CN110909813A - Business abnormal electricity utilization detection method based on edge algorithm - Google Patents

Abstract

The invention discloses a business body abnormal electricity utilization detection method based on an edge algorithm, which comprises the following steps: acquiring power consumption data to be detected, determining the date of the power consumption data to be detected, and determining training set data according to the date of the power consumption data to be detected; carrying out edge detection on the training set data by using an edge detection algorithm to obtain a stable edge column; acquiring sample data according to the stable edge column and the training set data; establishing an abnormal rule through sample data to obtain upper and lower bound thresholds of the electricity consumption of the business body; and carrying out electricity utilization abnormity detection on electricity utilization data to be detected through upper and lower threshold values to obtain an electricity utilization abnormity result for the business. The invention fully utilizes the periodic characteristic of business day of the business by monitoring the electricity consumption of the business, avoids the problem of different direction difference between hours, can realize the alarm of abnormal electricity consumption, and is beneficial to the manager to carry out deviation analysis and abnormal judgment on the business.

Description

Business abnormal electricity utilization detection method based on edge algorithm

Technical Field

The invention belongs to the field of commercial body electricity utilization detection, and particularly relates to a commercial body abnormal electricity utilization detection method based on an edge algorithm.

Background

The electric characteristics of the business are influenced by service and equipment types, the loop is complex, the number of loops is large, the electricity utilization difference of the loops is large, and the business is provided with an independent equipment loop and a composite loop consisting of a plurality of equipment. According to previous research, the types of electricity used by businesses can be divided into: air conditioning, elevator, power, fire control, illumination etc. return circuit. According to the analysis of the loops, the differences of seasonality, working days and non-working days, whether fixed starting and stopping are carried out, whether stable power utilization is carried out and the like exist in some loop sequences, and therefore the situation that the detection scene of abnormal power utilization of commercial power utilization is complex is determined. The existing power utilization anomaly detection technology comprises curve fitting, namely, anomaly is judged according to residual deviation of fitting, which belongs to fluctuation anomaly and is not suitable for unstable power utilization time sequence detection, namely, the time sequence anomaly of variance is difficult to judge when encountering the anomaly; whether the abnormal condition exists is judged by searching similar curves according to the current data historical similar days, the calculation method of the similar days comprises the methods of KNN, DTW, clustering and the like, but the distance parameter and the clustering number are set more complicatedly, and the start-stop boundary can not be judged well for the loop with fixed start-stop, and some fluctuations in the fixed start-stop time sequence can be judged as abnormal conditions by mistake. Whereas LOF and random forests easily identify the change points in the time sequence as anomalies and are less explanatory. The accuracy of the current prior art commercial power utilization anomaly identification is not high.

Disclosure of Invention

Aiming at the defects in the prior art, the business body abnormal electricity utilization detection method based on the edge algorithm solves the problem that the accuracy of business body abnormal electricity utilization identification is low.

In order to achieve the purpose of the invention, the invention adopts the technical scheme that: a business body abnormal electricity utilization detection method based on an edge algorithm comprises the following steps:

s1, collecting power consumption data to be detected, determining the date of the power consumption data to be detected, and determining training set data according to the date of the power consumption data to be detected;

s2, carrying out edge detection on the training set data by using an edge detection algorithm to obtain a stable edge column;

s3, obtaining sample data according to the stable edge column and the training set data;

s4, establishing an abnormal rule through sample data to obtain upper and lower bound thresholds of the electricity consumption of the business;

and S5, carrying out abnormal electricity utilization detection on the electricity data to be detected through the upper and lower threshold values to obtain the detection result of the abnormal electricity utilization of the commercial object.

Further, the step S1 includes the following sub-steps:

s1.1, collecting power consumption data to be detected, and determining the date d and the month t of the power consumption data to be detected;

and S1.2, acquiring data according to the date d and the month t, and taking the acquired data as training set data.

Further, the specific method for acquiring data according to the date d and the month t in the step S1.2 is as follows: and acquiring the electricity utilization data from No. 1 to date d in the month t, the electricity utilization data of the last two months in the month t, the electricity utilization data of two months adjacent to the same month in the last year and the electricity utilization data of the same month in the last year.

