Power abnormal data cleaning method in photovoltaic power prediction system

By performing time interval normalization and least squares fitting of power data in the photovoltaic power prediction system, the problem of abnormal data cleaning in the photovoltaic power station is solved, and fast and reliable data cleaning is achieved, ensuring the normality of the photovoltaic power station run.

CN119066337APending Publication Date: 2024-12-03国网天津市电力公司高压分公司 +2
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411096906.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

There are abnormal data in the power prediction system of photovoltaic power stations, resulting in uncertainty in grid-connected power. The prior art is difficult to effectively clean these data, especially the normal mutations and abnormal data of photovoltaic power generation cannot be distinguished.

Method used

By assuming that the time interval of power discrete data is 1 to 6 minutes, after 10 consecutive power data normalization, the line or quadratic curve is fitted using the least squares method to calculate the error of each sampling point, and the labeling error is greater than 20%. The data is cleaned of suspicious data.

Benefits of technology

This method can quickly and reliably screen and clean abnormal data, ensure the normal grid-connected power data of the photovoltaic power station and ensure the normal operation of the power station.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure FDA0004988201910000011
    Figure FDA0004988201910000011
  • Figure FDA0004988201910000012
    Figure FDA0004988201910000012
Patent Text Reader

Abstract

The invention relates to a power abnormal data cleaning method in a photovoltaic power prediction system. The method comprises the following steps: S1, assuming that the time interval of power discrete data is T, the number of data is N, and the photovoltaic installed power is P; s2, normalizing the power data of continuous 10 points to obtain a sequence B; s3, if the normalized power data B of the power data at continuous 10 points is less than 0.2, directly ending without considering that abnormal data exist; and S4, fitting the normalized power data B of the continuous 10-point power data into a straight line and a quadratic curve by using a least square method, then calculating the error of each sampling point, if the error is less than 20%, determining that there is no abnormality, and if the error is greater than 20%, determining that the data is suspicious data and needs to be cleaned. The method is scientific and reasonable in design, is simple and reliable, can quickly complete the cleaning and screening of the abnormal data, guarantees the data of the grid-connected power of the photovoltaic power station to be normal, and guarantees the normal operation of the photovoltaic power station.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of regional photovoltaic power prediction in an electric power system, and in particular relates to a method for cleaning abnormal power data in a photovoltaic power prediction system. Background Art

[0002] In recent years, in order to achieve the goals of carbon peak and carbon neutrality, clean energy that supports carbon peak and carbon neutrality has been rapidly developed. Among clean energy, wind power and photovoltaics have developed most rapidly, and photovoltaics have been widely used due to their easy installation and wide application.

[0003] Photovoltaic power generation has certain intermittency and uncertainty, but the power grid has certain certainty requirements for the power of photovoltaic grid-connected. Therefore, photovoltaic power stations need to predict future power generation to reduce the uncertainty of grid-connected power.

[0004] The main energy source of photovoltaic power generation is sunlight, and the output power of photovoltaic power generation is closely related to solar radiation. The current prediction method for photovoltaic power generation mainly uses weather forecast data such as solar radiation intensity to estimate the energy that photovoltaic modules can receive, and calculates the grid-connected power of photovoltaic power generation according to the corresponding physical laws.

[0005] The photovoltaic power station grid-connected power prediction system is based on artificial intelligence training based on existing data. The power data in the existing data may generate abnormal data due to communication transmission and other reasons, so these abnormal data need to be cleaned out. At the same time, there may also be a sudden change in photovoltaic power generation, so the regular photovoltaic power sudden change should not be cleared out during data cleaning.

[0006] Therefore, a method is needed to clean the abnormal data. Summary of the invention

[0007] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method for cleaning abnormal power data in a photovoltaic power prediction system.

[0008] The present invention solves the technical problem by the following technical solutions:

[0009] A method for cleaning abnormal power data in a photovoltaic power prediction system, the method comprising the following steps:

[0010] S1. Assume that the time interval of power discrete data is T, the number of data is N, and the N photovoltaic power data are expressed as: A(0), A(T), A(2T), ..., A(NT-T). Assume that the photovoltaic installed power is P;

[0011] S2, normalize the 10 consecutive power data to obtain sequence B, expressed as B(KT)=A(KT) / P;

[0012] S3. If the normalized power data B is less than 0.2 for 10 consecutive points, it is not considered abnormal data and ends directly;

[0013] S4. Fit the normalized power data B of 10 consecutive points into a straight line y=aT+b using the least squares method; calculate the error of each sampling point at the same time. If the error between the real-time data of all points and the power of the fitted straight line is less than 20%, it is considered that there is no abnormality; wherein,

[0014] a) The power of the k-point fitting straight line is akT+b;

[0015] b) The real-time data of point k is B(kT);

[0016] c) The error is

[0017] S5. If any data in S4 is greater than 20%, the power data B after normalization for 10 consecutive points is fitted into a quadratic curve y=aT using the least squares method. 2 +bT+c; calculate the error of each sampling point at the same time. If the error between the real-time data of all points and the fitted linear power is less than 20%, it is considered that there is no abnormality; among them,

[0018] a) The power of the k-point fitting straight line is a(kT) 2 +bkT+c;

[0019] b) The real-time data of point k is B(kT);

[0020] c) The error is

[0021] S6. If any data in S5 is greater than 20%, it is recorded as the error of point m being greater than 20%, and the data of point m is marked as suspicious data and needs to be cleaned.

