CN106548270B - Photovoltaic power station power abnormity data identification method and device - Google Patents

Photovoltaic power station power abnormity data identification method and device Download PDF

Info

Publication number
CN106548270B
CN106548270B CN201610875514.0A CN201610875514A CN106548270B CN 106548270 B CN106548270 B CN 106548270B CN 201610875514 A CN201610875514 A CN 201610875514A CN 106548270 B CN106548270 B CN 106548270B
Authority
CN
China
Prior art keywords
power
data
abnormal
daily
photovoltaic power
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201610875514.0A
Other languages
Chinese (zh)
Other versions
CN106548270A (en
Inventor
陶颍军
邱俊宏
李贞�
孔波利
李现伟
陈斌
沈志广
崔丽艳
刘永华
张秀娟
戚振伟
王四伟
熊焰
陈强
王兆庆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
State Grid Corp of China SGCC
Xuji Group Co Ltd
XJ Electric Co Ltd
Xuchang XJ Software Technology Co Ltd
Original Assignee
State Grid Corp of China SGCC
Xuji Group Co Ltd
XJ Electric Co Ltd
Xuchang XJ Software Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by State Grid Corp of China SGCC, Xuji Group Co Ltd, XJ Electric Co Ltd, Xuchang XJ Software Technology Co Ltd filed Critical State Grid Corp of China SGCC
Priority to CN201610875514.0A priority Critical patent/CN106548270B/en
Publication of CN106548270A publication Critical patent/CN106548270A/en
Application granted granted Critical
Publication of CN106548270B publication Critical patent/CN106548270B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Development Economics (AREA)
  • Health & Medical Sciences (AREA)
  • Educational Administration (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Theoretical Computer Science (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Game Theory and Decision Science (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Photovoltaic Devices (AREA)

Abstract

The invention relates to a method and a device for identifying abnormal power data of a photovoltaic power station, wherein the method takes active power data of the photovoltaic power station as a research object, combines various influence factors of the power data of the photovoltaic power station and clusters a power curve by using a fuzzy clustering algorithm; and identifying the obvious abnormal data according to the abnormal data judgment criterion of the photovoltaic power station. The data set with the abnormal data removed is adopted to train the photovoltaic power prediction model, so that the prediction precision and the prediction efficiency of the photovoltaic power station can be effectively improved, and the method has a wide engineering application value.