Further, the step S2 includes the following sub-steps:

s2.1, normalizing the training set data to be in a range of [0,255], wherein the normalized calculation formula is as follows:

s2.2, expressing the normalized electric quantity by days, and converting the electric quantity into a 24-column two-dimensional hour matrix X _ hour (day, hour _ cols);

s2.3, performing edge detection on the two-dimensional hour matrix by using an edge detection algorithm, representing edge points in the two-dimensional hour matrix by 255, and representing non-edge points in the two-dimensional hour matrix by 0 to obtain an edge matrix Y _ hour;

s2.4, acquiring the proportion of the number of times of occurrence of edge points in each hour column vector to the total number of rows according to the edge matrix Y _ hour;

and S2.5, judging whether the proportion of the edge points in each hour column vector exceeds 55%, if so, marking the corresponding column as a stable edge column as 1, otherwise, marking the corresponding column as a non-edge column as 0.

Wherein x' represents a normalized result vector of the power data at a fixed hour time point, x represents a historical power vector occurring at a fixed hour, PARAM _ x_maxRepresenting a 0.99 quantile of electrical quantity x, PARAM _ x_minRepresenting the 0.01 quantile, y, of the quantity of electricity x_maxRepresenting the upper bound of the normalized range, y_minLower bound, x, representing the normalized range_minDenotes the minimum value of the electric quantity, x_maxRepresenting the maximum value of the electricity, day representing the number of days the electricity x is located, and hour _ cols representing the 24 columns corresponding to 24 hours in the two-dimensional hour matrix.

Further, the formula for calculating the ratio of the number of occurrences of each row of edge points to the number of days in step S2.4 is:

wherein, cnt_dayRepresents the number of days of edge points and the total number of days of non-edge points, cnt, in the edge matrix Y _ hour₀Indicates the number of days, cnt, of non-edge points₂₅₅Represents the number of days of occurrence of edge points, z _ i and

the number of times of the edge points in the ith hour corresponding column vector in the edge matrix Y _ hour is shown, and f (z _ i) represents the proportion of the edge points in the total days.

Further, the step S3 includes the following sub-steps:

s3.1, taking the data corresponding to the stable edge column for hours as sample data; the sample data sample_jComprises the following steps:

sample_j＝{train_j}

wherein, train_jData representing the corresponding stable edge column at the j hour;

s3.2, corresponding the non-edge columns to the hour and beforeData of next adjacent hour as sample data sample thereof_iSample data sample_iComprises the following steps:

sample_i＝{train_{i_before},train_i,train_{i_end}}；

wherein, train_iIndicating data corresponding to a non-edge column in the ith hour, wherein j belongs to {0,1,2,. 23 }, i belongs to {0,1,2,. 23, }, j is not equal to i, train_{i_before}Electricity data, train, representing the hour i _ before the ith hour_{i_end}Electricity consumption data representing an hour i _ end after the ith hour; when i is 0, i _ before is 23; when i is 23, i _ end is 0.

Further, the sub-steps of step S4 are:

s4.1, sample data sample is utilized_iAnd sample_jGenerating a first quartile q75_iAnd a third quartile q25_i；

S4.2, according to the first quartile q75_iAnd a third quartile q25_iThe upper and lower bound thresholds for generating the electricity for the business are:

wherein thresh _ max_iUpper threshold, thresh _ min, representing commercial power usage_iA lower threshold representing the electricity usage of the business.

Further, the step S5 includes the following sub-steps:

s5.1, judging whether the electricity data to be detected is larger than an upper bound threshold thresh _ max_iIf yes, abnormal power utilization is achieved, otherwise, the step S5.2 is carried out;

s5.2, judging whether the power consumption data is smaller than a lower bound threshold thresh _ min_iIf so, judging the power consumption is abnormal, otherwise, judging the power consumption is normal, and obtaining the detection result of the abnormal power consumption of the business body.

The invention has the beneficial effects that:

(1) the invention automatically identifies the position with large gradient change by utilizing edge detection, is beneficial to effectively identifying the time point position of starting and stopping the loop and assists in judging whether the loop starting time in the business body service is on time or not; for non-edge points, the front hour and the back hour have reference effects, so that samples in a single hour are increased by adopting the front data and the back data, and the effective boundary corresponding to the hour can be better judged, and further abnormity can be judged.

