Power grid customer service work order seasonal feature identification method

Abstract

The invention relates to a power grid customer service work order seasonal feature identification method. The seasonal feature recognition method comprises the following steps: sorting the number of work orders to be recognized in a plurality of time periods to obtain a feature data sequence; the sorting sequence is from large to small; determining at least one optimal segmentation point based on the feature data sequence; according to the at least one optimal segmentation point and a preset time period number threshold value, determining an identification sub-sequence from sub-sequences of the feature data sequence; and judging whether the month to be identified is a seasonal month or not according to feature data included in the identification sub-sequence. The power grid customer service work order seasonality is judged through a quantitative method, and a feasible scheme is provided for large-scale work order processing of a computer.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method for identifying seasonal characteristics of power grid customer service work orders. Background Art

[0002] At present, with the accelerating transformation of the energy structure and the rapid development of technology, the power industry is undergoing profound changes. Its service quality is not only the key to ensuring stable and reliable power supply in the region, but also an important support for maintaining the smooth operation of the economic society.

[0003] As the unified service hotline of the State Grid facing customers, the work orders and reply forms generated by 95598 directly reflect various problems encountered by customers during the power consumption process. By deeply analyzing the content of these work orders and reply forms, service problems can be accurately identified.

[0004] From a seasonal perspective, a large number of work orders regarding power outage repair during the summer peak electricity consumption period can help relevant units quickly locate the overloaded areas of the load and understand the prominent points of seasonal power supply and demand contradictions. Based on this information, the provincial company can prepare seasonal emergency plans in advance, rationally allocate resources, conduct equipment maintenance and capacity expansion in key areas before the peak arrives, and effectively improve the seasonal power service guarantee ability.

[0005] For temporary service problems, the 95598 work orders and reply forms can play an immediate feedback role. Once there are temporary problems such as line failures and electricity prices and fees caused by extreme weather, the information such as the time and location of the failure recorded in the 95598 work order, the customer description, and the repair process and results can help the provincial company quickly review the temporary events, optimize the emergency handling process, and improve the emergency response speed.

[0006] By analyzing the content of 95598 work orders and reply forms and mining the data value therein, a comprehensive and detailed insight into service problems can be provided. Collaborating with the provincial company to improve service strategies and optimize resource allocation in a targeted manner can comprehensively improve service quality, better meet customer needs, ensure the safe and stable power supply in the region, and lay a solid power foundation for the economic and social development.

[0007] In view of this, there is an urgent need for a method that can identify the seasonal characteristics of power grid customer service work orders. Summary of the Invention

[0008] The present invention provides a method for identifying seasonal characteristics of power grid customer service work orders, and the specific technical solution is as follows:

[0009] A method for identifying seasonal characteristics of power grid customer service work orders includes the following processes:

[0010] Step S100: sorting the number of work orders to be identified in multiple time periods to obtain a feature data sequence; the sorting order is arranged from large to small;

[0011] Step S200: Determine at least one optimal segmentation point based on the feature data sequence; specifically includes:

[0012] Step S210: Any two consecutive numbers in the feature data sequence can be used as a split point; sequence s = (s1, s2, ..., s n ) has n-1 split points, s p-1 With s p The split point between the two subsequences is:

[0013] s L =(s1,s2,...,s p-1 ),

[0014] s R =(s p ,s p+1 ,...,s n );

[0015] Step S220: for each split point of the feature data sequence, determine its split ratio; the split ratio is the ratio of the mean values ​​of the feature data in the two subsequences corresponding to the split point;

[0016] Step S230: taking the segmentation point with the largest segmentation ratio as the optimal segmentation point;

[0017] Step S240: determine whether the number of feature data in the subsequence corresponding to the best split point is greater than a preset number threshold; if greater, determine an initial subsequence from the two subsequences corresponding to the best split point according to the recognition requirements; for each split point of the initial subsequence, determine its split ratio, and determine the best split point according to the split ratio of each split point of the initial subsequence, and return to step S240; if less, go to the next step;

[0018] Step S300: According to at least one optimal segmentation point and a preset time period threshold, select a sub-segment of the feature data sequence.

[0019] Determine the identification subsequence in the sequence; specifically include the following process:

[0020] For each subsequence extracted, determine whether the number of feature data in the subsequence is greater than a preset time period threshold; if greater, determine the subsequence as an identification subsequence;

[0021] Step S400: judging whether the month to be identified is a seasonal month according to the characteristic data included in the identification subsequence.

