CN103187804A - Station area electricity utilization monitoring method based on bad electric quantity data identification - Google Patents

Abstract

The invention provides a station area electricity utilization monitoring method based on bad electric quantity data identification. The station area electricity utilization monitoring method based on bad electric quantity data identification comprises the following steps of: calculating a line loss rate according to the acquired meter-reading electric quantity data of a station area during a time period, and initially judging whether an abnormal electricity utilization phenomenon exists or not; in the case that the abnormal electricity utilization phenomenon is found, turning to a B-spline function identification method, and identifying the line loss rate, and abnormal data in the daily electricity consumption data of each user; and locating suspicious electricity-stealing users by comparing the abnormal time periods corresponding to the line loss rate data of the power grid of the station area and the daily electricity consumption data of each user. The method provided by the invention only needs the meter-reading electric quantity data of the power grid of the station area, as well as is low in requirements on original data, strong in engineering applicability, and easy to be popularized and applied in the existing power grids of station areas in China. The method provided by the invention is adequate in theoretical basis and good in identification effect by identifying and locating the suspicious electricity-stealing users based on a B-spline function.

Description

A kind of platform district electricity consumption monitoring method based on bad electric quantity data identification

Technical field

The invention belongs to electric power system electricity consumption monitoring and analysis technical field, be specifically related to a kind of platform district electricity consumption monitoring method based on bad electric quantity data identification.

Background technology

The management line loss of China's platform district electrical network is high for a long time, brings enormous economic loss to power department, wherein stealing or be the one of the main reasons that causes the management line loss to remain high with electrical phenomena unusually.Owing to reasons such as stealing mode variation and Prevention Stealing Electricity Technology level are low, the stealing problem is perplexing power supply departments at different levels always.Electricity filching behavior has not only seriously been upset for the electricity consumption order, has also had a strong impact on the quality of power supply and the power supply reliability of platform district electrical network.Therefore, how to implement electricity consumption monitoring, thereby in time find, locate and forbid electricity filching behavior, become the economic benefit that ensures power supply enterprise and the just rights and interests of validated user, improve the quality of power supply of platform district electrical network and guarantee the major issue of power supply reliability.

For a long time, the electricity consumption monitoring of China mainly relies on the manpower inspection, anti-" anti-electricity-theft " design of opposing electricity-stealing and mainly being based on electric supply meter, for example, adopt the special-purpose lock of special-purpose batch meter or metering, ammeter is installed additional pick-proof lead sealing and lead-in wire sleeve pipe, adopt the inverse-stopping type ammeter.The major defect of manpower inspection is influenced greatly by subjective factor, is difficult to accurately, in time note abnormalities and uses electrical phenomena; And " anti-electricity-theft " of ammeter design can't prevent from showing before wiring or hidden electricity filching behavior such as sunken cord.

In recent years, along with the development of load data acquisition systems such as remote meter reading, occurred based on table counting electricity consumption monitoring method according to one's analysis.As Chinese patent " a kind of electricity consumption monitoring method " (patent No. CN101477163), disclosed method is at first to gather voltage, electric current and power freezing data by power consumption monitoring terminal to be sent to main website, judge relatively that in main website management of computing line loss rate and with the setting threshold values no abnormal electricity consumption phenomenon is arranged, if find multiplexing electric abnormality then utilize nonlinear programming approach to calculate each user power, locate suspicious stealing user by this rated output relatively and the size of freezing power.The major defect of this method is voltage, electric current and the power data that all users are gathered in requirement, and the rack data in known district, and to the requirement height of data platform, engineering practicability is poor.At present, the most of load data acquisition terminals of China can only provide the electric quantity data of checking meter, and have limited the applicability of said method.

Summary of the invention

The objective of the invention is the deficiency at existing electricity consumption monitoring method, a kind of platform district electricity consumption monitoring method based on bad electric quantity data identification is provided.This method only needs platform district user's the electric quantity data of checking meter, and locatees suspicious stealing user by the bad electric quantity data of identification, has the advantages that identification effect is good, engineering practicability strong, be easy to apply.For this reason, the present invention is by the following technical solutions:

It is earlier to the platform district of a certain period of the collecting electric quantity data of checking meter, calculate line loss rate and tentatively judged whether the unusual electrical phenomena of using, if note abnormalities with electrical phenomena then change over to based on the abnormal data in B-spline function discrimination method identification line loss rate and each user's daily power consumption data, by comparing platform district grid line loss rate data and each corresponding unusual period of user's daily power consumption data, locate suspicious stealing user.

On the basis of adopting above-mentioned scheme rapidly, the present invention also can adopt following further technical scheme:

The present invention includes following steps:

(1), provides platform district summary table in a certain period and the electric quantity data He Tai district electrical network day line loss rate threshold value β that checks meter day of each user's submeter _c

(2), computer board district day line loss rate;

At first distinguish computer board district electrical network summary table and the daily power consumption of each user's submeter in period T according to the data of step (1),

The daily power consumption computing formula of electrical network summary table in period T is:

W(i)=P(i+l)-P(i),i＝1,2,…,n （1）

In the formula, W (i) is platform district electrical network summary table in period T i days daily power consumption, P (i) is platform district electrical network summary table electric weight of checking meter of i days in period T, and P (i+1) is platform district electrical network summary table electric weight of checking meter of (i+1) day in period T, and n is electricity consumption fate in the period T;

The daily power consumption computing formula of each user's submeter in period T is in the platform district electrical network:

W ^k(i)=P ^k(i+1)-P ^k(i);i=1,2,…,n,k=1,2,…m. （2）

In the formula, W ^k(i) be user k in period T i days daily power consumption, P ^k(i) be user k electric weight of checking meter of i days in period T, P ^k(i+1) be the electric weight of checking meter in user k (i+1) sky in period T, n is the electricity consumption fate in the period T, and m is the user's number in the platform district electrical network;

The average day line loss rate in period T according to platform district electrical network summary table and the daily power consumption computer board district electrical network of each user's submeter in period T, its computing formula is:

