CN-121097665-B - Power load prediction analysis system and method based on big data

CN121097665BCN 121097665 BCN121097665 BCN 121097665BCN-121097665-B

Abstract

The invention discloses a power load prediction analysis system and a power load prediction analysis method based on big data, which relate to the technical field of power load analysis and comprise a data statistics module, a period marking module, a power load data classification module, a response filling demand model construction module, a data type division module, a scene characteristic output module, a filling data set generation module and a real-time response filling module; by classifying the missing data according to the missing reasons and introducing three quantization indexes of the output data completion ratio, the output data stability ratio and the prediction accuracy value, the influence of different types of missing data on the prediction result can be accurately estimated, the 'one-cut' estimation blind area in the traditional technology is avoided, and a scientific basis is provided for the subsequent filling decision.

Inventors

QIAO HUAGUO
WANG JIANDE
ZHANG YINCHANG
NIE YUAN
WANG CONG
WANG JIANHONG
ZOU XIN
Cui ao
WANG JINXIN

Assignees

青岛裕华电子科技有限公司

Dates

Publication Date: 20260508
Application Date: 20250905

Claims (7)

1. The power load prediction analysis method based on big data is characterized by comprising the following steps: Step S100, counting historical recorded power load data and predicted events in each monitoring area, wherein the predicted events refer to events for executing power prediction targets based on the power load data in a time period, dividing the historical data into a plurality of event periods according to the recording period of the power load data required by the predicted events, identifying data missing sections in the event periods, marking the event periods with the data missing sections as investigation periods and the event periods without the data missing sections as safety periods, extracting missing power load data of the data missing sections in all investigation periods, and carrying out data identification and classification aiming at missing reasons; Step 200, acquiring the missing duration of a data missing segment record where each type of missing power load data is located, constructing a response filling demand model, and outputting a safe power load data type and a missing power data type to be analyzed; Traversing and searching prediction precision values of all prediction events of the various types of missing power load data records, if the prediction precision values of all the prediction events are larger than a threshold value, marking the corresponding type of missing power load data as a safe power load data type, and extracting a maximum value of missing duration as a critical missing monitoring value; Step 300, extracting a missing period set of a missing power data history record to be analyzed and a investigation period, summarizing missing time length and a type of the period in the investigation period and specific missing reasons, and determining scene characteristics of power load monitoring; Step 400, when the response identifies the real-time data missing segment, judging the type of the power load data, and matching the response filling of the filling data set based on the judging result; The step S400 includes the following steps: Acquiring a missing reason corresponding to a real-time data missing segment, acquiring the missing duration of a real-time data missing segment record if the missing reason is the same as the missing reason corresponding to the safety power load data, and not responding to data filling if the missing duration is less than or equal to a critical missing monitoring value of the corresponding safety power load data; When the missing reason is the same as the missing reason corresponding to the missing power data to be analyzed, acquiring the real-time scene characteristics of the real-time data missing segment record, searching the filling data set corresponding to the missing power data record to be analyzed, and matching the scene monitoring characteristics identical to the real-time scene characteristics and the corresponding filling data to respond.
2. The method for predictive analysis of electrical load based on big data as set forth in claim 1, wherein said step S100 comprises the steps of: Extracting the minimum node interval duration and the average node interval duration recorded by the adjacent power load data acquisition time nodes in each event period, and outputting a section corresponding to the node interval duration as a data missing section when the difference between the recorded node interval duration and the minimum node interval duration is larger than a first difference threshold and larger than the average node interval duration; The method comprises the steps of obtaining the missing reasons of missing power load data records in a data missing section, wherein the missing reasons comprise equipment faults, environmental interference and manual operation, marking the missing power load data which record the same missing reasons in the same monitoring area as one type of data, and distinguishing the missing power load data recorded in different events under the same type of data by using a prediction event as a marking index.
3. The method for predictive analysis of electrical load based on big data as set forth in claim 1, wherein said step S200 comprises the steps of: Step S210, acquiring the average data record quantity Q of each type of missing power load data in the monitoring area in unit time length and the total unit time length T of each investigation period record of the type of missing power load data, extracting the actual load data storage quantity Q of the corresponding investigation period, and calculating the output data completion ratio D, D=Q/qT of the corresponding type of missing power load data under a prediction event; Step S220, taking the unit time length recorded in each investigation period where the same type of missing power load data is located as a data acquisition node, taking the data acquisition node as an abscissa and the power load data acquired in the unit time length corresponding to the acquisition node as an ordinate, drawing a power load data acquisition graph of each prediction event, taking nodes with the power load data difference value recorded by the adjacent acquisition nodes being more than or equal to a difference threshold as segmentation nodes, traversing all the data acquisition nodes to finish time period segmentation, taking the time period contained by the adjacent segmentation nodes as a characteristic time period, extracting peaks Gu Chazhi Q Difference of difference recorded in each characteristic time period, wherein the peaks Gu Chazhi represent the difference value between the maximum value and the minimum value of the power load data in the characteristic time period, extracting the sum Q 1 of the peak-valley difference values in the corresponding safety period of the same monitoring area of the historical record, and solving the average Q 1Are all of the sum of the peak-valley difference values in all the safety periods, and calculating the output data stability ratio F in the prediction event corresponding to the power load data of the corresponding type of the prediction event by using the formula of F= (ΣQ Difference of difference )/Q 1Are all ; Step S230, extracting a predicted result value U 1 and an actual load value U 2 of a predicted model of the