted that power dispatching is closely related to environmental data such as time and temperature. That is to say, in the data preparation stage, multiple fluctuation curves should be drawn, and the time, temperature, etc. of all fluctuation curves should be marked. When the target power data needs to be found, the calculated total score is used to find the fluctuation curve corresponding to the environmental data similar to the total score moment, and the target power data is found in the fluctuation curve. The difference between the target power data and the power data at the current moment is judged. If the difference is greater than the preset threshold, the difference is used to generate an alarm message; the threshold should be determined according to the actual dispatching capability of the power dispatching department. If the dispatching capability is strong, a larger threshold should be set. If the dispatching capability is weak, the threshold should be smaller; the alarm message is sent to a preset terminal, which can be the management system of the power dispatching department.

[0078] In this embodiment, the difference is the part that needs to be dispatched for power. If it exceeds the threshold, it is necessary to prepare for power dispatch in advance to avoid causing drastic fluctuations in power load and affecting the stability of the power system.

[0079] In Example 2, Figure 2 The implementation process of the power dispatch risk warning method integrating power data provided by an embodiment of the present invention is shown. The steps of extracting chat records in social applications, retrieving sensitive words in the chat records, integrating short sentences containing sensitive words, and creating a data set are described in detail as follows:

[0080] S101: Export chat records from social applications, and use a preset tagging tool to perform part-of-speech tagging on sensitive words.

[0081] Export all chat records from multiple social applications and use a preset annotation tool to perform part-of-speech tagging. The preset annotation tool can be provided by a third party or a machine learning model.

[0082] S102: Classifying the short sentences into different data sets according to the parts of speech, and defining data set attributes, where the attributes include: weather, temperature and humidity, and travel.

[0083] According to the part of speech of sensitive words, multiple data sets are created and the attributes of the data sets are defined.

[0084] In Example 3, Figure 3 The implementation process of the power dispatch risk warning method integrating power data provided by an embodiment of the present invention is shown. The steps of creating a pipeline, inserting a classifier trained by a naive Bayes algorithm into the pipeline, and hanging at least one filter under the classifier are described in detail as follows:

[0085] S201: The classifier is trained using the acquired sentiment dictionary, and the short sentences are assigned values ​​according to the training results, and the standard sentiment value of the short sentences is calculated based on a preset formula.

[0086] The emotional dictionary is used again to train the classifier, where the emotional dictionary can be obtained from the existing data. The classifier is used to assign values ​​to sensitive words, and after the assignment is completed, the preset formula is used to calculate the standard emotional value of the short sentence. By calculating the standard emotional value, the emotional color in the chat record can be judged, and by continuing to process the chat record, the development trend of power dispatching can be predicted.

[0087] S202: Delimiting score segments of standard sentiment values, and superimposing the score segments into the filters, wherein each score segment corresponds to a filter.

[0088] According to the calculated range of the standard sentiment value, the score segment of the standard sentiment value is delineated, and a score segment is set for each filter.

[0089] In Example 4, Figure 3 The implementation process of the power dispatch risk warning method integrating power data provided by an embodiment of the present invention is shown. The steps of calculating the standard sentiment value of a short sentence are described in detail as follows:

[0090] S203: Determine the basic sentiment value of the sensitive word based on the sentiment dictionary, and record the number of occurrences of the sensitive word;

[0091] The weighted average of the sentiment values ​​is calculated using the preset sentiment value formula;

[0092] The sentiment value formula is:

[0093]

[0094] where v i is the sentiment base value of the i-th sensitive word, c i is the number of occurrences of the i-th sensitive word, and V is the weighted average;

[0095] Substitute the calculated weighted average into the standardization formula to calculate the standard sentiment value;

[0096] The standardized formula is:

[0097]

[0098] Where V′ is the standard sentiment value, V is the weighted average, A is -1, B is 1, a is the minimum value among the sentiment base values, and b is the maximum value among the sentiment base values.

[0099] After the assignment is completed, the assignment of each sensitive word is defined as the basic sentiment value, the number of occurrences of each sensitive word is recorded, and the sentiment value formula is used to perform weighted average on the sentiment value. The calculated result is the sentiment value of the short sentence. For example, a short sentence is "The weather will be very good tomorrow, very suitable for camping", in which the sensitive words "very good", "very" and "camping" are assigned values ​​and weighted averaged. The result is determined as the sentiment value of the short sentence, and then the sentiment value of the short sentence is substituted into the standard sentiment value formula to determine the standard sentiment value.