[0022] Preferably, after the step S400, the following process is performed:

[0023] Generate a seasonal information annotation vector for the data to be recognized according to the judgment result of whether it is a seasonal month and the time period corresponding to the recognition data; the seasonal information annotation vector season_label = (is_season, l1, l2,..., l n ), where "is_season" represents a seasonal work order, and l1, l2,..., l n Each corresponds to each of a plurality of time periods, and "0" and "1" respectively represent a non-seasonal time period and a seasonal time period.

[0024] Preferably, the step S400 includes the following process:

[0025] Judge whether the month to be recognized is a seasonal month according to the time period mean value of the recognition data and the time period mean value of the feature data in the feature data sequence; specifically, judge whether the time period mean value of the recognition data is greater than a preset multiple of the time period mean value of the feature data in the feature data sequence; if it is greater, the month to be recognized is a seasonal month; otherwise, the month to be recognized is a non-seasonal month.

[0026] Preferably, the step S400 includes the following process:

[0027] Judge whether the month to be recognized is a seasonal month by comparing the ratio or difference between the time period mean value of the recognition data and the time period mean value of the feature data in the feature data sequence with the corresponding preset value.

[0028] The present invention determines the seasonality of power grid customer service work orders through a quantitative method, providing a feasible solution for large-scale computer processing of work orders. Description of the Drawings

[0029] Figure 1 It is a schematic flowchart of the method of the present invention. Detailed Embodiments

[0030] Embodiment 1

[0031] A method for identifying the seasonal characteristics of power grid customer service work orders includes the following process:

[0032] Step S100: Sort the number of work orders to be recognized in multiple time periods to obtain a feature data sequence; the sorting order is from large to small.

[0033] Step S200: Based on the feature data sequence, determine at least one optimal cut-off point; specifically including:

[0034] Step S210: Between any two consecutive numbers in the feature data sequence, it can be used as a segmentation point; for the sequence s=(s1, s2,..., s n ) there are n - 1 segmentation points. The segmentation point between s p-1 and s p divides the sequence s into two subsequences:

[0035] s L =(s1, s2,..., s p-1 ),

[0036] s R =(s p , s p+1 ,..., s n ).

[0037] Step S220: For each segmentation point of the feature data sequence, determine its segmentation ratio; the segmentation ratio is the ratio of the means of the feature data in the two subsequences corresponding to this segmentation point.

[0038] Step S230: Determine the optimal segmentation point according to the segmentation ratios of each segmentation point of the feature data sequence. Since this segmentation ratio can reflect the relationship between the data in the two subsequences corresponding to the segmentation point, determining the optimal segmentation point based on this segmentation ratio is beneficial to improving the accuracy of work order seasonal recognition; take the segmentation point with the largest segmentation ratio as the optimal segmentation point.

[0039] Step S240: Judge whether the number of feature data in the subsequence corresponding to the optimal segmentation point is greater than the preset number threshold; if it is greater, then according to the recognition requirements, determine the initial subsequence from the two subsequences corresponding to the optimal segmentation point; for each segmentation point of the initial subsequence, determine its segmentation ratio, and according to the segmentation ratios of each segmentation point of the initial subsequence, determine the optimal segmentation point, and return to Step S240; if it is less, then go to the next step.

[0040] For example: The average quantities of a certain work order type in 12 months are 20, 22, 30, 90, 60, 40, 35, 28, 26, 120, 80, 30 respectively; after sorting in descending order, the obtained sequence s=(120, 90, 80, 60, 40, 35, 30, 30, 28, 26, 22, 20), and the sorted months and corresponding work order data are as follows in the table:

[0041] Month 10 4 11 5 6 7 3 12 8 9 2 1 Work order data 120 90 80 60 40 35 30 30 28 26 22 20

[0042] For the sequence s=(120, 90, 80, 60, 40, 35, 30, 30, 28, 26, 22, 20) there are 11 segmentation points, and these 11 segmentation points are respectively marked as 1, 2, 3…11; the segmentation point 1 divides the sequence s into sL =(120) and s R =(90, 80, 60..., 20), these two subsequences are the subsequences corresponding to split point 1; Split point 2 divides sequence s into s L =(120, 90) and s R =(80, 60, 40..., 20), these two subsequences are the subsequences corresponding to split point 2; The split ratio of the split point is r = mean(s L ) / mean(s R ), where mean(s L ) represents the average value of the characteristic data included in subsequence s L . The split ratio of split point 1 is