same type of missing power load data in each predicted event, and calculating a predicted precision value Y of the same type of missing power load data corresponding to one predicted event by using a formula Y=1- [ |U 2 -U 1 ｜/U 2 ; Step S240, taking the output data completion ratio and the output data stability ratio of the same type of missing power load data records as input variables, taking a prediction precision value Y corresponding to the same prediction event as an output variable, generating a data group A, A= [ (D, F) and Y ], summarizing the data groups A corresponding to all the prediction events of the same type of missing power load data records, and constructing a response filling demand function Y, wherein y=k 1 *d D +k 2 *f F +epsilon; step S250, a prediction precision value threshold Y 0 is set, a response filling demand model is constructed, namely, the output data completion ratio and the output data stability ratio which are acquired in real time are input into a response filling demand function Y, if the theoretical prediction precision value Y Measuring ≥Y 0 is obtained, the response is not triggered, and if the theoretical prediction precision value Y Measuring <Y 0 is obtained, the filling demand response is triggered.
4. The method for predictive analysis of electrical load based on big data as set forth in claim 3, wherein said step S300 comprises the steps of: Step S310, the type of the time period is the time period difference divided by the peak-valley difference value in the characteristic time period, the peak Gu Chazhi average value Q Flat plate and standard deviation Q Standard deviation of all characteristic time periods recorded in each prediction event of the same type of the missing power data to be analyzed are calculated, and the [ Q Flat plate -Q Standard deviation ,Q Flat plate +Q Standard deviation ] is used as the interval length to divide all the characteristic time periods of the complete investigation period to generate the corresponding time period type; step 320, taking a feature collection chart, a deletion reason, a deletion time length and a feature period corresponding to the deletion as scene monitoring features of the corresponding type of power deletion data to be analyzed; Step S330, the predicted event in the safety period when the scene features are the same refers to the predicted event which corresponds to the safety period when the feature collection diagram similarity is greater than a similarity threshold, the reasons of the absence are the same, the absolute value of the difference value of the duration of the absence is less than a duration difference threshold, and the feature period corresponding to the absence is the same, the predicted event is marked as a comparison event, the power load data of the comparison event corresponding to the overlapping part of the absence period is extracted as the filling data of the predicted event corresponding to each absence period, and a filling data set of the associated filling data under the condition that the power data to be analyzed corresponds to different scene monitoring features is generated.
5. A big data-based power load prediction analysis system, as set forth in any one of claims 1-4, characterized in that the system comprises a data statistics module, a period marking module, a power load data classification module, a response filling demand model construction module, a data type division module, a scene feature output module, a filling data set generation module and a real-time response filling module; the data statistics module is used for counting historical recorded power load data and predicted events in each monitoring area; The period marking module is used for dividing historical data into a plurality of event periods according to the recording period of the power load data required by the predicted event, identifying a data missing section in the event periods, marking the event period with the data missing section as a investigation period and marking the event period without the data missing section as a safety period; the power load data classification module is used for carrying out data identification classification with the missing reason as a target; the response filling demand model construction module is used for acquiring the missing duration of the data missing segment record where each type of missing power load data is located and constructing a response filling demand model; the data type dividing module is used for outputting a safe power load data type and a missing power data type to be analyzed; The scene characteristic output module is used for summarizing the missing duration and the type of the time period in the investigation period and the specific missing reason, and determining the scene characteristics of the power load monitoring; The filling data set generation module is used for analyzing and outputting filling data sets of the missing power data to be analyzed of each type; And the real-time response filling module is used for judging the type of the power load data when the real-time data missing segment is identified in response, and matching the filling data set based on the judging result to respond and fill.
6. The power load prediction analysis system based on big data as set forth in claim 5, wherein the response filling demand model construction module comprises an output data completion ratio calculation unit, an output data stability ratio calculation unit, a prediction accuracy value calculation unit, a response filling demand function analysis unit and a model construction unit; the output data completion ratio calculation unit is used for calculating the output data completion ratio of the corresponding type of missing power load data under a prediction event; The output data stability ratio calculation unit is used for calculating the output data stability ratio of the corresponding type of missing power load data under a prediction event; the prediction precision value calculation unit is used for extracting a prediction result value and an actual load value of a prediction model of the same type of missing power load data in each prediction event and calculating a prediction precision value; The response filling demand function analysis unit is used for constructing a response filling demand function by taking the output data completion ratio and the output data stability ratio of the same type of missing power load data records as input variables and taking a prediction precision value corresponding to the same prediction event as an output variable; The model construction unit is used for setting a prediction precision value threshold value and constructing a response filling demand model.
7. The system for predictive analysis of electrical loads based on big data as set forth in claim 6, wherein said data type partitioning module comprises a safe electrical load data type marking unit and a missing electrical data type marking unit to be analyzed; The safety power load data type marking unit is used for traversing and searching the prediction precision values of all the prediction events of the various types of missing power load data records, and if the prediction precision values of all the prediction events are larger than a threshold value, marking the corresponding type of missing power load data as the safety power load data type; The to-be-analyzed missing power data type marking unit is used for marking the to-be-analyzed missing power data type when the judging condition of the safety power load data type marking unit is not met.

Description

Power load prediction analysis system and method based on big data Technical Field The invention relates to the technical field of power load analysis, in particular to a power load prediction analysis system and method based on big data. Background In the field of power load prediction analysis, the traditional technology depends on a complete historical power load data support prediction model (such as ARIMA model and artificial neural network) to operate, but has obvious defects in practical application, the power load data acquisition process is easily interfered by multiple factors, so that data is lost, and the lost data can form a data missing section to directly influence the data integrity. In the prior art, the data deletion type and the deletion scene are not finely distinguished, and only the data is filled in a single mode, so that filling requirements under the characteristics of different deletion reasons, deletion time length and deletion time period cannot be adapted, and the large deviation between the filling data and the actual load is easily caused. The lack of a quantitative evaluation mechanism for influence on data loss can not judge whether the data loss can reduce the prediction precision, and the situations of blind operation without filling and untimely response without filling can occur, so that the prediction precision of the power load is low, the reliability is poor, and the actual requirements of power dispatching and supply and demand balance are difficult to meet. Disclosure of Invention The invention aims to provide a power load prediction analysis system and method based on big data, which are used for solving the problems in the prior art. In order to achieve the purpose, the invention provides the following technical scheme that the power load prediction analysis method based on big data comprises the following steps: Step S100, counting historical recorded power load data and predicted events in each monitoring area, wherein the predicted events refer to events of executing power prediction targets based on the power load data in a time period, dividing the historical data into a plurality of event periods according to the recording period of the power load data required by the predicted events, identifying data missing sections in the event periods, marking the event periods with the data missing sections as investigation periods and the event periods without the data missing sections as safety periods, extracting missing power load data of the data missing sections in all investigation periods, and carrying out data identification and classification with missing reasons as targets; Step 200, acquiring the missing duration of a data missing segment record where each type of missing power load data is located, constructing a response filling demand model, and outputting a safe power load data type and a missing power data type to be analyzed; Step 300, extracting a missing period set of a missing power data history record to be analyzed and a investigation period, summarizing missing time length and a type of the period in the investigation period and specific missing reasons, and determining scene characteristics of power load monitoring; and step 400, judging the type of the power load data when the response identifies the real-time data missing segment, and matching the filling data set based on the judging result to respond and fill. Further, the step S100 includes the following specific steps: Extracting the minimum node interval duration and the average node interval duration recorded by the adjacent power load data acquisition time nodes in each event period, and outputting a section corresponding to the node interval duration as a data missing section when the difference between the recorded node interval duration and the minimum node interval duration is larger than a first difference threshold and larger than the average node interval duration; The method comprises the steps of obtaining the missing reasons of missing power load data records in a data missing section, wherein the missing reasons comprise equipment faults, environmental interference and manual operation, marking the missing power load data with the same missing reasons recorded in the same monitoring area as one type of data, and distinguishing the missing power load data recorded in different events under the same type of data by taking a prediction event as a marking index. The classification aims at determining data division under different reasons causing data loss, so that a data analysis basis is made for intelligent filling. Further, the step S200 includes the following specific steps: Step S210, acquiring the average data record quantity Q of various types of missing power load data in a monitoring area in unit time length and the total unit time length T of the type of missing power load data recorded in each investigation period, extracting the actual load data storage quantity Q of