[0100] The benefit of calculating the standard sentiment value is that it can standardize the data and reduce the complexity and processing difficulty of the data.

[0101] In Example 5, Figure 3 The implementation process of the power dispatch risk warning method integrating power data provided by an embodiment of the present invention is shown. The steps of integrating the pipeline, classifier and filter, forming a data pipeline, inputting the data set into the data pipeline, and outputting the total score of the sensitive words are described in detail as follows:

[0102] S204: Using the data set as a source, using the pipeline to connect the classifier and the filter to form a data pipeline.

[0103] Taking the data set as the source, the classifier and all filters are connected by pipeline to form a data pipeline. By forming a data pipeline, the processing efficiency of chat records can be greatly improved and the accuracy of the target power data can be ensured.

[0104] S205: Input the data set into the data pipeline, determine whether the standard sentiment value of the short sentence satisfies the score segment of the filter, and if so, classify the short sentence into the filter corresponding to the score segment.

[0105] The data set is input into the data pipeline, and the short sentences are classified into the corresponding filters; for example, two filters are installed on the pipeline. If the standard sentiment value of the short sentence meets the score segment requirement of the first filter, it will be retained in the first filter. If it does not meet the requirement, it will enter the second filter through the pipeline, and so on, so that the short sentence is classified into the corresponding filter.

[0106] S206: superimpose the standard sentiment values ​​of all the short sentences in the filter, and determine the superimposed result as the total score.

[0107] The standard sentiment values ​​of all short sentences in the filter are superimposed, and the superposition result is determined as the total score. It should be noted that the total score here is the superposition of the standard sentiment values ​​of all short sentences in all filters.

[0108] In Example 6, Figure 4The implementation process of the power dispatch risk warning method integrating power data provided by an embodiment of the present invention is shown. The steps of defining the target power data, determining the corresponding relationship between the total score and the target power data, and drawing the fluctuation curve with the total score as the horizontal coordinate and the target power data as the vertical coordinate are described in detail as follows:

[0109] S301: Based on the corresponding relationship, draw a fluctuation curve, and insert a label consisting of the attributes into the fluctuation curve.

[0110] According to the correspondence between the total score and the target power data, a fluctuation curve is drawn, where the target power data needs to be summarized based on the actual power dispatching data. At the same time, labels are generated according to the attributes of the data set and inserted into the fluctuation curve.

[0111] S302: Select a turning point in the fluctuation curve, use the turning point as a trigger, and embed a trigger mechanism.

[0112] Find the turning point in the fluctuation curve. The turning point here is the point where the power dispatching data changes dramatically. Use the turning point as the trigger to establish a trigger mechanism. The trigger mechanism can be an automatic adjustment of the power load or an emergency measure.

[0113] In this embodiment, the coordinates of the turning point are found, and when the total score is equal to the horizontal coordinate of the turning point, the vertical coordinate of the turning point is used to modify the trigger mechanism.

[0114] In Example 7, Figure 5 The implementation process of the power dispatch risk warning method integrating power data provided by an embodiment of the present invention is shown. The following is a detailed description of the steps of substituting the acquired real-time total score into the fluctuation curve, locating the target power data, determining the difference between the target power data and the real-time power data, and if the difference is greater than a preset threshold, generating an alarm message using the difference, and sending the alarm message to a preset terminal, as follows:

[0115] S401: using the real-time total score, traverse corresponding points in the fluctuation curve, and define the ordinate of the corresponding point as target power data.

[0116] The real-time total score is the total score of the short sentences in the chat record within the predetermined time. After the total score is calculated, the corresponding point is traversed in the fluctuation curve, and the vertical coordinate of the corresponding point is defined as the target power data.

[0117] S402: Divide the difference into at least two difference segments, determine the alarm level based on a preset alarm strategy, and integrate the difference to generate alarm information.

[0118] Real-time power data refers to the power data under actual working conditions. When the difference between the target power data and the real-time power data is calculated, the difference is divided into multiple segments, and different alarm levels are set for each segment according to the preset alarm strategy. The alarm strategy can be to set different transmission methods and expression methods for different alarm levels. The advantage of this is that it can more intuitively display the urgency of the alarm information and better provide decision support for power dispatching.