[0022] Moreover, the value of T in S1 is 1 to 6 minutes.

[0023] The advantages and beneficial effects of the present invention are:

[0024] The method of the present invention is simple and reliable, and can quickly complete the cleaning and screening of abnormal data, ensure that the data of the photovoltaic power station grid-connected power is normal, and ensure the normal operation of the photovoltaic power station. DETAILED DESCRIPTION

[0025] The present invention is further described in detail below through specific examples. The following examples are only illustrative and not restrictive, and the protection scope of the present invention cannot be limited thereto.

[0026] A method for cleaning abnormal power data in a photovoltaic power prediction system, the innovation of which lies in that the method comprises the following steps:

[0027] S1. Assume that the time interval of power discrete data is T, which is 1 to 6 minutes, the number of data is N, and the N photovoltaic power data are expressed as: A(0), A(T), A(2T), ..., A(NT-T). Assume that the photovoltaic installed power is P;

[0028] S2, normalize the 10 consecutive power data to obtain sequence B, expressed as B(KT)=A(KT) / P;

[0029] S3. If the normalized power data B is less than 0.2 for 10 consecutive points, it is not considered abnormal data and ends directly;

[0030] S4. Fit the normalized power data B of 10 consecutive points into a straight line y=aT+b using the least squares method; calculate the error of each sampling point at the same time. If the error between the real-time data of all points and the power of the fitted straight line is less than 20%, it is considered that there is no abnormality; wherein,

[0031] d) The power of the k-point fitting straight line is akT+b;

[0032] e) The real-time data of point k is B(kT);

[0033] f) The error is

[0034] S5. If any data in S4 is greater than 20%, the power data B after normalization for 10 consecutive points is fitted into a quadratic curve y=aT using the least squares method. 2 +bT+c; calculate the error of each sampling point at the same time. If the error between the real-time data of all points and the fitted linear power is less than 20%, it is considered that there is no abnormality; among them,

[0035] a) The power of the k-point fitting straight line is a(kT) 2 +bkT+c;

[0036] b) The real-time data of point k is B(kT);

[0037] c) The error is

[0038] S6. If any data in S5 is greater than 20%, it is recorded as the error of point m being greater than 20%, and the data of point m is marked as suspicious data and needs to be cleaned.

[0039] Although the embodiments of the present invention are disclosed for illustrative purposes, those skilled in the art will appreciate that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments.

Claims

1. A method for cleaning abnormal power data in a photovoltaic power prediction system. Features: The method comprises the following steps: S1. Assume that the time interval of power discrete data is T, the number of data is N, and the N photovoltaic power data are expressed as: A(0), A(T), A(2T), ..., A(NT-T). Assume that the photovoltaic installed power is P; S2, normalize the 10 consecutive power data to obtain sequence B, expressed as B(KT)=A(KT) / P; S3. If the normalized power data B is less than 0.2 for 10 consecutive points, it is not considered abnormal data and ends directly; S4. Fit the normalized power data B of 10 consecutive points into a straight line y=aT+b using the least squares method; calculate the error of each sampling point at the same time. If the error between the real-time data of all points and the power of the fitted straight line is less than 20%, it is considered that there is no abnormality; wherein, a) The power of the k-point fitting straight line is akT+b; b) The real-time data of point k is B(kT); c) The error is S5. If any data in S4 is greater than 20%, the power data B after normalization for 10 consecutive points is fitted into a quadratic curve y=aT using the least squares method. 2 +bT+c; calculate the error of each sampling point at the same time. If the error between the real-time data of all points and the fitted linear power is less than 20%, it is considered that there is no abnormality; among them, a) The power of the k-point fitting straight line is a(kT) 2 +bkT+c; b) The real-time data of point k is B(kT); c) The error is S6. If any data in S5 is greater than 20%, it is recorded as the error of point m being greater than 20%, and the data of point m is marked as suspicious data and needs to be cleaned.

2. The power abnormality data cleaning method in the photovoltaic power prediction system according to claim 1, Features: The T value in S1 is 1 to 6 minutes.

Citation Information

Cited By

  • Linear power networking distributed optical storage system and implementation method thereof

    CN120414657A

  • Ultra-short-term photovoltaic power prediction method and system based on golden day physical calibration

    CN122869308A