${\mathrm{\β}}_i=\frac{W\left(i\right)-\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{W}^k\left(i\right)}{W\left(i\right)}\×100\%---\left(3\right)$

${\mathrm{\β}}_{\mathrm{av}}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\β}}_i}{n}---\left(4\right)$

In the formula, β _AvBe the average day line loss rate of platform district electrical network in period T, β _iBe platform district electrical network day line loss rate of i days in period T, n is electricity consumption fate in the period T;

(3), relatively average day line loss rate β _AvWith threshold value β _c, judging has no abnormal electricity consumption phenomenon, works as β _Av〉=β _cThe time, illustrate to exist and use electrical phenomena unusually, go to step (4); Otherwise, do not have the unusual electrical phenomena of using;

(4), utilize the bad line loss rate data of coming identification platform district electrical network based on the B-spline function discrimination method,

Concrete steps are as follows:

4-1) form knot vector

At first calculate n inscribed polygon i bar length of side l that data point constitutes on the line loss rate primitive curve _iWith inscribed polygon length of side length overall L, computing formula is:

l _i=|β _i+1-β _i|, $L=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{l}_i,i=\mathrm{1,2},...,n-1---\left(5\right)$

In the formula, β _iBe i data point on the line loss rate primitive curve, L is line loss rate primitive curve inscribed polygon length of side length overall, and n is the data point number of line loss rate on the primitive curve;

Calculate the knot vector X=(x of three uniform B-Spline function curves of line loss rate then ₁, x ₂, x ₃..., x _N+3) be

$X=\left[\begin{array}{cccccccccc}0,& 0,& 0,& \frac{\underset{j=1}{\overset{2}{\mathrm{\Σ}}}{l}_j}{L},& \frac{\underset{j=1}{\overset{3}{\mathrm{\Σ}}}{l}_j}{L},& \·\·\·\·\·\·,& \frac{\underset{j=1}{\overset{n-2}{\mathrm{\Σ}}}{l}_j}{L},& 1,& 1,& 1\end{array}\right]---\left(6\right)$

In the formula: l _j, L, n the same formula of meaning (5);

4-2) the cubic B-spline basic function sytem matrix N of calculating line loss rate, its computing formula is as follows:

In the formula: N _{I, 1}(X), N _I+1,1(X) be respectively i and i+1 function base of the B-spline function of line loss rate; N _{I, 2}(X), N _I+1,2(X) be respectively i and i+1 function base of the Quadric Spline function of line loss rate; N _{I, 3}(X) be i function base of the cubic B-spline function of line loss rate; x _i, x _I+1, x _I+2, x _I+3All represent the nodal value in the knot vector of line loss rate.Can form the n rank square formation N of the cubic B-spline basic function system of line loss rate according to the n on the line loss rate primitive curve day line loss rate numerical point;

4-3) the control point vector C of calculating spline function curve, its computing formula is as follows:

C=(N ^TN+λR) ^-11N ^Tβ （8）

In the formula: β is original day line loss rate column vector data, and N is the cubic B-spline basic function matrix of line loss rate, N ^TBe the transposed matrix of the cubic B-spline basic function matrix N of line loss rate, λ is the spline curve smoothing parameter, chooses empirical value λ=10 ^-9.5, R is square formation, and R=∫ [N (x) N ^T(x)] " dx;

4-4) the confidential interval of calculating line loss rate, its calculation procedure is as follows:

4-4-1) the predicated error of calculating line loss rate B-spline curves match value

At first calculate the mean square deviation MSE of line loss rate spline curve, computing formula is:

$\mathrm{MSE}=\frac{1}{n-f}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{\β}}_i-{\widehat{\mathrm{\β}}}_i)}^2---\left(9\right)$

In the formula, f is the degree of freedom of spline curve fitting, and f=trace(S),, namely f equals the summation of the diagonal entry of n rank coefficient square formation S, coefficient matrix S=N (N ^TN+ λ R) ^-1N ^T, N, N ^T, λ and the same formula of R meaning (8), β _iBe the line loss rate value of line loss rate primitive curve at i days,

Be the line loss rate match value of B-spline curves at i days, i.e. line loss rate match value vector

In i element, and

Calculate a day line loss rate match value then Predicated error s _i, computing formula is:

$s_i=\sqrt{\mathrm{MSE}+S*S^T*\mathrm{MSE}}---\left(10\right)$

In the formula, s _iIt is i days day line loss rate match value

Predicated error, coefficient matrix S, the same formula of mean square deviation MSE (8), S ^TTransposed matrix for coefficient matrix S;

4-4-2) utilize the parameter method of interval estimation, according to the mathematical definition of probability statistics confidential interval, calculate a day line loss rate match value

Confidential interval upper limit β i at i days ₊, its computing formula is:

${\mathrm{\β}}_{i+}={\widehat{\mathrm{\β}}}_i+Z_{1-\mathrm{\α}/2}*s_i---\left(11\right)$

In the formula: β i ₊Be the upper limit of confidential interval,

For the B-spline curves of line loss rate at i days line loss rate match values, Z _1-a/2Be 100* (1-α) the % percentile of standardized normal distribution, α is significance level, s _iIt is i days day line loss rate match value Predicated error;

4-5) scoring table district grid line loss rate abnormal time section

Pointwise is platform district electrical network day line loss rate β relatively _iWith corresponding confidential interval upper limit β i ₊Magnitude relationship, work as β _i＞β i ₊The time, call wire loss rate β _iUnusual corresponding time point T _i, platform district electrical network day line loss rate β then _iThe abnormal time section is [T _i-1, T _i+ 1], adds at last and all line loss rate abnormal times just constitute platform district electrical network line loss rate abnormal time section set T in period T _Loss

(5) the bad electric quantity data of each user of identification

Employing comes identification platform each user's of district bad daily power consumption data based on the B-spline function discrimination method, and concrete steps are as follows:

5-1) form knot vector

At first calculate n inscribed polygon i bar length of side that data point constitutes on the daily power consumption primitive curve of user k With inscribed polygon length of side length overall L ^k, computing formula is:

$l_i^k=|W^k(i+1)-W^k\left(i\right)|,$ $L^k=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{l}_i^k,i=\mathrm{1,2},...,n-1---\left(12\right)$

In the formula: W ^k(i) be i days daily power consumptions on the daily power consumption primitive curve of user k in period T, W ^k(i+1) be i+1 days daily power consumptions on the daily power consumption primitive curve of user k in period T, L ^kBe the daily power consumption primitive curve inscribed polygon length of side length overall of user k, n is daily power consumption data point number on the primitive curve;

Calculate the knot vector X of three uniform B-Spline function curves of the daily power consumption of user k then ^k=(x ^k ₁, x ^k ₂, x ^k ₃..., x ^k _N+3) be:

$X^k=\left[\begin{array}{cccccccccc}0,& 0,& 0,& \frac{\underset{j=1}{\overset{2}{\mathrm{\Σ}}}{l}_j^k}{L^k},& \frac{\underset{j=1}{\overset{3}{\mathrm{\Σ}}}{l}_j^k}{L^k},& \·\·\·\·\·\·,& \frac{\underset{j=1}{\overset{n-2}{\mathrm{\Σ}}}{l}_j^k}{L^k},& 1,& 1,& 1\end{array}\right]---\left(13\right)$

In the formula:

Be the j bar length of side of the line loss rate primitive curve inscribed polygon of user k, L ^kBe line loss rate primitive curve inscribed polygon length of side length overall, n is data point number on day electricity consumption primitive curve;

5-2) the cubic B-spline basic function sytem matrix N of the daily power consumption of calculating user k ^k, its computing formula is as follows:

In the formula: N ^k _{I, 1}(X), N ^k _I+1,1(X) be respectively i and i+1 function base of the B-spline function of user k; N ^k _{I, 2}(X), N ^k _I+1,2(X) be respectively i and i+1 function base of the Quadric Spline function of user k; N ^k _{I, 3}(X) be i function base of the cubic B-spline function of user k; x ^k _i, x ^k _I+1, x ^k _I+2, x ^k _I+3The knot vector X that all represents user k ^kIn nodal value.Can form the n rank square formation N of cubic B-spline basic function system according to n daily power consumption numerical point on the daily power consumption primitive curve of user k ^k

5-3) the control point vector C of the daily power consumption spline function curve of calculating user k ^k, its computing formula is as follows:

C ^k=(N ^kTN ^k+λR ^k) ^-1N ^kTW ^k （15）

In the formula, W ^kBe the original daily power consumption data rows vector of user k, N ^kBe the cubic B-spline basic function matrix of user k, N ^KTCubic B-spline basic function matrix N for user k ^kTransposed matrix, λ is the spline curve smoothing parameter, chooses empirical value λ=10 ^-9.5, R ^kBe m dimension square formation, wherein m is platform district electrical network user number, and R ^k=∫ [N ^k(x ^k) (N ^k(x ^k)) T] " dx ^k

5-4) the confidential interval of the daily power consumption of calculating user k, its calculation procedure is as follows:

5-4-1) the predicated error of the daily power consumption B-spline curves match value of calculating user k

At first calculate the daily power consumption mean square deviation MSE of user k ^k, computing formula is:

${\mathrm{MSE}}^k=\frac{1}{n-f^k}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(W^k\left(i\right)-{\hat{W}}^k\left(i\right))}^2---\left(16\right)$

In the formula, MSE ^kBe the mean square deviation of the daily power consumption curve match of user k, f ^kBe the degree of freedom of the daily power consumption spline curve fitting of user k, and f ^k=trace(S ^k), i.e. f ^kEqual n rank coefficient square formation S ^kThe summation of diagonal entry, coefficient matrix S ^k=N ^k(N ^KTN ^k+ λ R ^k) ^-1N ^KT, W ^k(i) be i days daily power consumptions on the daily power consumption primitive curve of user k in period T,

Be the B-spline curves of the user k daily power consumption match value at i days;

Calculate the predicated error of the daily power consumption match value of user k then, computing formula is:

$s_i^k=\sqrt{{\mathrm{MSE}}^k+S^k*S^{\mathrm{kT}}*{\mathrm{MSE}}^k}---\left(17\right)$

In the formula:

Be the daily power consumption match value of user k at i days

Predicated error, coefficient matrix S ^k, mean square deviation MSE ^kSame formula (15), S ^KTBe coefficient matrix S ^kTransposed matrix;

5-4-2) utilize the parameter method of interval estimation, according to the mathematical definition of probability statistics confidential interval, calculate the daily power consumption match value of user k

Confidential interval [W at i days ^ki ₁, W ^ki ₂] bound, its computing formula is:

$\left\{\begin{array}{c}{W}_{i1}^k={\hat{W}}^k\left(i\right)-Z_{1-\mathrm{\α}/2}*s_i^k\\ W_{i2}^k={\hat{W}}^k\left(i\right)+Z_{1-\mathrm{\α}/2}*s_i^k\end{array}---\left(18\right)\right.$

In the formula: W ^ki ₁Be the lower limit of user k i days daily power consumption confidential intervals, W ^ki ₂Be the upper limit of user k i days daily power consumption confidential intervals,

Be i days the daily power consumption match value of B-spline curves of user k, Z _1-a/2Be 100* (1-α) the % percentile of standardized normal distribution, α is significance level,

It is i days daily power consumption match value

Predicated error;

5-5) recording user daily power consumption abnormal time section

The magnitude relationship of day electric weight Wk (i) with corresponding confidential interval [Wki1, the Wki2] of user k is compared in pointwise, if all daily power consumption Wk (i) ∈ [Wki1, Wki2] of user k, then user k does not find daily power consumption data exception point, judges that namely user k is normal users; Otherwise work as

The time, judge that then user k is undesired user, the unusual corresponding time T ik of daily power consumption Wk (i) of recording user k, the unusual electricity consumption time period of daily power consumption Wk (i) of stipulating user k is [Tik-1, Tik+1], add and all daily power consumption abnormal times just constitute user k daily power consumption abnormal time section set T in period T ^k

(6), change different users, repeating step (5) to all user's electric weight bad data identifications in the platform district, is judged all suspicious stealing users of platform district electrical network.