Description

Photovoltaic power station power abnormity data identification method and device
Technical Field
The invention belongs to the technical field of new energy control, and particularly relates to a method and a device for identifying abnormal power data of a photovoltaic power station.
Background
The field acquisition of photovoltaic power data is the basis of work such as photovoltaic power generation amount analysis calculation, power prediction and the like, however, due to the fact that the photovoltaic power abnormal data are generated, such as communication abnormality, equipment failure, artificial power limitation and the like, the quality of the power data acquired by a plurality of photovoltaic power stations on the field is poor, the abnormal data can seriously affect the real rules of the factors of photovoltaic power extraction, irradiance and temperature, the accuracy and effectiveness of photovoltaic power prediction can be greatly reduced by directly utilizing the abnormal data, and adverse effects can also be generated on photovoltaic power station operation management and power grid scheduling.
Disclosure of Invention
The invention aims to provide a method and a device for identifying abnormal power data of a photovoltaic power station, so as to identify the abnormal power data of the photovoltaic power and improve the accuracy of power prediction.
In order to solve the technical problem, the invention provides a method for identifying abnormal power data of a photovoltaic power station, which comprises six method schemes:
the first method scheme comprises the following steps:
1) preprocessing the daily power data, and calculating the correlation degree of a daily power curve by combining the influence factors of the photovoltaic power station power data;
2) clustering the daily power curves by adopting a clustering algorithm according to the relevance of the daily power curves of different influence factors to obtain data classification results under different clustering numbers, namely obtaining characteristic curves under different influence factors;
3) according to an abnormal data judgment criterion of the power data of the photovoltaic power station, identifying abnormal data with obvious characteristics in daily power, wherein the abnormal data judgment criterion is as follows:
A) the photovoltaic power value is higher than the characteristic curve value within the continuous set time and does not change along with the irradiance;
B) the photovoltaic power is lower than the characteristic curve value within the continuous set time and does not change along with the irradiance;
C) the total irradiance is obviously not 0 in the continuous set time, and the photovoltaic power is kept at 0 or close to 0;
and if any one of the conditions occurs, judging the data to be abnormal data.
And a second method scheme, based on the first method scheme, the method further comprises the step of performing secondary abnormality identification on the data which are not judged to be abnormal by adopting a longitudinal method and a transverse method after the daily power data are judged according to an abnormal data judgment criterion.
In a third method, on the basis of the first method, the method for calculating the correlation degree is a gray prediction correlation degree method, and includes the following steps:
converting the influence factors of the power data into daily feature vectors as samples for calculating grey correlation degrees, wherein the correlation coefficient of each point is as follows:
Figure BDA0001125346680000021
wherein ξ (k) is the sequence x0And xiThe grey correlation coefficient at the point k is,
Figure BDA0001125346680000022
Figure BDA0001125346680000023
2-level minimum range and maximum range are respectively adopted, rho is a resolution coefficient, and 0.5 is selected between 0 and 1;
and (3) calculating the average value of the correlation coefficients:
Figure BDA0001125346680000024
on the basis of the first method scheme or the third method scheme, the influence factors of the photovoltaic power station power data comprise:
weather type:
Figure BDA0001125346680000025
season:
Figure BDA0001125346680000026
the highest temperature:
Figure BDA0001125346680000027
lowest temperature:
Figure BDA0001125346680000028
2. temperature at 8, 14, 20: t is tk＝tTime of day/tmax
Converting the influencing factors into daily feature vectors (d, s, t)max,tmin,t2,t8,t14,t20) As a sample for calculating the gray correlation.
And a fifth method scheme, wherein on the basis of the first method scheme, the clustering algorithm is a fuzzy mean clustering algorithm, and the power sample space is X ═ X1,x2,…,xnAnd (n is the number of input samples), the method comprises the following steps:
s1, giving the cluster category number c (c is more than or equal to 2 and less than or equal to n), setting an iteration stop threshold value according to needs, and setting the initialized cluster prototype mode as P(0)The iteration counter b is 0;
s2, dividing the matrix U according to the average value of the correlation coefficient(b): if any of k (k is 1,2, …, n) and i (i is 1,2, …, c)
Figure BDA0001125346680000031
Then there are:
Figure BDA0001125346680000032
where m > 1 (m is generally 2) is called a blurring coefficient, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
if i, k are present, such that
Figure BDA0001125346680000033
Then there is
Figure BDA0001125346680000034
And for j ≠ k,
Figure BDA0001125346680000035
s3, calculating a clustering prototype pattern matrix P(b+1)：
Figure BDA0001125346680000036
Wherein u isikPartition matrix calculated from correlation coefficient for the previous step, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
s4, judging P(b)-P(b+1)And the relationship: if P | |(b)-P(b+1)If | | > is greater than or equal to |, b ═ b +1, and the iteration process is repeated until | | | P(b)-P(b+1)I < I; where the iteration stop threshold is indicated.
