96-point electric quantity multi-source fusion fitting and anomaly correction method for new energy power plant

Abstract

The invention discloses a 96-point electric quantity multi-source fusion fitting and anomaly correction method for a new energy power plant, and belongs to the field of electric quantity metering and data processing of a new energy power system. The multi-source data fusion sensing module adopts a three-layer structure of a data access layer, a data fusion layer and a sensing analysis layer, the data access layer synchronously accesses acquisition side data, scheduling side data and marketing side data, the data fusion layer performs association and time alignment on multi-source data, and the sensing analysis layer performs sensing analysis on the multi-source data; the perceptual analysis layer constructs an algorithm model based on the fusion data; executing an adaptive fitting algorithm: on the basis of the multi-source fusion data, automatically complementing decimal digits of the electric energy meter and acquiring missing data for back-stepping fitting to obtain 96-point electric quantity fitting data; and carrying out automatic verification and abnormity correction on the collected fitting data: judging whether an abnormity is triggered through cross-side data comparison, if the abnormity is triggered, automatically returning to a data fitting link for reprocessing, and pushing the data to a marketing side until the data meets requirements.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electric quantity measurement and data processing of new energy power system, and particularly relates to a method of fusing multi-source data of dispatching side, collection side and marketing side, dynamically fitting 96-point electric quantity of Qinghai new energy power plant, and realizing abnormal adaptive correction. BACKGROUND

[0002] In the new energy power system, the 96-point electric quantity data of Qinghai new energy power plant (such as centralized photovoltaic and wind power plant) is the key basis for electricity settlement and power grid dispatching. At present, the fitting of 96-point electric quantity in the industry mainly relies on the original meter data of the collection side, without fully considering the special scene and multi-dimensional influencing factors of Qinghai new energy power plant, resulting in significant deviation between the fitted data and the actual power generation, and low efficiency of abnormal processing.

[0003] Specifically, the existing technology has the following defects: firstly, the loss allocation logic of multiple power plants and the influence of mutual inductor ratio are not included, and the characteristics of Qinghai new energy power plant output being affected by light and wind speed and fluctuating dramatically are ignored, resulting in insufficient accuracy of electric quantity fitting results; secondly, the identification of electric quantity data abnormality caused by sudden change of new energy output lags behind, and the abnormal processing adopts a "passive response mode", which needs to be completed manually after the abnormality is found, not only consuming time and effort, but also possibly delaying the progress of electricity settlement.

[0004] In the existing related technology, such as patent CN118779826A, only the time scale of multi-source data of Qinghai power system is unified, without limiting the scene of new energy power plant, and there is no 96-point electric quantity fitting and abnormal correction mechanism; patent CN117633431A focuses on fitting 96-point curve of 24-point load of power users, without involving electric quantity calculation and multi-source data fusion of new energy power plant; patents CN118604717B and CN118818414A only focus on single-body abnormal detection and correction of electric energy meter, without considering multi-source data cooperation and new energy output fluctuation and loss allocation and other systematic factors. In summary, the existing technology cannot solve the core problems of large deviation of Qinghai new energy power plant 96-point electric quantity fitting and passive abnormal processing, and an all-link intelligent technical solution is urgently needed. SUMMARY

[0005] The present application aims to solve the technical problems of "large data fitting deviation, late abnormal identification and dependence on manual processing" in the existing Qinghai new energy power plant 96-point electric quantity fitting, and to realize deep fusion of multi-source data, dynamic accurate fitting of electric quantity and adaptive correction of abnormality, and to upgrade the management mode from "passive manual processing" to "active intelligent correction".

[0006] To achieve the above-mentioned purpose, the present application provides the following steps: Step One: Constructing a multi-source data fusion perception module: The multi-source data fusion perception module adopts a "data access layer - data fusion layer - perception analysis layer" three-layer structure. The data access layer synchronously accesses the collection side data, the dispatching side data, and the marketing side data. The data fusion layer correlates and time-aligns the multi-source data. The perception analysis layer constructs an algorithm model based on the fused data. Step Two: Executing an adaptive fitting algorithm: Based on the multi-source fused data, the 96-point power fitting data is obtained through automatic decimal complement of the electric energy meter and reverse fitting of missing data. Step Three: Implementing automatic verification and abnormal correction of the collected fitting data: Whether an exception is triggered is determined through cross-side data comparison. If an exception is triggered, the data fitting link is automatically rolled back for reprocessing until the data meets the requirements and is pushed to the marketing side.