The present invention also comprises step (7):

According to the bad data identification result of step (4) and (5), relatively each user's daily power consumption abnormal time section is gathered T ^kWith platform district electrical network statistics line loss rate abnormal time section set T _LossThe juxtaposition degree; As time period T _LossWith T ^kWhen both are identical, judge that then user k is the suspicious stealing user of one-level; Work as T _LossWith T ^kBoth time periods judge that then user k is the suspicious stealing user of secondary, works as T when overlapping _LossWith T ^kBoth time periods judge that then user k is three grades of suspicious stealing users when not overlapping.

Above-mentioned λ=10 ^-9.5

After the present invention adopts technique scheme, mainly contain following effect:

1. the inventive method only needs the electric quantity data of checking meter of platform district electrical network, less demanding to initial data, and engineering adaptability is strong, is easy to apply in the present platform district electrical network of China.

2. the inventive method is come identification and the suspicious stealing user in location based on B-spline function, and theoretical foundation is abundant, and identification effect is good.

Description of drawings

Fig. 1 is the program flow chart of the inventive method;

Fig. 2 is the identification result figure of a Zhejiang Hangzhou bad line loss rate data of district's electrical network;

Fig. 3 is the identification result figure of a Zhejiang Hangzhou bad electric quantity data of district electrical network user k;

Fig. 4 is the identification result figure of a Zhejiang Hangzhou bad electric quantity data of district electrical network user j.

Embodiment

Below in conjunction with embodiment, be example with Zhejiang Hangzhou district's electrical network, further checking illustrates correctness of the present invention.

As shown in Figure 1, a kind of concrete steps of the electricity consumption monitoring method based on bad electric quantity data identification are as follows

(1), input basic data

Import platform district summary table in a certain period and check meter day electric quantity data and day line loss rate threshold value β of each user's submeter _c

(note is made period T to input Zhejiang Hangzhou district's electrical network between in July, 2011 ~ November ₁) platform district summary table and 39 users' the electric quantity data (amount to 151 days) of checking meter day, the threshold value of platform district electrical network statistics line loss rate: β _c=3.5%.

(2) computer board district daily power consumption and day line loss rate

1) after (1) step finished, respectively computer board district electrical network summary table and each user's submeter were at period T ₁The daily power consumption of interior every day, computing formula are formula (1) and the formula (2) in the technical scheme.

According to period T ₁In the day electric quantity data of checking meter of platform district electrical network, according to the formula in the technical scheme (1), calculate the daily power consumption of user k, its partial results is as shown in table 1 below:

The daily power consumption result of calculation of form 1 district summary table and user k

Attention: the time fences numeral in the table 1 with on July 1st, 2011 as the 1st day, all the other time points are by that analogy.

2) computer board district electrical network is at period T ₁Average day interior line loss rate β _Av, its computing formula is formula (2) and the formula (3) in the technical scheme.

According to period T ₁Interior platform district electrical network daily power consumption data computation result according to the formula in the technical scheme (3), calculates period T ₁Interior day line loss rate β, partial results is as shown in table 2 below:

Table 2 district's electrical network statistics line loss rate result of calculation

By the formula (4) in the technical scheme, computer board district electrical network is at period T ₁Average day interior line loss rate is β _Av=3.729%.

(3) judge whether to exist and use electrical phenomena unusually

(2) step was compared average day line loss rate β after finishing _AvWith threshold value β _cSize, work as β _Av〉=β _cThe time, judgement platform district electrical network exists uses electrical phenomena unusually, goes to step (3); Otherwise, owing to do not find that platform district electrical network is at period T ₁The interior existence used electrical phenomena, EP (end of program) unusually.

According to the step in the technical scheme (3), relatively average day line loss rate β _AvWith the initial threshold value β that sets _cMagnitude relationship because β _Av=3.529%＞β _c, illustrate to exist and use electrical phenomena unusually, so the step of going to (4) is carried out the bad data identification.

(4) the bad line loss rate data of identification

(3) step was utilized the bad line loss rate data of B-spline function discrimination method identification platform district electrical network after finishing, and concrete calculation procedure is as follows:

4-1) form knot vector

At first calculate the knot vector X of line loss rate spline curve, computing formula is formula (5) and the formula (6) in the technical scheme.

According to period T ₁The electric quantity data of interior platform district electrical network, according to the formula in the technical scheme (5) and formula (6), the knot vector that calculates Hangzhou district's electrical network day line loss rate curve is as shown in table 3 below:

The knot vector X of table 3 district's electrical network statistics line loss rate curve

4-2) calculate the B spline base function

After (4-1) goes on foot and finish, calculate the cubic B-spline basic function sytem matrix N of line loss rate, computing formula is the formula (7) in the technical scheme.

According to period T ₁Interior platform district electrical network day line loss rate result of calculation, the knot vector X of joint line loss rate spline curve, by the formula (7) in the technical scheme, recursion calculates cubic B-spline basic function sytem matrix N.

4-3) inverse control point vector

After (4-2) finishes, the control point of computer board district grid line loss rate spline function curve vector C, computing formula is the formula (8) in the technical scheme.

According to period T ₁Interior platform district electrical network day line loss rate result of calculation by the formula (8) in the technical scheme, calculates the control point vector C of platform district electrical network statistics line loss rate spline function curve, and is as shown in table 4 below:

Table 4 control point vector C

4-4) calculate confidential interval

After finish (4-3), calculate the confidential interval of line loss rate, its calculation procedure is as follows:

4-4-1) the predicated error of calculating line loss rate spline curve fitting value, computing formula is formula (9) and the formula (10) in the technical scheme.