In a sixth method, on the basis of the second method, the transverse method is used for identifying abnormal data according to the power change situation between s points of the power curve, and comprises the following steps:
q1, finding the power change rate of the ith point of the selected sample day:
Δli＝(li-li-1)/li
wherein,. DELTA.liIs the rate of change of power between 2 consecutive power points, i is the number of power points, and i is 1,2, …, s;
q2, calculating the average power change rate of the same moment in the previous n days:
Figure BDA0001125346680000037
wherein,. DELTA.li,avThe average power change rate of n sample data, i is the number of power points, and i is 1,2, …, s;
q3, if | Δ li|≥kΔli,avJudging the data to be abnormal data, wherein k is a power mutation coefficient;
the longitudinal method is characterized in that power data of n days are selected as samples, and power values at the same moment on a daily power curve are compared, and the longitudinal method comprises the following steps:
p1, regarding the daily power data of s points of n days as an array with the horizontal vector as s points and the vertical vector as n, finding the expected value of each point
Figure BDA0001125346680000041
Sum variance
Figure BDA0001125346680000042
P2 determining the offset ratio of each value in the two-dimensional array
Figure BDA0001125346680000043
Comparing the offset with a set threshold lambda, and if the offset rate is greater than lambda, determining abnormal data, wherein sigmajIs the variance at point j.
The invention also provides a photovoltaic power station power abnormal data identification device, which comprises six device schemes:
the device scheme I comprises the following units:
1) the unit is used for preprocessing the daily power data and calculating the correlation degree of the daily power curve by combining the influence factors of the photovoltaic power station power data;
2) a unit for clustering the daily power curves by using a clustering algorithm according to the association degrees of the daily power curves of different influence factors to obtain data classification results under different clustering numbers, namely obtaining characteristic curves under different influence factors;
3) the device comprises a unit for identifying abnormal data with obvious characteristics in daily power according to an abnormal data judgment criterion of the power data of the photovoltaic power station, wherein the abnormal data judgment criterion is as follows:
A) the photovoltaic power value is higher than the characteristic curve value within the continuous set time and does not change along with the irradiance;
B) the photovoltaic power is lower than the characteristic curve value within the continuous set time and does not change along with the irradiance;
C) the total irradiance is obviously not 0 in the continuous set time, and the photovoltaic power is kept at 0 or close to 0;
and if any one of the conditions occurs, judging the data to be abnormal data.
And the device scheme II is based on the device scheme I and further comprises a unit for performing secondary abnormality identification on the data which is not judged to be abnormal by adopting a longitudinal method and a transverse method after the daily power data is judged according to an abnormal data judgment criterion.
In the third embodiment, on the basis of the first embodiment, the method for calculating the correlation degree is a gray prediction correlation degree method, and includes the following modules:
the module is used for converting the influence factors of the power data into daily feature vectors, the daily feature vectors are used as samples for calculating the grey correlation degree, and the correlation coefficient of each point is as follows:
Figure BDA0001125346680000051
wherein ξ (k) is the sequence x0And xiThe grey correlation coefficient at the point k is,
Figure BDA0001125346680000052
Figure BDA0001125346680000053
2-level minimum range and maximum range are respectively adopted, rho is a resolution coefficient, and 0.5 is selected between 0 and 1;
means for averaging the correlation coefficients:
Figure BDA0001125346680000054
and on the basis of the first device scheme or the third device scheme, the influence factors of the power data of the photovoltaic power station comprise:
weather type:
Figure BDA0001125346680000055
season:
Figure BDA0001125346680000056
the highest temperature:
Figure BDA0001125346680000057
lowest temperature:
Figure BDA0001125346680000058
2. temperature at 8, 14, 20: t is tk＝tTime of day/tmax
Converting the influencing factors into daily feature vectors (d, s, t)max,tmin,t2,t8,t14,t20) As a sample for calculating the gray correlation.
In a fifth embodiment, based on the first embodiment, the clustering algorithm is a fuzzy mean clustering algorithm, and the power sample space is X ═ X1,x2,…,xnAnd (n is the number of input samples), the method comprises the following modules:
s1, setting iteration stop threshold value according to requirement for given cluster category number c (c is more than or equal to 2 and less than or equal to n), and setting initialized cluster prototype mode as P(0)A module in which an iteration counter b is 0;
s2, dividing the matrix U according to the average value of the correlation coefficient(b)The module (2) comprises the following modules: if any of k (k is 1,2, …, n) and i (i is 1,2, …, c)
Figure BDA0001125346680000061
Then there are:
Figure BDA0001125346680000062
where m > 1 (m is generally 2) is called a blurring coefficient, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
if i, k are present, such that
Figure BDA0001125346680000063
Then there is
Figure BDA0001125346680000064
And for j ≠ k,
Figure BDA0001125346680000065
s3, calculating clustering prototype pattern matrix P(b+1)The module (2) comprises the following modules:
Figure BDA0001125346680000066
wherein u isikCalculated according to the correlation coefficient for the last stepPartition matrix of the calculation, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