[0119] Figure 6 The structure block diagram of the power dispatch risk early warning system integrating power data provided by an embodiment of the present invention is shown. The power dispatch risk early warning system integrating power data 1 includes:

[0120] A retrieval module 11 is used to extract chat records in social applications, retrieve sensitive words in the chat records, integrate short sentences containing sensitive words, and create a data set;

[0121] The output module 12 is used to create a pipeline, insert a classifier trained by the naive Bayes algorithm into the pipeline, and hang at least one filter under the classifier, integrate the pipeline, the classifier and the filter, form a data pipeline, input the data set into the data pipeline, and output the total score of all the short sentences;

[0122] A drawing module 13 is used to define target power data, determine the corresponding relationship between the total score and the target power data, and draw a fluctuation curve with the total score as the horizontal coordinate and the actually collected target power data as the vertical coordinate;

[0123] The alarm module 14 is used to substitute the acquired real-time total score into the fluctuation curve, locate the target power data, determine the difference between the target power data and the real-time power data, and if the difference is greater than a preset threshold, generate an alarm message using the difference, and send the alarm message to a preset terminal.

[0124] Figure 7 The structure diagram of the electric power dispatch risk early warning system integrating electric power data provided by an embodiment of the present invention is shown, and the retrieval module 11 includes:

[0125] A tagging unit 111 is used to export chat records from social applications and use a preset tagging tool to tag the sensitive words with parts of speech;

[0126] The definition unit 112 is used to classify the short sentences into different data sets according to the parts of speech and define data set attributes.

[0127] Figure 8The structure block diagram of the power dispatch risk early warning system integrating power data provided by an embodiment of the present invention is shown, and the output module 12 includes:

[0128] The assigning unit 121 is used to train the classifier using the acquired sentiment dictionary, assign values ​​to the short sentences according to the training results, and calculate the standard sentiment value of the short sentences based on a preset formula;

[0129] A delimiting unit 122, used for delimiting the score segments of the standard sentiment value and superimposing the score segments into the filter;

[0130] The recording unit 123 can determine the basic sentiment value of the sensitive word based on the sentiment dictionary, record the number of occurrences of the sensitive word, calculate the weighted average of the sentiment value using a preset sentiment value formula, and substitute the calculated weighted average into a standardization formula to calculate a standard sentiment value;

[0131] A building unit 124, configured to use the data set as a source, connect the classifier and the filter using the pipeline, and build a data pipeline;

[0132] The classification unit 125 is used to input the data set into the data pipeline, determine whether the standard sentiment value of the short sentence meets the score segment of the filter, and if so, classify the short sentence into the filter corresponding to the score segment;

[0133] The superposition unit 126 is used to superimpose the standard sentiment values ​​of all the short sentences in the filter and determine the superposition result as the total score.

[0134] Fig. 9 The structure diagram of the electric power dispatch risk warning system integrating electric power data provided by an embodiment of the present invention is shown, and the drawing module 13 includes:

[0135] An inserting unit 131, capable of drawing a fluctuation curve based on the corresponding relationship, and inserting a label consisting of the attributes into the fluctuation curve;

[0136] The trigger unit 132 is used to select a turning point in the fluctuation curve and use the turning point as a trigger head to embed a trigger mechanism.

[0137] Fig.10 The structure block diagram of the electric power dispatch risk early warning system integrating electric power data provided by an embodiment of the present invention is shown, and the alarm module 14 includes:

[0138] A traversal unit 141 is used to traverse corresponding points in the fluctuation curve using the real-time total score, and define the ordinate of the corresponding point as target power data;

[0139] The generating unit 142 is used to divide the difference into at least two difference segments, determine the alarm level based on a preset alarm strategy, and integrate the difference to generate alarm information.

[0140] The retrieval module 11 is mainly used to complete step S100, the output module 12 is mainly used to complete step S200, the drawing module 13 is mainly used to complete step S300, and the alarm module 14 is mainly used to complete step S400;

[0141] The labeling unit 111 is mainly used to complete step S101, and the defining unit 112 is mainly used to complete step S102;

[0142] The assignment unit 121 is mainly used to complete step S201, the delimitation unit 122 is mainly used to complete step S202; the recording unit 123 is mainly used to complete step S203, the formation unit 124 is mainly used to complete step S204; the classification unit 125 is mainly used to complete step S205, and the superposition unit 126 is mainly used to complete step S206;

[0143] The inserting unit 131 is mainly used to complete step S301, and the triggering unit 132 is mainly used to complete step S302;

[0144] The traversal unit 141 is mainly used to complete step S401, and the generation unit 142 is mainly used to complete step S402.