According to period T ₁Interior platform district electrical network day line loss rate data, by the formula (9) in the technical scheme, a mean square deviation MSE who calculates day line loss rate spline curve is: MSE=0.2342; By the formula (10) in the technical scheme, can calculate the predicated error s of line loss rate spline curve fitting value _i

4-4-2) the confidential interval of calculating line loss rate, computing formula is the formula (11) in the technical scheme.

According to period T ₁The electric quantity data of interior platform district electrical network by the formula (11) in the technical scheme, calculates the confidential interval upper limit of line loss rate spline curve fitting value, and part result of calculation is as shown in table 5 below:

The confidential interval of table 5 day line loss rate spline curve fitting value

Line loss rate	...	3.101	4.294	4.237	4.531	4.091	4.405	4.105	4.112	3.784	...
												The upper limit	...	4.398	4.579	4.604	4.701	4.670	4.617	4.472	4.393	4.412	...

4-5) scoring table district grid line loss rate abnormal time section

After (4-4) finished, pointwise is platform district electrical network day line loss rate β relatively _iWith corresponding confidential interval upper limit β i ₊Magnitude relationship, work as β _i＞β i ₊The time, call wire loss rate β _iUnusual corresponding time point T _i, platform district electrical network day line loss rate β then _iThe abnormal time section is [T _i-1, T _i+ 1], adds at last and all line loss rate abnormal times just constitute platform district electrical network line loss rate abnormal time section set T in period T _Loss

According to period T ₁The electric quantity data of interior platform district electrical network, pointwise is platform district electrical network day line loss rate β relatively _iWith corresponding confidential interval upper limit β i ₊Magnitude relationship, obtain line loss rate abnormal time point, identification result as shown in Figure 2, the record result as shown in table 6 below:

Table 6 line loss rate abnormal time segment record table

(5) the bad electric quantity data of each user of identification

(4) step utilized the B-spline function discrimination method to come identification platform each user's of district bad daily power consumption data after finishing, and was example with user k below, specified user's bad daily power consumption data identification.Step is as follows:

5-1) form knot vector

At first calculate n inscribed polygon i bar length of side that data point constitutes on the daily power consumption primitive curve of user k

With inscribed polygon length of side length overall L ^k, computing formula is formula in the technical scheme (11) and formula (12).

According to period T ₁The daily power consumption data of interior platform district user k, according to the formula in the technical scheme (11) and formula (12), the knot vector of the daily power consumption curve of a calculating Hangzhou district electrical network user k is as shown in table 7 below:

The knot vector X of the daily power consumption curve of table 7 user k ^k

5-2) calculate the B spline base function

After (5-1) finishes, calculate the cubic B-spline basic function sytem matrix N of the daily power consumption of user k ^k, its computing formula is the formula (13) in the technical scheme

According to period T ₁The daily power consumption data of interior platform district user k are in conjunction with the knot vector X of daily power consumption spline curve ^k, by the formula (13) in the technical scheme, recursion calculates the cubic B-spline basic function sytem matrix N of the daily power consumption spline curve of user k.

5-3) inverse control point vector

After finish (5-2), calculate the control point vector C of the daily power consumption spline function curve of user k ^k, its computing formula is the formula (14) in the technical scheme.

According to period T ₁The daily power consumption data of interior platform district user k by the formula (14) in the technical scheme, calculate the control point vector C of platform district electrical network statistics line loss rate spline function curve, and are as shown in table 8 below:

Table 8 control point vector C

5-4) calculate confidential interval

After (5-3) finishes, calculate the confidential interval of the daily power consumption of user k, its calculation procedure is as follows:

5-4-1) at first calculate the predicated error of the daily power consumption B-spline curves match value of user k, computing formula is formula (15) and the formula (16) in the technical scheme.

According to period T ₁The daily power consumption data of interior platform district user k, by the formula (15) in the technical scheme, a mean square deviation MSE who calculates day line loss rate spline curve is: MSE=4.3862; By the formula (16) in the technical scheme, can calculate the predicated error of the daily power consumption spline curve fitting value of user k

5-4-2) the confidential interval of the daily power consumption match value of calculating user k, computing formula is the formula (16) in the technical scheme.

According to period T ₁The daily power consumption data of interior platform district user k by the formula (16) in the technical scheme, calculate the confidential interval bound of the daily power consumption spline curve fitting value of user k, and part result of calculation is as shown in table 9 below:

The confidential interval of the daily power consumption spline curve fitting value of table 9 user k

Daily power consumption	...	6.66	10.4	14.5	16.58	21.82	20.55	21.72	12.7	12.86	...
												The upper limit	...	5.516	9.214	11.64	15.74	14.62	13.76	19.03	16.43	11.90	...
Lower limit	...	11.08	14.97	17.48	21.41	20.17	19.44	24.85	22.01	17.26	...

5-5) recording user daily power consumption abnormal time section

After finished (5-4), the magnitude relationship of day electric weight Wk (i) with corresponding confidential interval [Wki1, the Wki2] of user k was compared in pointwise, when

The time, the unusual corresponding time T jk of daily power consumption Wk (i) of recording user k stipulates that then the unusual electricity consumption time period of daily power consumption Wk (i) of user k is [Tjk-1, Tjk+1], then adds and all daily power consumption abnormal times just constitute user k at period T ₁Interior daily power consumption abnormal time section set T ^k

According to period T ₁The daily power consumption data of interior platform district user k, pointwise is the daily power consumption W of user k relatively ^k(i) with the corresponding confidential interval upper limit [W ^ki ₁, W ^ki ₂] magnitude relationship, obtain the daily power consumption abnormal time point of user k, identification result as shown in Figure 3, the result is as shown in table 10 below for record:

The daily power consumption abnormal time segment record table of table 10 user k

(6) the suspicious stealing user in location

According to the bad data identification result of preceding step (4) and (5), relatively each user's daily power consumption abnormal time section is gathered T ^kWith platform district electrical network statistics line loss rate abnormal time section set T _LossThe juxtaposition degree.As time period T _LossWith T ^kWhen both are identical, judge that then user k is the suspicious stealing user of one-level; Work as T _LossWith T ^kBoth time periods overlap, and judge that then user k is the suspicious stealing user of secondary, works as T _LossWith T ^kBoth time periods are not overlapping, judge that then user k is three grades of suspicious stealing users.