[0007] In Step One, the collection side data includes original meter collection data, electric energy meter decimal digit configuration, and mutual inductor comprehensive multiplier information. The dispatching side data includes 96-point output curve and real-time load data. The marketing side data includes electricity billing formula and customer mutual inductor and electric energy meter profile information.

[0008] In Step One, the correlation processing of the data fusion layer includes binding "electric energy meter original data - mutual inductor comprehensive multiplier - electric energy meter decimal digit configuration" and aligning "dispatching side 96-point output curve and real-time load data - collection side original meter data" according to a 15-minute / point time scale.

[0009] In Step Two, the rule for automatic complement of the electric energy meter decimal digit is: when the mutual inductor comprehensive multiplier is <100000, complement to 4 digits; when 100000≤multiplier<1000000>, complement to 5 digits; when 1000000≤multiplier<1000000>0, complement to 6 digits; and when the multiplier≥10000000, complement to 7 digits.

[0010] In Step Two, the trigger condition for reverse fitting of missing data is that the collection side continuous missing number is≥50 points. The fitting method is: through a "load - power" mapping model, the dispatching side real-time load data is converted into power data to fill in the 96-point power fitting value of the missing period.

[0011] In Step Three, the cross-side data comparison method is: the dispatching side real-time load data is converted into power data through a load-power conversion algorithm, and is compared with the marketing side billing data based on the collected fitting + decimal complement data point by point.

[0012] In Step Three, the condition for triggering an exception is that there are 10 or more time points in the day where the "converted power and settlement data difference exceeds 10%"; the method for correcting the exception is to automatically revert to the "data fitting link" and re-execute the "electricity meter decimal complement, missing data back-propagation fitting" process until the difference meets the requirements The specific steps are as follows: 1. Construct a multi-source data fusion perception module The multi-source data fusion perception module adopts a three-layer structure of "data access layer - data fusion layer - perception analysis layer", which is as follows: Data access layer: synchronously access three types of data. The collection side data includes original meter collection data, electricity meter decimal configuration, and mutual inductor comprehensive multiplier information; the dispatching side data includes 96-point output curve and real-time load data; the marketing side data includes electricity settlement formula and customer mutual inductor and electricity meter profile information.

[0013] Data fusion layer: correlate and process multi-source data. Bind "electricity meter original data - mutual inductor comprehensive multiplier - electricity meter decimal configuration" to ensure that the metering parameters match; align "dispatching side 96-point output curve and real-time load data - collection side original meter data" according to the time axis and unify the time scale (accurate to 15 minutes / point, matching 96-point time sequence).

[0014] Perception analysis layer: build an algorithm model based on the fused data to provide a data basis for subsequent adaptive fitting and anomaly recognition.

[0015] 2. Execute the adaptive fitting algorithm Based on multi-source fusion data, the adaptive fitting algorithm is used to achieve accurate fitting of 96-point power data, including the following sub-steps: Electricity meter decimal automatic completion: dynamically configure the decimal based on the mutual inductor comprehensive multiplier. When the mutual inductor comprehensive multiplier is <100000, complete the decimal to 4 digits; when 100000≤multiplier<1000000>, complete to 5 digits; when 1000000≤multiplier<1000000>0, complete to 6 digits; when the multiplier is ≥10000000, complete to 7 digits; for the case of "having load but the electricity meter does not move", use the intermediate difference method to calculate and complete the decimal data of the intermediate time point based on the trend of adjacent valid data points.

[0016] Collection missing data back-propagation fitting: when the collection side continuous defect number is greater than or equal to 50 points, the back-propagation fitting process is started. The "load-power" mapping model is constructed, the real-time load data of the dispatching side is converted into power data by the load-power conversion algorithm, and the 96-point power fitting value of the collection missing period is filled to ensure the data continuity.