[0145] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0146] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0147] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A power dispatch risk early warning method integrating power data, characterized in that: The method comprises: Extract chat records from social applications, retrieve sensitive words from the chat records, integrate short sentences containing sensitive words, and create a data set; Building a pipeline, inserting a classifier trained by the Naive Bayes algorithm into the pipeline, and attaching at least one filter to the classifier, integrating the pipeline, the classifier and the filter, forming a data pipeline, inputting the data set into the data pipeline, and outputting the total score of the sensitive words; Define target power data, determine the corresponding relationship between the total score and the target power data, and draw a fluctuation curve with the total score as the abscissa and the actually collected target power data as the ordinate; Substituting the acquired real-time total score into the fluctuation curve, locating the target power data, determining the difference between the target power data and the real-time power data, and if the difference is greater than a preset threshold, generating an alarm message using the difference, and sending the alarm message to a preset terminal; The steps of constructing a pipeline, inserting a classifier trained by the naive Bayes algorithm into the pipeline, and attaching at least one filter to the classifier include: The classifier is trained using the acquired sentiment dictionary, and the short sentences are assigned values ​​according to the training results, and the standard sentiment value of the short sentences is calculated based on a preset formula; Delimiting the score segments of the standard sentiment value, and superimposing the score segments into the filter; Each score segment corresponds to a filter; The step of calculating the standard sentiment value of the short sentence comprises: Based on the sentiment dictionary, determine the sentiment base value of the sensitive word and record the number of occurrences of the sensitive word; The weighted average of the sentiment values ​​is calculated using the preset sentiment value formula; The sentiment value formula is: where v i is the sentiment base value of the i-th sensitive word, c i is the number of occurrences of the i-th sensitive word, and V is the weighted average; Substitute the calculated weighted average into the standardization formula to calculate the standard sentiment value; The standardized formula is: Where V′ is the standard sentiment value, V is the weighted average, A is -1, B is 1, a is the minimum value in the sentiment base value, and b is the maximum value in the sentiment base value; The steps of integrating the pipeline, the classifier and the filter, forming a data pipeline, inputting the data set into the data pipeline, and outputting the total score of the sensitive words include: Using the data set as a source, connecting the classifier and the filter using the pipeline to form a data pipeline; Input the data set into the data pipeline, determine whether the standard sentiment value of the short sentence satisfies the score segment of the filter, and if so, classify the short sentence into the filter corresponding to the score segment; The standard sentiment values ​​of all short sentences in all filters are superimposed, and the superposition result is determined as the total score.

2. The power dispatch risk early warning method integrating power data according to claim 1 is characterized in that: The steps of extracting chat records from social applications, retrieving sensitive words from the chat records, integrating short sentences containing sensitive words, and creating a data set include: Exporting chat records from the social application and using a preset annotation tool to perform part-of-speech tagging on sensitive words; The short sentences are classified into different data sets according to the parts of speech, and data set attributes are defined, wherein the attributes include: weather, temperature and humidity, and travel.

3. The power dispatch risk early warning method integrating power data according to claim 2 is characterized in that: The steps of defining the target power data, determining the corresponding relationship between the total score and the target power data, and drawing the fluctuation curve with the total score as the horizontal coordinate and the target power data as the vertical coordinate include: Based on the corresponding relationship, a fluctuation curve is drawn, and a label consisting of the attributes is inserted into the fluctuation curve; A turning point in the fluctuation curve is selected, and the turning point is used as a trigger head to embed a trigger mechanism.

4. The power dispatch risk early warning method integrating power data according to claim 3 is characterized in that: The step of substituting the acquired real-time total score into the fluctuation curve, locating the target power data, determining the difference between the target power data and the real-time power data, and if the difference is greater than a preset threshold, generating an alarm message using the difference, and sending the alarm message to a preset terminal comprises: Using the real-time total score, traverse corresponding points in the fluctuation curve, and define the ordinates of the corresponding points as target power data; The difference is divided into at least two difference segments, an alarm level is determined based on a preset alarm strategy, and the difference is integrated to generate alarm information.