s4 for judging P(b)-P(b+1) The module of | and relationship: if P | |(b)-P(b+1)If | | > is greater than or equal to |, b ═ b +1, and the iteration process is repeated until | | | P(b)-P(b+1)I < I; where the iteration stop threshold is indicated.
In a sixth apparatus scheme, on the basis of the second apparatus scheme, the transverse method is used for identifying abnormal data according to power change conditions between s points of a power curve, and includes the following modules:
q1, module for finding the power change rate of the ith point of the selected sample day:
Δli＝(li-li-1)/li
wherein,. DELTA.liIs the rate of change of power between 2 consecutive power points, i is the number of power points, and i is 1,2, …, s;
q2, a module for calculating the average power change rate of the same time of the previous n days:
Figure BDA0001125346680000067
wherein,. DELTA.li,avThe average power change rate of n sample data, i is the number of power points, and i is 1,2, …, s;
q3 for if | Δ li|≥kΔli,avJudging the data to be abnormal data, wherein k is a power mutation coefficient;
the longitudinal method is characterized in that power data of n days are selected as samples, power values at the same moment on a daily power curve are compared, and the longitudinal method comprises the following modules:
p1, for determining the expected power per point by considering the daily power data of s points of n days as an array with the horizontal vector as s points and the vertical vector as n
Figure BDA0001125346680000071
Sum variance
Figure BDA0001125346680000072
The module of (1);
p2 for finding the offset ratio of each value in a two-dimensional array
Figure BDA0001125346680000073
And the module compares the offset with a set threshold lambda, and if the offset rate is greater than lambda, the abnormal data is judged, wherein sigmajIs the variance at point j.
The invention has the beneficial effects that: according to the method, the power curve is clustered by considering the field environment of the photovoltaic power station and combining the influence factors of the power data of the photovoltaic power station, so that the accuracy and reliability of identifying the abnormal data of the photovoltaic power station are improved; considering that photovoltaic power data has strong randomness and dispersity, and the characteristics can influence power prediction and power generation analysis and calculation, most obvious abnormal data can be identified according to the characteristics and an abnormal data judgment criterion summarized by field practical experience, so that the accuracy and reliability of power abnormal data identification are improved, and the precision of power prediction is improved.
Drawings
FIG. 1 is a flow chart of a method for identifying abnormal power data of a photovoltaic power station according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
Fig. 1 shows a flow chart of a method for identifying abnormal power data of a photovoltaic power station, specifically:
1) and carrying out normalization preprocessing on the daily power data.
2) Considering the influence factors of the photovoltaic power station power data, including weather type, temperature and season:
weather type:
Figure BDA0001125346680000074
season:
Figure BDA0001125346680000075
the highest temperature:
Figure BDA0001125346680000081
lowest temperature:
Figure BDA0001125346680000082
2. temperature at 8, 14, 20: t is tk＝tTime of day/tmax
And calculating the association degree of the daily power curve according to a grey association coefficient method by combining the influence factors of the photovoltaic power station power data:
Figure BDA0001125346680000083
wherein ξ (k) is the sequence x0And xiThe grey correlation coefficient at the point k is,
Figure BDA0001125346680000084
Figure BDA0001125346680000085
2-level minimum range and maximum range respectively, rho is a resolution coefficient, and 0.5 is taken between 0 and 1.
Converting the influencing factors of the power data into day eigenvectors (d, s, t)max,tmin,t2,t8,t14,t20) As a sample for calculating the gray correlation degree.
And (3) calculating the average value of the correlation coefficients:
Figure BDA0001125346680000086
3) according to different influence factorsThe relevance of the daily power curve is clustered by adopting a fuzzy mean clustering algorithm to obtain data classification results under different clustering numbers, and the power sample space is X ═ { X ═1,x2,…,xn} (n is the number of input samples), specifically:
s1, giving the cluster category number c (c is more than or equal to 2 and less than or equal to n), setting an iteration stop threshold value according to needs, and setting the initialized cluster prototype mode as P(0)The iteration counter b is 0;
s2, dividing the matrix U according to the average value of the correlation coefficient(b): if any of k (k is 1,2, …, n) and i (i is 1,2, …, c)
Figure BDA0001125346680000087
Then there are:
Figure BDA0001125346680000088
where m > 1 (m is generally 2) is called a blurring coefficient, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
if i, k are present, such that
Figure BDA0001125346680000091
Then there is
Figure BDA0001125346680000092
And for j ≠ k,
Figure BDA0001125346680000093
s3, calculating a clustering prototype pattern matrix P(b+1)：
Figure BDA0001125346680000094
Wherein u isikPartition matrix calculated from correlation coefficient for the previous step, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
s4, judging P(b)-P(b+1)And the relationship: if P | |(b)-P(b+1)If | | > is greater than or equal to |, b ═ b +1, and the iteration process is repeated until | | | P(b)-P(b+1)I < I; where the iteration stop threshold is indicated.
4) According to an abnormal data judgment criterion aiming at the power data of the photovoltaic power station, identifying abnormal data with obvious characteristics in daily power, wherein the abnormal data judgment criterion is as follows:
A. the photovoltaic power value is higher than the characteristic curve value within the continuous set time and does not change along with the irradiance, and a larger value of the photovoltaic power data record always kept under the high irradiance is mainly generated because of communication or measuring equipment failure;
B. the photovoltaic power is lower than the characteristic curve value within the continuous set time and does not change along with the irradiance, and the generation reason comprises photovoltaic power limiting, communication or measuring equipment failure, so that the recorded value of the photovoltaic power is always kept on the power value under the lower irradiance;
C. the total irradiance is obviously not 0 in the duration setting time, the photovoltaic power is kept at 0 or close to 0, and the generation reason is that the data record is always kept at 0 or close to 0 due to communication, measurement equipment failure or photovoltaic assembly failure.
And if any one of the conditions occurs, judging the data to be abnormal data.
In this embodiment, a fuzzy mean clustering algorithm is used to perform clustering processing on the daily power curve, so as to obtain data classification results under different clustering numbers. As other embodiments, other clustering algorithms may also be employed to achieve the purpose of obtaining data classification results under different clustering numbers.
In the embodiment, the influence factors of the photovoltaic power station power data select the weather type, the temperature and the season. As other implementation modes, the influence factors of the power of the photovoltaic power station can be selected according to actual conditions, the influence factors can be correspondingly increased or decreased, and the atmospheric pressure, the wind speed and the like can be increased.
In the embodiment, the power curves are clustered by combining the influence factors of the power data of the photovoltaic power station, the photovoltaic power data have strong randomness and dispersity, the characteristics influence power prediction and power generation analysis and calculation, and most obvious abnormal data can be identified by combining the abnormal data judgment criterion summarized by field practical experience according to the characteristics. As another embodiment, after most of the obvious abnormal data are identified according to the above-mentioned abnormal data judgment criterion summarized according to the actual experience in the field, the abnormal data are secondarily identified according to the characteristics of the photovoltaic power curve, so as to better improve the accuracy.
Specifically, the abnormal data of the daily power curve is secondarily identified by adopting a longitudinal method and a transverse method.
The transverse method is used for identifying abnormal data according to power change conditions among 96 points of a power curve and identifying the abnormal data according to power change rates among the points, and comprises the following steps:
q1, finding the power change rate of the ith point of the selected sample day:
Δli＝(li-li-1)/li
wherein,. DELTA.liIs the rate of change of power between 2 consecutive power points, i is the number of power points, and i is 1,2, …, 96;
q2, calculating the average power change rate of the same moment in the previous n days:
Figure BDA0001125346680000101
wherein,. DELTA.li,avThe average power change rate of n sample data, i is the number of power points, and i is 1,2, …, 96;
q3, if | Δ li|≥kΔli,avAnd judging the data to be abnormal data, wherein k is a power mutation coefficient.
The longitudinal method is characterized in that power data of n days are selected as samples, and power values at the same moment on a daily power curve are compared, and the longitudinal method comprises the following steps:
p1, regarding the daily power data of 96 points in n days as an array with the horizontal vector as 96 points and the vertical vector as n, and finding the expected value of each point
Figure BDA0001125346680000102
Sum variance
Figure BDA0001125346680000103
P2 determining the offset ratio of each value in the two-dimensional array
Figure BDA0001125346680000104
Comparing with a set threshold lambda, and if the offset rate is greater than lambda, determining abnormal data, wherein sigmajIs the variance at point j.
The invention also provides a photovoltaic power station power abnormal data identification device, which comprises the following units:
1) the unit is used for preprocessing the daily power data and calculating the correlation degree of the daily power curve by combining the influence factors of the photovoltaic power station power data;
2) a unit for clustering the daily power curves by using a clustering algorithm according to the association degrees of the daily power curves of different influence factors to obtain data classification results under different clustering numbers, namely obtaining characteristic curves under different influence factors;
3) the device comprises a unit for identifying abnormal data with obvious characteristics in daily power according to an abnormal data judgment criterion of the power data of the photovoltaic power station, wherein the abnormal data judgment criterion is as follows:
A) the photovoltaic power value is higher than the characteristic curve value within the continuous set time and does not change along with the irradiance;
B) the photovoltaic power is lower than the characteristic curve value within the continuous set time and does not change along with the irradiance;
C) the total irradiance is obviously not 0 in the duration setting time, and the photovoltaic power is kept at 0 or close to 0;
and if any one of the conditions occurs, judging the data to be abnormal data.
The device is actually a computer solution based on the method flow of the invention, namely a software framework, and each unit is each processing process or program corresponding to the method flow. The apparatus will not be described in detail since the description of the above method is sufficiently clear and complete.