[0017] 3. Implementing collection fitting data automatic checking and abnormal correction Cross-side data comparison: in the power audit link, the real-time load data of the dispatching side is converted into power data by the load-power conversion algorithm, and the 96-point point-by-point comparison is carried out with the settlement data (based on the collection fitting + decimal complemented data) of the marketing side.

[0018] Abnormal triggering and correction: if the "converted power and settlement data difference exceeds 10% of the time points" reaches 10 or more within the whole day, the power abnormality prompt is triggered immediately; the system automatically reverts to the "data fitting link", and the "electric energy meter decimal complement, collection missing data back-propagation fitting" process is re-executed; after the correction is completed, the cross-side data comparison is executed again until the difference meets the requirements (the time points with the difference exceeding 10% in the whole day is less than 10), and then the data is pushed to the marketing side for 96-point time settlement. Beneficial effects

[0019] Improve the power fitting accuracy: by fusing the multi-source data of the dispatching side, the collection side and the marketing side, the mutual inductor ratio, the loss allocation (reflected by the marketing side settlement formula) and the new energy output fluctuation characteristics are taken into account, the fitting deviation problem caused by the existing technology which only relies on the collection side data is solved, and the 96-point power fitting error can be reduced to within 10%.

[0020] Realize abnormal adaptive processing: an automatic checking and abnormal correction mechanism is constructed, abnormal identification, process rollback and data re-fitting can be completed without manual intervention, the abnormal processing response time is shortened from "hour level" to "minute level", and manual processing delay is avoided.

[0021] Adapt to the Qinghai new energy scene: according to the characteristics of the Qinghai new energy power plant output fluctuation, through the "load-power" mapping model and the differential decimal complement rule, the data fitting accuracy under extreme conditions is ensured, and compared with the general power measurement scheme, the scene adaptability is improved by more than 40%. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 It is a logic block diagram of the present application. DETAILED DESCRIPTION

[0023] Taking a 100MW centralized photovoltaic power station in Qinghai as an example 1, the specific implementation process of the present application is described in detail: 1. Multi-source data access and fusion Data access: Collect side accesses the original collection data (sampling interval 15 minutes) of 30 electric energy meters in the power plant, the initial configuration of the decimal places of the electric energy meter (default 2 bits), and the comprehensive ratio of the mutual inductor (80000); the dispatching side accesses the 96-point output curve (unit: MW) of the power plant on the same day and the real-time load data (sampling interval 5 minutes); the marketing side accesses the electric power settlement formula of the power plant (settlement power = collection power × mutual inductor ratio + line loss compensation) and the customer mutual inductor and electric energy meter archives (confirming that the mutual inductor ratio is correct).

[0024] Data fusion: binding “electric energy meter original data - ratio 80000 - decimal place configuration”, aligning the 96-point output curve of the dispatching side with the 15-minute interval original data of the collection side on the time axis, eliminating data points with timestamp deviation exceeding 1 minute, and ensuring time sequence uniformity.

[0025] 2. Self-adaptive fitting execution Decimal place completion: because the mutual inductor ratio is 80000 < 100000, the decimal places of the electric energy meter are completed from 2 bits to 4 bits; among them, during the period of 10:15-10:30 on the same day, “there is load (output 50MW) but the electric energy meter does not move”, based on the effective data of 9:45 (power 12.5000kWh) and 10:45 (power 13.2000kWh), the decimal place data of 10:15 (12.7750kWh) and 10:30 (13.0000kWh) are calculated and completed by the intermediate difference method.

[0026] Missing data fitting: during 14:00-16:00 on the same day, 48 consecutive data points are missing due to collection equipment failure (not reaching 50 points, no need to reverse), and 60 consecutive data points are missing from 16:00-18:30, starting reverse fitting: through the “load-power” mapping model (power = load × 0.25h, because the sampling interval is 15 minutes), the load data of the dispatching side at 16:00 (45MW) and 16:15 (42MW) are converted into power of 11.2500kWh and 10.5000kWh, respectively, to fill in the missing points.

[0027] 3. Automatic verification and abnormal correction Cross-side comparison: compare the 96-point power converted from the load of the dispatching side (such as 10:15 for 12.5000kWh) with the settlement data of the marketing side (completed as 12.7750kWh) point by point, and count the time points with a difference exceeding 10%.

[0028] Abnormality treatment: 12 time points with more than 10% difference were found in the first comparison, triggering an abnormality reminder; the system automatically rolled back to the fitting link, rechecked the mutual inductor ratio (confirmed to be correct), and completed the decimal place (corrected 1 mis-supplemented data point); after the second comparison, the time points with more than 10% difference decreased to 8, meeting the requirements, and the data was pushed to the marketing side for time settlement.

[0029] The three-layer structure block diagram of the multi-source data fusion perception module in FIG. 1; the diagram includes a data access layer (data interface of the collection side, the scheduling side, and the marketing side), a data fusion layer (a data correlation unit and a time alignment unit), and a perception analysis layer (an algorithm model construction unit), and each layer is connected through a data transmission link.

Claims

1. A method for multi-source fusion fitting and anomaly correction of 96 power points in a new energy power plant, characterized in that, Includes the following steps: Step 1: Constructing a multi-source data fusion perception module: The multi-source data fusion perception module adopts a three-layer structure of "data access layer - data fusion layer - perception and analysis layer". The data access layer synchronously accesses data from the collection side, scheduling side, and marketing side. The data fusion layer correlates and aligns the multi-source data in time. The perception and analysis layer builds an algorithm model based on the fused data. Step 2: Execute the adaptive fitting algorithm: Based on multi-source fusion data, the algorithm automatically completes the decimal places of the electricity meter and back-fits the missing data to obtain 96 points of electricity fitting data; Step 3: Implement automatic verification and anomaly correction of collected and fitted data: Determine whether an anomaly is triggered by cross-side data comparison. If an anomaly is triggered, automatically revert to the data fitting stage for reprocessing until the data meets the requirements before pushing it to the marketing side.

2. The method according to claim 1, characterized in that, In step one, the data collected on the acquisition side includes raw meter data, the decimal place configuration of the electricity meter, and the comprehensive ratio information of the current transformer; the data dispatched on the dispatch side includes 96-point output curves and real-time load data; and the data marketing on the marketing side includes electricity billing formulas and customer current transformer and electricity meter file information.

3. The method according to claim 1, characterized in that, In step one, the association processing of the data fusion layer includes: binding "original data of electricity meter - comprehensive ratio of current transformer - configuration of decimal places of electricity meter" and aligning "output curves and real-time load data of 96 points on the dispatch side - original meter data on the acquisition side" according to a time scale of 15 minutes / point.

4. The method according to claim 1, characterized in that, In step two, the rules for automatically filling in the decimal places of the electricity meter are as follows: when the comprehensive transformer ratio is less than 100,000, the decimal places are filled to 4 digits; when 100,000 ≤ ratio < 1000,000, the decimal places are filled to 5 digits; when 1000,000 ≤ ratio < 1000,000, the decimal places are filled to 6 digits; and when the ratio is greater than or equal to 10,000,000, the decimal places are filled to 7 digits.

5. The method according to claim 1, characterized in that, In step two, the trigger condition for backfitting missing data is that the number of consecutive defects on the acquisition side is ≥50. The fitting method is to convert the real-time load data on the dispatch side into power data through the "load-power" mapping model and fill in the missing power fitting values ​​for 96 points during the missing period.

6. The method according to claim 1, characterized in that, In step three, the cross-side data comparison method is as follows: the real-time load data of the dispatch side is converted into power data through the load-power conversion algorithm, and then compared with the settlement data of the marketing side based on the collection fitting + decimal place completion in 96 points.

7. The method according to claim 1, characterized in that, In step three, the condition for triggering the anomaly is that there are 10 or more "time points in the day where the difference between the converted electricity and the settlement data exceeds 10%"; the anomaly correction method is that the system automatically rolls back to the "data fitting stage" and re-executes the "filling in the decimal places of the electricity meter and backfitting the missing data" process until the difference meets the requirements.

Images