In conjunction with the daily power consumption bad data identification effect figure of Zhejiang district's grid line loss rate and user k, shown in accompanying drawing 2 and accompanying drawing 3, according to platform district electrical network day line loss rate abnormal time section T _LossWith the daily power consumption abnormal time section set record result of user k, shown in table 6 and table 10, T _Loss=[70,72] ∪ [74,88],

Significantly, owing to have the juxtaposition part in both time periods, and overlapping time, section was [74,77] ∪ [80,88], judged that then platform district electrical network user k is the suspicious stealing user of secondary.

Experiment effect

Be object with a Zhejiang Hangzhou district electrical network, design the validity of following simulation example checking the inventive method.

The accurate information that provides according to the Zhejiang Hangzhou power department is provided, confirmed that user k is the stealing user in September, 2011, at first the summary table of input table district electrical network and 39 users' day meter reading data according to the technical scheme in the summary of the invention, calculates the statistics line loss rate β of xx platform district, Zhejiang electrical network _iWith average day line loss rate β _Av, because average day line loss rate β _Av=3.729%＞β _cSo the step of going to (4) and step (5), bad data identification based on the B-spline function discrimination method, carry out the grid line loss rate bad data identification of platform district and each user's daily power consumption bad data identification respectively, obtain the electric quantity data identification effect figure of platform district electrical network day line loss rate and user k, shown in accompanying drawing 2 and accompanying drawing 3.Press the step (4) in the summary of the invention, can analyze and obtain platform district electrical network statistics line loss rate abnormal time section T _Loss, the result is as shown in table 6 for its record, then T _Loss=[70,72] ∪ [74,88]; The daily power consumption abnormal time section set T of user k ^k, the result is as shown in table 10 for record, T ^k=[72,77] ∪ [80,89].Significantly, owing to have the juxtaposition part in both time periods, and overlapping time, section was [74,77] ∪ [80,88], judged that then platform district electrical network user k is the suspicious stealing user of secondary.

Day meter reading data analytical calculation to user j simultaneously obtains its daily power consumption data, utilize the identification of B-spline function method to analyze, according to the identification step in the technical scheme, the day electricity consumption bad data identification effect figure of user j as shown in Figure 4, no abnormal daily power consumption data are so judge that user j is normal electricity consumer.Verify through investigation through the Zhejiang Hangzhou power department, prove that user j does not have electricity filching behavior really.

Claims (4)

1. platform district electricity consumption monitoring method based on bad electric quantity data identification, it is characterized in that it is earlier to the platform district of a certain period of the collecting electric quantity data of checking meter, calculate line loss rate and tentatively judged whether the unusual electrical phenomena of using, if note abnormalities with electrical phenomena then change over to based on the abnormal data in B-spline function discrimination method identification line loss rate and each user's daily power consumption data, by comparing platform district grid line loss rate data and each corresponding unusual period of user's daily power consumption data, locate suspicious stealing user.

2. a kind of platform district electricity consumption monitoring method based on bad electric quantity data identification as claimed in claim 1 is characterized in that it may further comprise the steps:

(1), provides platform district summary table in a certain period and the electric quantity data He Tai district electrical network day line loss rate threshold value β that checks meter day of each user's submeter _c

(2), computer board district day line loss rate;

At first distinguish computer board district electrical network summary table and the daily power consumption of each user's submeter in period T according to the data of step (1),

The daily power consumption computing formula of electrical network summary table in period T is:

W(i)=P(i+l)-P(i),i＝1,2,…,n （1）

In the formula, W (i) is platform district electrical network summary table in period T i days daily power consumption, P (i) is platform district electrical network summary table electric weight of checking meter of i days in period T, and P (i+1) is platform district electrical network summary table electric weight of checking meter of (i+1) day in period T, and n is electricity consumption fate in the period T;

The daily power consumption computing formula of each user's submeter in period T is in the platform district electrical network:

W ^k(i)=P ^k(i+1)-P ^k(i);i=1,2,…,n,k=1,2,…m. （2）

In the formula, W ^k(i) be user k in period T i days daily power consumption, P ^k(i) be user k electric weight of checking meter of i days in period T, P ^k(i+1) be the electric weight of checking meter in user k (i+1) sky in period T, n is the electricity consumption fate in the period T, and m is the user's number in the platform district electrical network;

The average day line loss rate in period T according to platform district electrical network summary table and the daily power consumption computer board district electrical network of each user's submeter in period T, its computing formula is:

${\mathrm{\β}}_i=\frac{W\left(i\right)-\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{W}^k\left(i\right)}{W\left(i\right)}\×100\%---\left(3\right)$

${\mathrm{\β}}_{\mathrm{av}}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\β}}_i}{n}---\left(4\right)$

In the formula, β _AvBe the average day line loss rate of platform district electrical network in period T, β _iBe platform district electrical network day line loss rate of i days in period T, n is electricity consumption fate in the period T;

(3), relatively average day line loss rate β _AvWith threshold value β _c, judging has no abnormal electricity consumption phenomenon, works as β _Av〉=β _cThe time, illustrate to exist and use electrical phenomena unusually, go to step (4); Otherwise, do not have the unusual electrical phenomena of using;

(4), utilize the bad line loss rate data of coming identification platform district electrical network based on the B-spline function discrimination method, concrete steps are as follows:

4-1) form knot vector

At first calculate n inscribed polygon i bar length of side l that data point constitutes on the line loss rate primitive curve _iWith inscribed polygon length of side length overall L, computing formula is:

l _i=|β _i+1-β _i|, $L=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{l}_i,i=\mathrm{1,2},...,n-1---\left(5\right)$

In the formula, β _iBe i data point on the line loss rate primitive curve, L is line loss rate primitive curve inscribed polygon length of side length overall, and n is the data point number of line loss rate on the primitive curve;

Calculate the knot vector X=(x of three uniform B-Spline function curves of line loss rate then ₁, x ₂, x ₃..., x _N+3) be

$X=\left[\begin{array}{cccccccccc}0,& 0,& 0,& \frac{\underset{j=1}{\overset{2}{\mathrm{\Σ}}}{l}_j}{L},& \frac{\underset{j=1}{\overset{3}{\mathrm{\Σ}}}{l}_j}{L},& \·\·\·\·\·\·,& \frac{\underset{j=1}{\overset{n-2}{\mathrm{\Σ}}}{l}_j}{L},& 1,& 1,& 1\end{array}\right]---\left(6\right)$

In the formula: l _j, L, n the same formula of meaning (5);

4-2) the cubic B-spline basic function sytem matrix N of calculating line loss rate, its computing formula is as follows:

In the formula: N _{I, 1}(X), N _I+1,1(X) be respectively i and i+1 function base of the B-spline function of line loss rate; N _{I, 2}(X), N _I+1,2(X) be respectively i and i+1 function base of the Quadric Spline function of line loss rate; N _{I, 3}(X) be i function base of the cubic B-spline function of line loss rate; x _i, x _I+1, x _I+2, x _I+3All represent the nodal value in the knot vector of line loss rate.Can form the n rank square formation N of the cubic B-spline basic function system of line loss rate according to the n on the line loss rate primitive curve day line loss rate numerical point;

4-3) the control point vector C of calculating spline function curve, its computing formula is as follows:

C=(N ^TN+λR) ^-1N ^Tβ （8）

In the formula: β is original day line loss rate column vector data, and N is the cubic B-spline basic function matrix of line loss rate, N ^TBe the transposed matrix of the cubic B-spline basic function matrix N of line loss rate, λ is the spline curve smoothing parameter, gets empirical value λ=10 ^-9.5, R is the m square formation, m is platform district electrical network user number, and R=∫ [N (x) N ^T(x)] " dx;

4-4) the confidential interval of calculating line loss rate, its calculation procedure is as follows:

4-4-1) the predicated error of calculating line loss rate B-spline curves match value

At first calculate the mean square deviation MSE of line loss rate spline curve, computing formula is:

$\mathrm{MSE}=\frac{1}{n-f}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{\β}}_i-{\widehat{\mathrm{\β}}}_i)}^2---\left(9\right)$

In the formula, f is the degree of freedom of spline curve fitting, and f=trace(S),, namely f equals the summation of the diagonal entry of n rank coefficient square formation S, coefficient matrix S=N (N ^TN+ λ R) ^-1N ^T, N, N ^T, λ and the same formula of R meaning (8), β _iBe the line loss rate value of line loss rate primitive curve at i days,

Be the line loss rate match value of B-spline curves at i days, i.e. line loss rate match value vector In i element, and

Calculate a day line loss rate match value then

Predicated error s _i, computing formula is:

$s_i=\sqrt{\mathrm{MSE}+S*S^T*\mathrm{MSE}}---\left(10\right)$

In the formula, s _iIt is i days day line loss rate match value

Predicated error, coefficient matrix S, the same formula of mean square deviation MSE (8), S ^TTransposed matrix for coefficient matrix S;

4-4-2) utilize the parameter method of interval estimation, according to the mathematical definition of probability statistics confidential interval, calculate a day line loss rate match value

Confidential interval upper limit β i at i days ₊, its computing formula is:

${\mathrm{\β}}_{i+}={\widehat{\mathrm{\β}}}_i+Z_{1-\mathrm{\α}/2}*s_i---\left(11\right)$

In the formula: β i ₊Be the upper limit of confidential interval,

For the B-spline curves of line loss rate at i days line loss rate match values, Z _1-a/2Be 100* (1-α) the % percentile of standardized normal distribution, α is significance level, s _iIt is i days day line loss rate match value

Predicated error;

4-5) scoring table district grid line loss rate abnormal time section

Pointwise is platform district electrical network day line loss rate β relatively _iWith corresponding confidential interval upper limit β i ₊Magnitude relationship, work as β _i＞β i ₊The time, call wire loss rate β _iUnusual corresponding time point T _i, platform district electrical network day line loss rate β then _iThe abnormal time section is [T _i-1, T _i+ 1], adds at last and all line loss rate abnormal times just constitute platform district electrical network line loss rate abnormal time section set T in period T _Loss

(5) the bad electric quantity data of each user of identification

Employing comes identification platform each user's of district bad daily power consumption data based on the B-spline function discrimination method, and concrete steps are as follows:

5-1) form knot vector

At first calculate n inscribed polygon i bar length of side that data point constitutes on the daily power consumption primitive curve of user k With inscribed polygon length of side length overall L ^k, computing formula is:

$l_i^k=|W^k(i+1)-W^k\left(i\right)|,$ $L^k=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{l}_i^k,i=\mathrm{1,2},...,n-1---\left(12\right)$

In the formula: W ^k(i) be i days daily power consumptions on the daily power consumption primitive curve of user k in period T, W ^k(i+1) be i+1 days daily power consumptions on the daily power consumption primitive curve of user k in period T, L ^kBe the daily power consumption primitive curve inscribed polygon length of side length overall of user k, n is daily power consumption data point number on the primitive curve;

Calculate the knot vector X of three uniform B-Spline function curves of the daily power consumption of user k then ^k=(x ^k ₁, x ^k ₂, x ^k ₃..., x ^k _N+3) be:

$X^k=\left[\begin{array}{cccccccccc}0,& 0,& 0,& \frac{\underset{j=1}{\overset{2}{\mathrm{\Σ}}}{l}_j^k}{L^k},& \frac{\underset{j=1}{\overset{3}{\mathrm{\Σ}}}{l}_j^k}{L^k},& \·\·\·\·\·\·,& \frac{\underset{j=1}{\overset{n-2}{\mathrm{\Σ}}}{l}_j^k}{L^k},& 1,& 1,& 1\end{array}\right]---\left(13\right)$

In the formula:

Be the j bar length of side of the line loss rate primitive curve inscribed polygon of user k, L ^kBe line loss rate primitive curve inscribed polygon length of side length overall, n is data point number on day electricity consumption primitive curve;

5-2) the cubic B-spline basic function sytem matrix N of the daily power consumption of calculating user k ^k, its computing formula is as follows:

In the formula: N ^k _{I, 1}(X), N ^k _I+1,1(X) be respectively i and i+1 function base of the B-spline function of user k; N ^k _{I, 2}(X), N ^k _I+1,2(X) be respectively i and i+1 function base of the Quadric Spline function of user k; N ^k _{I, 3}(X) be i function base of the cubic B-spline function of user k; x ^k _i, x ^k _I+1, x ^k _I+2, x ^k _I+3The knot vector X that all represents user k ^kIn nodal value, can form the n rank square formation N of cubic B-spline basic function system according to n daily power consumption numerical point on the daily power consumption primitive curve of user k ^k

5-3) the control point vector C of the daily power consumption spline function curve of calculating user k ^k, its computing formula is as follows:

C ^k=(N ^kTN ^k+λR ^k) ^-1N ^kTW ^k （15）

In the formula, W ^kBe the original daily power consumption data rows vector of user k, N ^kBe the cubic B-spline basic function matrix of user k, N ^KTCubic B-spline basic function matrix N for user k ^kTransposed matrix, λ is the spline curve smoothing parameter, chooses empirical value λ=10 ^-9.5, R ^kBe m dimension square formation, wherein m is platform district electrical network user number, and R ^k=∫ [N ^k(x ^k) (N ^k(x ^k)) ^T] " dx ^k

5-4) the confidential interval of the daily power consumption of calculating user k, its calculation procedure is as follows:

5-4-1) the predicated error of the daily power consumption B-spline curves match value of calculating user k

At first calculate the daily power consumption mean square deviation MSE of user k ^k, computing formula is:

${\mathrm{MSE}}^k=\frac{1}{n-f^k}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(W^k\left(i\right)-{\hat{W}}^k\left(i\right))}^2---\left(16\right)$

In the formula, MSE ^kBe the mean square deviation of the daily power consumption curve match of user k, f ^kBe the degree of freedom of the daily power consumption spline curve fitting of user k, and f ^k=trace(S ^k), i.e. f ^kEqual n rank coefficient square formation S ^kThe summation of diagonal entry, coefficient matrix S ^k=N ^k(N ^KTN ^k+ λ R ^k) ^-1N ^KT, W ^k(i) be i days daily power consumptions on the daily power consumption primitive curve of user k in period T,

Be the B-spline curves of the user k daily power consumption match value at i days;

Calculate the predicated error of the daily power consumption match value of user k then, computing formula is:

$s_i^k=\sqrt{{\mathrm{MSE}}^k+S^k*S^{\mathrm{kT}}*{\mathrm{MSE}}^k}---\left(17\right)$

In the formula:

Be the daily power consumption match value of user k at i days

Predicated error, coefficient matrix S ^k, mean square deviation MSE ^kSame formula (15), S ^KTBe coefficient matrix S ^kTransposed matrix;

5-4-2) utilize the parameter method of interval estimation, according to the mathematical definition of probability statistics confidential interval, calculate the daily power consumption match value of user k

Confidential interval [W at i days ^ki ₁, W ^ki ₂] bound, its computing formula is:

$\left\{\begin{array}{c}{W}_{i1}^k={\hat{W}}^k\left(i\right)-Z_{1-\mathrm{\α}/2}*s_i^k\\ W_{i2}^k={\hat{W}}^k\left(i\right)+Z_{1-\mathrm{\α}/2}*s_i^k\end{array}---\left(18\right)\right.$

In the formula: W ^ki ₁Be the lower limit of user k i days daily power consumption confidential intervals, W ^ki ₂Be the upper limit of user k i days daily power consumption confidential intervals,

Be i days the daily power consumption match value of B-spline curves of user k, Z _1-a/2Be 100* (1-α) the % percentile of standardized normal distribution, α is significance level,

It is i days daily power consumption match value

Predicated error;

5-5) recording user daily power consumption abnormal time section

The magnitude relationship of day electric weight Wk (i) with corresponding confidential interval [Wki1, the Wki2] of user k is compared in pointwise, if all daily power consumption Wk (i) ∈ [Wki1, Wki2] of user k, then user k does not find daily power consumption data exception point, judges that namely user k is normal users; Otherwise work as

The time, judge that then user k is undesired user, the unusual corresponding time T ik of daily power consumption Wk (i) of recording user k, the unusual electricity consumption time period of daily power consumption Wk (i) of stipulating user k is [Tik-1, Tik+1], add and all daily power consumption abnormal times just constitute user k daily power consumption abnormal time section set T in period T ^k

(6), change different users, repeating step (5) to all user's electric weight bad data identifications in the platform district, is judged the suspicious stealing user of platform district electrical network.

3. a kind of platform district electricity consumption monitoring method based on bad electric quantity data identification as claimed in claim 2 is characterized in that, it is characterized in that it also comprises step (7):

According to the bad data identification result of step (4) and (5), relatively each user's daily power consumption abnormal time section is gathered T ^kWith platform district electrical network statistics line loss rate abnormal time section set T _LossThe juxtaposition degree; As time period T _LossWith T ^kWhen both are identical, judge that then user k is the suspicious stealing user of one-level; Work as T _LossWith T ^kBoth time periods judge that then user k is the suspicious stealing user of secondary, works as T when overlapping _LossWith T ^kBoth time periods judge that then user k is three grades of suspicious stealing users when not overlapping.

4. a kind of platform district electricity consumption monitoring method based on bad electric quantity data identification as claimed in claim 2 is characterized in that, it is characterized in that λ=10 ^-9.5

Images