Claims (10)

1. A photovoltaic power station power anomaly data identification method is characterized by comprising the following steps:
1) preprocessing the daily power data, and calculating the correlation degree of a daily power curve by combining the influence factors of the photovoltaic power station power data;
2) clustering the daily power curves by adopting a clustering algorithm according to the relevance of the daily power curves of different influence factors to obtain data classification results under different clustering numbers, namely obtaining characteristic curves under different influence factors;
3) according to an abnormal data judgment criterion of the power data of the photovoltaic power station, identifying abnormal data with obvious characteristics in daily power, wherein the abnormal data judgment criterion is as follows:
A) the photovoltaic power value is higher than the characteristic curve value within the continuous set time and does not change along with the irradiance;
B) the photovoltaic power is lower than the characteristic curve value within the continuous set time and does not change along with the irradiance;
C) the total irradiance is obviously not 0 in the continuous set time, and the photovoltaic power is kept at 0 or close to 0;
if any one of the conditions occurs, judging the data to be abnormal data;
the clustering algorithm is a fuzzy mean clustering algorithm, and the power sample space is X ═ X1,x2,…,xnN is the number of input samples, and the method comprises the following steps:
s1, giving the cluster category number c, c is more than or equal to 2 and less than or equal to n, setting an iteration stop threshold value according to needs, and setting an initialized cluster prototype mode as P(0)The iteration counter b is 0;
s2, dividing the matrix U according to the average value of the correlation coefficient(b): for any k-1, 2, …, n, i-1, 2, …, c, if
Figure FDA0002225842200000011
Then there are:
Figure FDA0002225842200000012
where m > 1 is called the blur coefficient, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
if i, k are present, such that
Figure FDA0002225842200000013
Then there is
Figure FDA0002225842200000014
And for j ≠ k,
Figure FDA0002225842200000015
s3, calculating a clustering prototype pattern matrix P(b+1)：
Figure FDA0002225842200000016
Wherein u isikPartition matrix calculated from correlation coefficient for the previous step, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
s4, judging P(b)-P(b+1)And the relationship: if P | |(b)-P(b+1)If | | > is greater than or equal to |, b ═ b +1, and the iteration process is repeated until | | | P(b)-P(b+1)I < I; where the iteration stop threshold is indicated.
2. The method for identifying the abnormal power data of the photovoltaic power station as claimed in claim 1, further comprising the step of performing secondary abnormal identification on data which is not judged to be abnormal by adopting a longitudinal method and a transverse method after the daily power data is judged according to an abnormal data judgment criterion.
3. The method for identifying abnormal power data of a photovoltaic power plant as claimed in claim 1, wherein the method for calculating the correlation degree is a gray prediction correlation degree method, and comprises the following steps:
converting the influence factors of the power data into daily feature vectors as samples for calculating grey correlation degrees, wherein the correlation coefficient of each point is as follows:
Figure FDA0002225842200000021
wherein ξ (k) is the sequence x0And xiThe grey correlation coefficient at the point k is,
Figure FDA0002225842200000022
Figure FDA0002225842200000023
2 levels of minimum range and maximum range are respectively adopted, and rho is a resolution coefficient and is between 0 and 1;
and (3) calculating the average value of the correlation coefficients:
Figure FDA0002225842200000024
4. the method for identifying abnormal power data of photovoltaic power plants according to claim 1 or 3, wherein the influence factors of the power data of photovoltaic power plants include:
weather type:
Figure FDA0002225842200000025
season:
Figure FDA0002225842200000026
the highest temperature:
Figure FDA0002225842200000027
lowest temperature:
Figure FDA0002225842200000031
2. temperature at 8, 14, 20: t is tk＝tTime of day/tmax
Converting the influencing factors into daily feature vectors (d, s, t)max,tmin,t2,t8,t14,t20) As a sample for calculating the gray correlation.
5. The method for identifying abnormal power data of the photovoltaic power station as claimed in claim 2, wherein the transverse method is used for identifying abnormal data according to power change conditions between s points of a power curve, and comprises the following steps:
q1, finding the power change rate of the ith point of the selected sample day:
Δli＝(li-li-1)/li
wherein,. DELTA.liIs the rate of change of power between 2 consecutive power points, i is the number of power points, and i is 1,2, …, s;
q2, calculating the average power change rate of the same moment in the previous n days:
Figure FDA0002225842200000032
wherein,. DELTA.li,avThe average power change rate of n sample data, i is the number of power points, and i is 1,2, …, s;
q3, if | Δ li|≥kΔli,avJudging the data to be abnormal data, wherein k is a power mutation coefficient;
the longitudinal method is characterized in that power data of n days are selected as samples, and power values at the same moment on a daily power curve are compared, and the longitudinal method comprises the following steps:
p1, regarding the daily power data of s points of n days as an array with the horizontal vector as s points and the vertical vector as n, finding the expected value of each point
Figure FDA0002225842200000033
Sum variance
Figure FDA0002225842200000034
P2 determining the offset ratio of each value in the two-dimensional array
Figure FDA0002225842200000035
Comparing the offset with a set threshold lambda, and if the offset rate is greater than lambda, determining abnormal data, wherein sigmajIs the variance at point j.
6. The photovoltaic power station power anomaly data identification device is characterized by comprising the following units:
1) the unit is used for preprocessing the daily power data and calculating the correlation degree of the daily power curve by combining the influence factors of the photovoltaic power station power data;