(2) The invention fully utilizes the periodic characteristic of business day of the business by monitoring the electricity consumption of the business, avoids the problem of different direction difference between hours, can realize the alarm of abnormal electricity consumption, and is beneficial to the manager to carry out deviation analysis and abnormal judgment on the business.

Drawings

Fig. 1 is a flowchart of a method for detecting abnormal electricity consumption of a business based on an edge algorithm according to the present invention.

Detailed Description

The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

As shown in fig. 1, a method for detecting abnormal electricity consumption of a business based on an edge algorithm includes the following steps:

s1, collecting power consumption data to be detected, determining the date of the power consumption data to be detected, and determining training set data according to the date of the power consumption data to be detected;

s2, carrying out edge detection on the training set data by using an edge detection algorithm to obtain a stable edge column;

s3, obtaining sample data according to the stable edge column and the training set data;

s4, establishing an abnormal rule through sample data to obtain upper and lower bound thresholds of the electricity consumption of the business;

and S5, carrying out abnormal electricity utilization detection on the electricity data to be detected through the upper and lower threshold values to obtain the detection result of the abnormal electricity utilization of the commercial object.

The step S1 includes the following sub-steps:

s1.1, collecting power consumption data to be detected, and determining the date d and the month t of the power consumption data to be detected;

and S1.2, acquiring data according to the date d and the month t, and taking the acquired data as training set data.

The specific method for acquiring data according to the date d and the month t in the step S1.2 is as follows: and acquiring the electricity utilization data from No. 1 to date d in the month t, the electricity utilization data of the last two months in the month t, the electricity utilization data of two months adjacent to the same month in the last year and the electricity utilization data of the same month in the last year.

The step S2 includes the following sub-steps:

s2.1, normalizing the training set data to be in a range of [0,255], wherein the normalized calculation formula is as follows:

s2.2, expressing the normalized electric quantity by days, and converting the electric quantity into a 24-column two-dimensional hour matrix X _ hour (day, hour _ cols);

s2.3, performing edge detection on the two-dimensional hour matrix by using an edge detection algorithm, representing edge points in the two-dimensional hour matrix by 255, and representing non-edge points in the two-dimensional hour matrix by 0 to obtain an edge matrix Y _ hour;

s2.4, acquiring the proportion of the number of times of occurrence of edge points in each hour column vector to the total number of rows according to the edge matrix Y _ hour;

and S2.5, judging whether the proportion of the edge points in each hour column vector exceeds 55%, if so, marking the corresponding column as a stable edge column as 1, otherwise, marking the corresponding column as a non-edge column as 0.

Wherein x' represents a normalized result vector of the electricity quantity data at a fixed hour time point, and x represents a history occurring at a fixed hourVector of electric quantity, PARAM _ x_maxRepresenting a 0.99 quantile of electrical quantity x, PARAM _ x_minRepresenting the 0.01 quantile, y, of the quantity of electricity x_maxRepresenting the upper bound of the normalized range, y_minLower bound, x, representing the normalized range_minDenotes the minimum value of the electric quantity, x_maxRepresenting the maximum value of the electricity, day representing the number of days the electricity x is located, and hour _ cols representing the 24 columns corresponding to 24 hours in the two-dimensional hour matrix.

In this embodiment, each column of the two-dimensional hour matrix X _ hour (day, hour _ cols) includes a date specific to a certain day and a power consumption amount corresponding thereto for 24 hours, and the total number of columns of the two-dimensional hour matrix X _ hour (day, hour _ cols) is the total number of days in the training set data.

The formula for calculating the proportion of the occurrence frequency of each row of edge points in the number of days in the step S2.4 is as follows:

wherein, cnt_dayRepresents the number of days of edge points and the total number of days of non-edge points, cnt, in the edge matrix Y _ hour₀Indicates the number of days, cnt, of non-edge points₂₅₅Represents the number of days of occurrence of edge points, z _ i and

the number of times of the edge points in the ith hour corresponding column vector in the edge matrix Y _ hour is shown, and f (z _ i) represents the proportion of the edge points in the total days.

The step S3 includes the following sub-steps:

s3.1, taking the data corresponding to the stable edge column for hours as sample data; the sample data sample_jComprises the following steps:

sample_j＝{train_j}

wherein, train_jData representing the corresponding stable edge column at the j hour;

s3.2, taking the data of the hour corresponding to the non-edge column and the hours adjacent to the hour as sample data sample_iSample data sample_iComprises the following steps:

sample_i＝{train_{i_before},train_i,train_{i_end}}；

wherein, train_iIndicating data corresponding to a non-edge column in the ith hour, wherein j belongs to {0,1,2,. 23 }, i belongs to {0,1,2,. 23, }, j is not equal to i, train_{i_before}Electricity data, train, representing the hour i _ before the ith hour_{i_end}Electricity consumption data representing an hour i _ end after the ith hour; when i is 0, i _ before is 23; when i is 23, i _ end is 0.

The sub-steps of the step S4 are as follows:

s4.1, sample data sample is utilized_iAnd sample_jGenerating a first quartile q75_iAnd a third quartile q25_i；

S4.2, according to the first quartile q75_iAnd a third quartile q25_iThe upper and lower bound thresholds for generating the electricity for the business are:

wherein thresh _ max_iUpper threshold, thresh _ min, representing commercial power usage_iA lower threshold representing the electricity usage of the business.

The step S5 includes the following sub-steps:

s5.1, judging whether the electricity data to be detected is larger than an upper bound threshold thresh _ max_iIf yes, abnormal power utilization is achieved, otherwise, the step S5.2 is carried out;

s5.2, judging whether the power consumption data is smaller than a lower bound threshold thresh _ min_iIf so, judging the power consumption is abnormal, otherwise, judging the power consumption is normal, and obtaining the detection result of the abnormal power consumption of the business body.

The invention automatically identifies the position with large gradient change by utilizing edge detection, is beneficial to effectively identifying the time point position of starting and stopping the loop and assists in judging whether the loop starting time in the business body service is on time or not; for non-edge points, the front hour and the back hour have reference effects, so that samples in a single hour are increased by adopting the front data and the back data, and the effective boundary corresponding to the hour can be better judged, and further abnormity can be judged.

The invention fully utilizes the periodic characteristic of business day of the business by monitoring the electricity consumption of the business, avoids the problem of different direction difference between hours, can realize the alarm of abnormal electricity consumption, and is beneficial to the manager to carry out deviation analysis and abnormal judgment on the business.

Claims (8)

1. A business body abnormal electricity utilization detection method based on an edge algorithm is characterized by comprising the following steps:

s1, collecting power consumption data to be detected, determining the date of the power consumption data to be detected, and determining training set data according to the date of the power consumption data to be detected;

s2, carrying out edge detection on the training set data by using an edge detection algorithm to obtain a stable edge column;

s3, obtaining sample data according to the stable edge column and the training set data;

s4, establishing an abnormal rule through sample data to obtain upper and lower bound thresholds of the electricity consumption of the business;

and S5, carrying out abnormal electricity utilization detection on the electricity data to be detected through the upper and lower threshold values to obtain the detection result of the abnormal electricity utilization of the commercial object.

2. The method for detecting abnormal electricity consumption of business body based on edge algorithm as claimed in claim 1, wherein said step S1 includes the following sub-steps:

s1.1, collecting power consumption data to be detected, and determining the date d and the month t of the power consumption data to be detected;

and S1.2, acquiring data according to the date d and the month t, and taking the acquired data as training set data.

3. The business body abnormal electricity utilization detection method based on the edge algorithm as claimed in claim 2, wherein the specific method for data collection according to the date d and the month t in the step S1.2 is as follows: and acquiring the electricity utilization data from No. 1 to date d in the month t, the electricity utilization data of the last two months in the month t, the electricity utilization data of two months adjacent to the same month in the last year and the electricity utilization data of the same month in the last year.

4. The method for detecting abnormal electricity consumption of business body based on edge algorithm as claimed in claim 1, wherein said step S2 includes the following sub-steps:

s2.1, normalizing the training set data to be in a range of [0,255], wherein the normalized calculation formula is as follows:

s2.2, expressing the normalized electric quantity by days, and converting the electric quantity into a 24-column two-dimensional hour matrix X _ hour (day, hour _ cols);

s2.3, performing edge detection on the two-dimensional hour matrix by using an edge detection algorithm, representing edge points in the two-dimensional hour matrix by 255, and representing non-edge points in the two-dimensional hour matrix by 0 to obtain an edge matrix Y _ hour;

s2.4, acquiring the proportion of the number of times of occurrence of edge points in each hour column vector to the total number of rows according to the edge matrix Y _ hour;

and S2.5, judging whether the proportion of the edge points in each hour column vector exceeds 55%, if so, marking the corresponding column as a stable edge column as 1, otherwise, marking the corresponding column as a non-edge column as 0.

Wherein x' represents a normalized result vector of the power data at a fixed hour time point, x represents a historical power vector occurring at a fixed hour, PARAM _ x_maxRepresenting a 0.99 quantile of electrical quantity x, PARAM _ x_minRepresenting the 0.01 quantile, y, of the quantity of electricity x_maxRepresenting the upper bound of the normalized range, y_minLower bound, x, representing the normalized range_minDenotes the minimum value of the electric quantity, x_maxRepresenting the maximum value of the electricity, day representing the number of days the electricity x is located, and hour _ cols representing the 24 columns corresponding to 24 hours in the two-dimensional hour matrix.

5. The business body abnormal electricity utilization detection method based on the edge algorithm as claimed in claim 4, wherein the proportion of the number of the edge points appearing in each column in the step S2.4 to the number of days is calculated by the formula:

wherein, cnt_dayRepresents the number of days of edge points and the total number of days of non-edge points, cnt, in the edge matrix Y _ hour₀Indicates the number of days, cnt, of non-edge points₂₅₅Represents the number of days of occurrence of edge points, z _ i and

the number of times of the edge points in the ith hour corresponding column vector in the edge matrix Y _ hour is shown, and f (z _ i) represents the proportion of the edge points in the total days.

6. The method for detecting abnormal electricity consumption of business body based on edge algorithm as claimed in claim 1, wherein said step S3 includes the following sub-steps:

s3.1, taking the data corresponding to the stable edge column for hours as sample data; the sample data sample_jComprises the following steps:

sample_j＝{train_j}

wherein, train_jData representing the corresponding stable edge column at the j hour;

s3.2, taking the data of the hour corresponding to the non-edge column and the hours adjacent to the hour as sample data sample_iSample data sample_iComprises the following steps:

sample_i＝{train_{i_before},train_i,train_{i_end}}；

wherein, train_iIndicating data corresponding to a non-edge column in the ith hour, wherein j belongs to {0,1,2,. 23 }, i belongs to {0,1,2,. 23, }, j is not equal to i, train_{i_before}Electricity data, train, representing the hour i _ before the ith hour_{i_end}Electricity consumption data representing an hour i _ end after the ith hour; when i is 0, i _ before is 23; when i is 23, i _ end is 0.

7. The method for detecting abnormal electricity consumption of commercial body based on edge algorithm as claimed in claim 6, wherein the sub-steps of step S4 are:

s4.1, sample data sample is utilized_iAnd sample_jGenerating a first quartile q75_iAnd a third quartile q25_i；

S4.2, according to the first quartile q75_iAnd a third quartile q25_iThe upper and lower bound thresholds for generating the electricity for the business are:

wherein thresh _ max_iUpper threshold, thresh _ min, representing commercial power usage_iA lower threshold representing the electricity usage of the business.

8. The method for detecting abnormal electricity consumption of commercial body based on edge algorithm as claimed in claim 7, wherein said step S5 includes the following sub-steps:

s5.1, judging whether the electricity data to be detected is larger than an upper bound threshold thresh _ max_iIf yes, abnormal power utilization is achieved, otherwise, the step S5.2 is carried out;

s5.2, judging whether the power consumption data is smaller than a lower bound threshold thresh _ min_iIf so, judging the power consumption is abnormal, otherwise, judging the power consumption is normal, and obtaining the detection result of the abnormal power consumption of the business body.

Images