[0043] 120 / [(90 + 80 + 60 +... + 20) / 11], approximately 2.863962. Similarly, the split ratios of split points 2, 3…11 can be determined, as shown in the following table:

[0044]

[0045] Take the split point with the largest split ratio as the optimal split point. In sequence s, the optimal split point is split point 4, and the subsequences corresponding to split point 4 are (120, 90, 80, 60) and (40, 35, 30, 30, 28, 26, 22, 20). In the embodiment of the present invention, the preset number threshold is 1. As can be seen from the above, the number of characteristic data in the subsequences corresponding to split point 4 is greater than 1, so the optimal split point needs to be further determined.

[0046] In the embodiment of the present invention, to determine the seasonality of the work order quantity, it is necessary to further judge the months with a large number of work orders. On this basis, based on the sequence (120, 90, 80, 60), the optimal split point is further determined. This sequence has 3 split points, and their split ratios are 1.57, 1.50, and 1.61 respectively; Then, based on the determined second optimal split, the determined subsequences are respectively (120, 90, 80) and (60); At this time, the number of characteristic data in the subsequence (60) is not greater than 1, so the optimal split point is not continued to be determined; In the embodiment of the present invention, based on the determined two optimal split points, multiple subsequences are intercepted from sequence s, which are: (120, 90, 80), (60), and (40, 35, 30, 30, 28, 26, 22, 20).

[0047] Step S300: Determine the recognition subsequence from the subsequences of the characteristic data sequence according to at least one optimal split point and the preset time period number threshold; The specific process is as follows:

[0048] For each extracted subsequence, determine whether the number of feature data in the subsequence is greater than a preset time period threshold; if it is greater, then determine the subsequence as an identification subsequence.

[0049] Set the time period threshold according to the work order quantity pattern or the requirement of identification accuracy. If the number of feature data included in the subsequence is greater than the time period threshold, it means that the feature data included in the subsequence is not concentrated enough, which is not conducive to the judgment of the seasonality of the work order quantity.

[0050] For example: the time period threshold is set to 6, which means that when determining seasonal months, the monthly work order quantity included in the determined identification subsequence cannot exceed the work order quantity of 6 months; in the above example, based on determining two optimal cut-off points, multiple subsequences are extracted from the sequence s, which are: (120, 90, 80), (60), and (40, 35, 30, 30, 28, 26, 22, 20); the time period threshold is set to 6, then according to this time period threshold, the determined identification subsequences are (120, 90, 80) and (60).

[0051] Step S400: According to the feature data included in the identification subsequence, determine whether the month to be identified is a seasonal month; according to the time period mean of the identification data and the time period mean of the feature data in the feature data sequence, determine whether the month to be identified is a seasonal month; specifically, determine whether the time period mean of the identification data is greater than a preset multiple of the time period mean of the feature data in the feature data sequence; if it is greater, then the month to be identified is a seasonal month; otherwise, the month to be identified is a non-seasonal month.

[0052] This preset multiple can be set according to the work order characteristics or identification requirements. For example, if this preset multiple is set to 2, then when the time period mean of the identification data is greater than 2 times the time period mean of the feature data in the feature data sequence, the month to be identified is a seasonal month.

[0053] Step S500: Generate a seasonal information annotation vector for the data to be identified according to the judgment result of whether it is a seasonal month and the time period corresponding to the identification data; this seasonal information annotation vector season_label = (is_season, l1, l2,..., l n ), where, "is_season" represents a seasonal work order, and l1, l2,..., l n correspond one by one to each of multiple time periods, and are represented by "0" and "1" respectively for non-seasonal time periods and seasonal time periods.

[0054] For example, if the seasonal information annotation vector of work order A is (is_season, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0), it means that work order A is a seasonal work order, and its seasonal time periods are October and November.

[0055] Example 2

[0056] Steps S100 - S300, and S500 in this example are the same as those in the example. The difference lies in step S400. The said step S400 includes the following process:

[0057] By comparing the ratio or difference between the period mean of the identified data and the period mean of the characteristic data in the characteristic data sequence with the corresponding preset value, it is determined whether the month to be identified is a seasonal month.

[0058] The characteristic data included in the identification subsequence are 120, 90, 80, and 60, and the corresponding months are October, April, November, and May respectively; among them, April and May are consecutive months, and October and November are consecutive months. Based on this period order, the characteristic data included in the subsequence are combined to obtain 150 and 200; if the preset screening threshold is 2, two data are selected from the combined characteristic data as the identified data, and if the preset screening threshold is 1, one data is selected from the combined characteristic data as the identified data.

[0059] In the above example, assuming the preset screening threshold is 2, the identified data screened out are (150, 200); the work order quantities in 4 months (i.e., 4 periods) are included in 150 and 200, and the period mean is (150 + 200) / 4 = 87.5; and, the work order quantities in 12 months (i.e., 12 periods) are included in sequence s, and its period mean is (120, 90, 80, 60, 40, 35, 30, 30, 28, 26, 22, 20) / 12 = 48.42.

[0060] In the above example, the preset multiple is set to 2, then 87.5 is less than twice of 48.42. Therefore, in this example, the month to be identified is a non - seasonal month.

[0061] If, the preset screening threshold is 1, the identified data screened out are 150 or 200. The work order quantities in 2 months (i.e., 2 periods) are included in 150, and its period mean is 150 / 2 = 75; the work order quantities in 2 months (i.e., 2 periods) are included in 200, and its period mean is 200 / 2 = 100; at this time, 100 is greater than twice of 48.42, and the month to be identified can be determined as a seasonal month.

Claims

1. A method for identifying seasonal characteristics of power grid customer service work orders, characterized in that: The process includes the following: Step S100: sorting the number of work orders to be identified in multiple time periods to obtain a feature data sequence; The sorting order is from largest to smallest; Step S200: Determine at least one optimal segmentation point based on the feature data sequence; specifically includes: Step S210: Any two consecutive numbers in the feature data sequence can be used as a split point; sequence s = (s1, s2, ..., s n ) has n-1 split points, s p-1 With s p The split point between the two subsequences is: s L =(s1,s2,...,s p-1 ), s R =(s p ,s p+1 ,...,s n ); Step S220: for each split point of the feature data sequence, determine its split ratio; the split ratio is the ratio of the mean values ​​of the feature data in the two subsequences corresponding to the split point; Step S230: taking the segmentation point with the largest segmentation ratio as the optimal segmentation point; Step S240: determine whether the number of feature data in the subsequence corresponding to the best split point is greater than a preset number threshold; if greater, determine an initial subsequence from the two subsequences corresponding to the best split point according to the recognition requirements; for each split point of the initial subsequence, determine its split ratio, and determine the best split point according to the split ratio of each split point of the initial subsequence, and return to step S240; if less, go to the next step; Step S300: determining an identification subsequence from a subsequence of a feature data sequence according to at least one optimal segmentation point and a preset time period number threshold; specifically comprising the following process: For each subsequence extracted, determine whether the number of feature data in the subsequence is greater than a preset time period threshold; if greater, determine the subsequence as an identification subsequence; Step S400: judging whether the month to be identified is a seasonal month according to the characteristic data included in the identification subsequence.

2. The method for identifying seasonal characteristics of power grid customer service work orders according to claim 1, characterized in that: The step S400 is followed by the following process: According to the judgment result of whether it is a seasonal month and the time period corresponding to the identified data, a seasonal information labeling vector of the data to be identified is generated; the seasonal information labeling vector season_label=(is_season, l1, l2, ..., l n ), where "is_season" indicates seasonal work orders, l1, l2, ..., l n Corresponding to each of the plurality of time periods one by one, "0" and "1" are used to represent the non-seasonal time period and the seasonal time period, respectively.

3. The method for identifying seasonal characteristics of power grid customer service work orders according to claim 1 or 2, characterized in that: The step S400 includes the following process: Based on the period mean of the identification data and the period mean of the characteristic data in the characteristic data sequence, determine whether the month to be identified is a seasonal month; wherein, determine whether the period mean of the identification data is greater than a preset multiple of the period mean of the characteristic data in the characteristic data sequence; if so, the month to be identified is a seasonal month; otherwise, the month to be identified is a non-seasonal month.

4. The method for identifying seasonal characteristics of power grid customer service work orders according to claim 1 or 2, characterized in that: The step S400 includes the following process: By comparing the ratio or difference between the period mean of the identification data and the period mean of the feature data in the feature data sequence with the corresponding preset value, it is determined whether the month to be identified is a seasonal month.

Images