2) a unit for clustering the daily power curves by using a clustering algorithm according to the association degrees of the daily power curves of different influence factors to obtain data classification results under different clustering numbers, namely obtaining characteristic curves under different influence factors;
3) the device comprises a unit for identifying abnormal data with obvious characteristics in daily power according to an abnormal data judgment criterion of the power data of the photovoltaic power station, wherein the abnormal data judgment criterion is as follows:
A) the photovoltaic power value is higher than the characteristic curve value within the continuous set time and does not change along with the irradiance;
B) the photovoltaic power is lower than the characteristic curve value within the continuous set time and does not change along with the irradiance;
C) the total irradiance is obviously not 0 in the continuous set time, and the photovoltaic power is kept at 0 or close to 0;
if any one of the conditions occurs, judging the data to be abnormal data;
the clustering algorithm is a fuzzy mean clustering algorithm, and the power sample space is X ═ X1,x2,…,xnN is the number of input samples, and comprises the following modules:
s1, setting the number of the cluster categories c2 and c as well as n as required, setting the iteration stop threshold value as P, and setting the initialized cluster prototype mode(0)A module in which an iteration counter b is 0;
s2, dividing the matrix U according to the average value of the correlation coefficient(b)The module (2) comprises the following modules: for any k-1, 2, …, n, i-1, 2, …, c, if
Figure FDA0002225842200000041
Then there are:
Figure FDA0002225842200000042
where m > 1 is called the blur coefficient, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
if i, k are present, such that
Figure FDA0002225842200000043
Then there is
Figure FDA0002225842200000044
And for j ≠ k,
Figure FDA0002225842200000045
s3, calculating clustering prototype pattern matrix P(b+1)The module (2) comprises the following modules:
Figure FDA0002225842200000046
wherein u isikPartition matrix calculated from correlation coefficient for the previous step, dikRepresenting the degree of association between the samples in the kth class and the typical sample center of the ith class;
s4 for judging P(b)-P(b+1)The module of | and relationship: if P | |(b)-P(b+1)If | | > is greater than or equal to |, b ═ b +1, and the iteration process is repeated until | | | P(b)-P(b+1)I < I; where the iteration stop threshold is indicated.
7. The device for identifying abnormal power data of a photovoltaic power station as claimed in claim 6, further comprising a unit for performing secondary abnormal identification on data which is not judged to be abnormal by a longitudinal method and a transverse method after the daily power data is judged according to an abnormal data judgment criterion.
8. The device for identifying abnormal power data of a photovoltaic power plant as claimed in claim 6, wherein the method for calculating the correlation degree is a grey prediction correlation degree method, and comprises the following modules:
the module is used for converting the influence factors of the power data into daily feature vectors, the daily feature vectors are used as samples for calculating the grey correlation degree, and the correlation coefficient of each point is as follows:
Figure FDA0002225842200000051
wherein ξ (k) is the sequence x0And xiThe grey correlation coefficient at the point k is,
Figure FDA0002225842200000052
Figure FDA0002225842200000053
2-level minimum range and maximum range, respectively, rho is resolution coefficient, between 0 and 1A (c) is added;
means for averaging the correlation coefficients:
Figure FDA0002225842200000054
9. the device for identifying abnormal power data of photovoltaic power plants according to claim 6 or 8, wherein the influence factors of the power data of photovoltaic power plants include:
weather type:
Figure FDA0002225842200000055
season:
Figure FDA0002225842200000056
the highest temperature:
Figure FDA0002225842200000057
lowest temperature:
Figure FDA0002225842200000058
2. temperature at 8, 14, 20: t is tk＝tTime of day/tmax
Converting the influencing factors into daily feature vectors (d, s, t)max,tmin,t2,t8,t14,t20) As a sample for calculating the gray correlation.
10. The device for identifying abnormal power data of the photovoltaic power station as claimed in claim 7, wherein the transverse method is used for identifying abnormal data according to power change conditions between s points of a power curve, and comprises the following modules:
q1, module for finding the power change rate of the ith point of the selected sample day:
Δli＝(li-li-1)/li
wherein,. DELTA.liIs the rate of change of power between 2 consecutive power points, i is the number of power points, and i is 1,2, …, s;
q2, a module for calculating the average power change rate of the same time of the previous n days:
Figure FDA0002225842200000061
wherein,. DELTA.li,avThe average power change rate of n sample data, i is the number of power points, and i is 1,2, …, s;
q3 for if | Δ li|≥kΔli,avJudging the data to be abnormal data, wherein k is a power mutation coefficient;
the longitudinal method is characterized in that power data of n days are selected as samples, power values at the same moment on a daily power curve are compared, and the longitudinal method comprises the following modules:
p1, for determining the expected power per point by considering the daily power data of s points of n days as an array with the horizontal vector as s points and the vertical vector as n
Figure FDA0002225842200000062
Sum variance
Figure FDA0002225842200000063
The module of (1);
p2 for finding the offset ratio of each value in a two-dimensional array
Figure FDA0002225842200000064
And the module compares the offset with a set threshold lambda, and if the offset rate is greater than lambda, the abnormal data is judged, wherein sigmajIs the variance at point j.
CN201610875514.0A 2016-09-30 2016-09-30 Photovoltaic power station power abnormity data identification method and device Active CN106548270B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610875514.0A CN106548270B (en) 2016-09-30 2016-09-30 Photovoltaic power station power abnormity data identification method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610875514.0A CN106548270B (en) 2016-09-30 2016-09-30 Photovoltaic power station power abnormity data identification method and device

Publications (2)

Publication Number Publication Date
CN106548270A CN106548270A (en) 2017-03-29
CN106548270B true CN106548270B (en) 2020-08-14

Family

ID=58368503

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610875514.0A Active CN106548270B (en) 2016-09-30 2016-09-30 Photovoltaic power station power abnormity data identification method and device

Country Status (1)

Country Link
CN (1) CN106548270B (en)

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108107312B (en) * 2017-12-04 2018-11-30 国网江苏省电力有限公司电力科学研究院 Non- register one's residence does not power off wrong wiring of electric energy meter detection device and method
CN109842371A (en) * 2019-03-19 2019-06-04 黎和平 A kind of method and apparatus positioning photovoltaic power generation exception
CN110164102B (en) * 2019-04-22 2021-05-25 创维互联（北京）新能源科技有限公司 Photovoltaic power station string abnormity alarm method and alarm device
CN110674864B (en) * 2019-09-20 2024-03-15 国网上海市电力公司 Wind power abnormal data identification method comprising synchronous phasor measurement device
CN110995153B (en) * 2019-12-18 2020-11-24 国网电子商务有限公司 Abnormal data detection method and device for photovoltaic power station and electronic equipment
CN111258787B (en) * 2020-01-07 2023-06-20 浙江零跑科技股份有限公司 Method for identifying abnormal NTC temperature sampling value based on battery pack
CN112816216B (en) * 2021-01-05 2022-08-16 三峡大学 Rolling bearing performance test bench and identification and correction method of abnormal test sample
CN113900370B (en) * 2021-09-30 2022-11-08 万帮数字能源股份有限公司 Time calibration method and time calibration device for photovoltaic system and photovoltaic system
CN115718901A (en) * 2022-11-15 2023-02-28 中国南方电网有限责任公司超高压输电公司广州局 Data processing method and device based on converter valve and computer equipment

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009169930A (en) * 2007-12-21 2009-07-30 Fuji Electric Systems Co Ltd Energy demand predicting device
KR20140018497A (en) * 2012-08-01 2014-02-13 한국전력공사 Prediction method of short-term wind speed and wind power and power supply line voltage prediction method therefore

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103390902B (en) * 2013-06-04 2015-04-29 国家电网公司 Photovoltaic power station super short term power prediction method based on least square method
CN104463349A (en) * 2014-11-11 2015-03-25 河海大学 Photovoltaic generated power prediction method based on multi-period comprehensive similar days
CN104881706B (en) * 2014-12-31 2018-05-25 天津弘源慧能科技有限公司 A kind of power-system short-term load forecasting method based on big data technology

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009169930A (en) * 2007-12-21 2009-07-30 Fuji Electric Systems Co Ltd Energy demand predicting device
KR20140018497A (en) * 2012-08-01 2014-02-13 한국전력공사 Prediction method of short-term wind speed and wind power and power supply line voltage prediction method therefore

Also Published As

Publication number Publication date
CN106548270A (en) 2017-03-29

Similar Documents

Publication Publication Date Title
CN106548270B (en) Photovoltaic power station power abnormity data identification method and device
CN110070226B (en) Photovoltaic power prediction method and system based on convolutional neural network and meta-learning
CN112257941B (en) Photovoltaic power station short-term power prediction method based on improved Bi-LSTM
CN106447098B (en) Photovoltaic ultra-short-term power prediction method and device
CN111199016A (en) DTW-based improved K-means daily load curve clustering method
CN113486078B (en) Distributed power distribution network operation monitoring method and system
CN111680820B (en) Distributed photovoltaic power station fault diagnosis method and device
CN109086527B (en) Practical equivalent modeling method based on running state of wind turbine generator
CN111625399A (en) Method and system for recovering metering data
CN114004139A (en) Photovoltaic power generation power prediction method
CN113344288B (en) Cascade hydropower station group water level prediction method and device and computer readable storage medium
CN106099932B (en) Day-ahead planning power flow analysis method considering uncertainty time-space correlation
CN112801332B (en) Short-term wind speed prediction method based on gray level co-occurrence matrix
CN111861023A (en) Statistical-based hybrid wind power prediction method and device
CN112784920A (en) Cloud-side-end-coordinated dual-anti-domain self-adaptive fault diagnosis method for rotating part
CN115115090A (en) Wind power short-term prediction method based on improved LSTM-CNN
CN105956708A (en) Grey correlation time sequence based short-term wind speed forecasting method
CN109615027B (en) Intelligent prediction method for extracting wind speed characteristics along high-speed railway
CN116702937A (en) Photovoltaic output day-ahead prediction method based on K-means mean value clustering and BP neural network optimization
CN116050666A (en) Photovoltaic power generation power prediction method for irradiation characteristic clustering
CN110020680B (en) PMU data classification method based on random matrix theory and fuzzy C-means clustering algorithm
CN108985563B (en) Electromechanical system service dynamic marking method based on self-organizing feature mapping
CN112508278A (en) Multi-connected system load prediction method based on evidence regression multi-model
CN113837096B (en) Rolling bearing fault diagnosis method based on GA random forest
CN113657687B (en) Power load prediction method based on feature engineering and multipath deep learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant