CN104794544A - Intelligent algorithm based electricity-larceny-preventive monitoring method for distributed type photovoltaic power generation - Google Patents

Abstract

The invention discloses an intelligent algorithm based electricity-larceny-preventive monitoring method for distributed type photovoltaic power generation. The method is characterized by including step 1, creating a distributed type photovoltaic power generation database, wherein database information includes historical meteorological data of a photovoltaic power station and photovoltaic power generation power data in the corresponding period; step 2, taking the meteorological data as input, taking photovoltaic power generation power values as output, and creating an electricity-larceny prevention model according to an intelligent algorithm; step 3, predicting photovoltaic power by the aid of the electricity-larceny prevention model, and calculating theoretical power generation quantity of the photovoltaic power station by the aid of a calculation model of the theoretical power generation quantity of the photovoltaic power station; step 4, inputting collected real-time power generation quantity of the photovoltaic power station and the theoretical power generation quantity of the photovoltaic power station into a power quantity abnormity judgment module to acquire an power generation electricity-larceny suspicion judgment result of the photovoltaic power station. The intelligent algorithm based electricity-larceny-preventive monitoring method has the advantages that photovoltaic power quantity abnormity judgment accuracy is high, electricity-larceny suspicion coefficients are given, auditing work efficiency is improved, and the basis is provided for effective monitoring of distributed type photovoltaic power generation.

Description

The anti-electricity-theft monitoring method of distributed photovoltaic power generation based on intelligent algorithm

Technical field

The present invention relates to the anti-electricity-theft monitoring method of a kind of distributed photovoltaic power generation based on intelligent algorithm, belong to photovoltaic generation and anti-theft electricity technology field thereof.

Background technology

Domestic still not for effective risk prevention instruments of distributed photovoltaic stealing at present, but in recent years, generation of electricity by new energy industry is constantly risen, based on distributed photovoltaic power generation its special generation mode and subsidy policy, also expedite the emergence of out much novel electricity filching means, the anti-electricity-theft work of power supply enterprise is had higher requirement.To measure power consumption less different from the stealing of conventional electric power user, and the object of distributed photovoltaic power generation stealing is the electricity volume of many metering photovoltaics, so that obtain extra subsidy.

The research object of anti-theft electricity technology lays particular emphasis on electric energy meter body, carries out mainly for legacy user's stealing mode.For the distributed photovoltaic power generation anti-theft electricity technology that this patent proposes, how to evade distributed photovoltaic power generation owner and make the many meterings of ammeter send out an amount aspect, have not yet to see correlative study achievement both at home and abroad at present.At present for the new feature of photovoltaic stealing, also do not realize the photovoltaic electricity anti-theft method of quick and precisely locating stealing suspicion user and stealing suspicion coefficient thereof.

Summary of the invention

Exist for current distributed photovoltaic power generation to gain the stealing electricity phenomenon for the purpose of subsidy by cheating, the invention provides a kind of anti-electricity-theft monitoring method based on neural network intelligent algorithm, improve the metering security capabilities of Guo Wang company distributed photovoltaic power generation, be conducive to the enforcement of distributed photovoltaic power generation subsidy policy.The present invention according to stealing number provide different stealing suspicion coefficients, judge accuracy high.

For solving the problem, technical scheme of the present invention is as follows: the anti-electricity-theft monitoring method of a kind of distributed photovoltaic power generation based on intelligent algorithm, is characterized in that, comprise the following steps:

Step one: set up distributed photovoltaic power generation database, database information comprise photovoltaic plant Historical Meteorological Information and the same period photovoltaic generation power data.

Step 2: take meteorological data as input quantity, with photovoltaic generation power value for exporting, adopting intelligent algorithm to set up anti-electricity-theft model, and verifying.

Step 3: utilize anti-electricity-theft model to predict photovoltaic power, by the theoretical generated energy computation model of photovoltaic plant, and then calculates the theoretical generated energy of photovoltaic plant.

Step 4: real-time for the photovoltaic plant collected generated energy and the theoretical generated energy of photovoltaic plant are input to electricity abnormality juding module, photovoltaic power station power generation stealing suspicion result of determination can be obtained, and according to different stealing suspicion coefficients, carry out corresponding stealing event alarm.

Preferably, the sample data of described database is classified according to fine day, cloudy, cloudy, rainy day four kinds of weather conditions, sample data is divided into training sample and test sample book, training sample and test sample book contain above-mentioned four kinds of weather patterns simultaneously, and the ratio that training sample and test sample book account for total sample is respectively 70% and 30%.

Preferably, utilize the BP neural network established, by inputting the weather information such as photovoltaic plant irradiation, temperature collected, just the prediction of photovoltaic plant output power can be realized, because photovoltaic plant output power is random variation amount, and comparatively large by the meteorological condition such as irradiation, temperature variable effect, single-point does not have reference value, for improving the accuracy of anti-electricity-theft monitoring, the present invention adopts photovoltaic power station power generation amount as stealing whether distinguishing rule.By photovoltaic plant output power predicted value, the theoretical generated energy of photovoltaic generating system can be calculated.The theoretical generated energy expression formula of photovoltaic generating system is:

$W={\∫}_0^t\mathrm{Pdt}=\mathrm{Pt}$

Wherein, P is photovoltaic generation power, and W is photovoltaic power generation quantity; Namely photovoltaic power generation quantity is the accumulation of photovoltaic generation power on time t.Consider that photovoltaic power generation quantity is statistical information, error small between photovoltaic electric amount calculated value and actual electrical measurement, under the effect of accumulated time, deviation therebetween also may be caused progressively to increase, and then cause stealing to be judged by accident.Thus, the theoretical generated energy computation model of described photovoltaic plant every ten days is a cycle for setting, in this computation period, judges that whether photovoltaic plant electricity is abnormal, carry out relevant alarm if abnormal by electricity abnormality juding module; If normal, next computation period starting point is photovoltaic plant actual electrical measurement, recalculates the theoretical generated energy of photovoltaic plant on this basis.Adopt the statistical method based on neural network, avoid the concrete modeling to photovoltaic system inversion model and opto-electronic conversion model, can systematic error be reduced, the accuracy that the theoretical generated energy improving photovoltaic generating system calculates.

Preferably, utilize distributed photovoltaic power generation electricity computation model, the theoretical generated energy of photovoltaic plant can be calculated.Consider that photovoltaic power generation quantity affects by meteorological condition larger, there is randomness and uncontrollable feature, for improving the accuracy of photovoltaic electricity abnormality juding, need to carry out uncertainty analysis, therefore stealing suspicion coefficient decision is introduced fiducial interval and is judged, the calculating of fiducial interval bound adopts sample average and standard deviation to calculate or adopt monte carlo method to obtain estimated value distribution, namely when 95% fiducial interval providing photovoltaic electricity theoretical value is [a b], can be understood as us and have the mean value of the confidence guarantee sample of 95% between a to b, and the probability made a mistake is 5%.

Preferably, the calculating of fiducial interval bound adopts sample average and standard deviation to calculate, account form is: when known sample average (M) and standard deviation (ST), lower limit of confidence interval: a=M-n*ST, the fiducial interval upper limit: b=M+n*ST; The n=1.645 when asking for 90% fiducial interval; The n=1.96 when asking for 95% fiducial interval; The n=2.576 when asking for 99% fiducial interval.

Preferably, photovoltaic generation metering ammeter is mounted in the photovoltaic power generation quantity clearing ammeter of photovoltaic DC-to-AC converter outlet, the photovoltaic power generation quantity that this table is uploaded is exactly the generated energy of this photovoltaic plant of metering, described electricity abnormality juding module is judge compared with the fiducial interval bound of photovoltaic theoretical capacity according to photovoltaic meter volume and electricity, if photovoltaic meter volume and electricity is less than the lower limit of confidence interval of photovoltaic theoretical capacity, then judge that its generating is abnormal, provide generating abnormality alarming; If photovoltaic meter volume and electricity is greater than the fiducial interval upper limit of photovoltaic theoretical capacity, then it is determined that the presence of stealing suspicion.By the collection value of distributed photovoltaic power station generated energy and comparing of this fiducial interval upper limit, obtain different stealing suspicion coefficients, thus improve the specific aim of inspection, improve inspecting efficiency.

Preferably, described stealing suspicion coefficient is that to get Δ be the number percent that difference that photovoltaic meter volume and electricity is greater than the fiducial interval upper limit of photovoltaic theoretical capacity accounts for the fiducial interval upper limit of photovoltaic theoretical capacity, namely

As 10%> Δ >0, stealing suspicion coefficient is 0.1; When 20%> Δ >=10%, stealing suspicion coefficient is 0.2; When 30%> Δ >=20%, stealing suspicion coefficient is 0.3; When 40%> Δ >=30%, stealing suspicion coefficient is 0.4; When 50%> Δ >=40%, stealing suspicion coefficient is 0.5; When 60%> Δ >=50%, stealing suspicion coefficient is 0.6; When 70%> Δ >=60%, stealing suspicion coefficient is 0.7; When 80%> Δ >=70%, stealing suspicion coefficient is 0.8; When >=80%, stealing suspicion coefficient is 0.9.According to different stealing suspicion coefficients, carry out corresponding stealing event alarm, simultaneously stealing suspicion coefficient is more close to 1, illustrates that stealing possibility is larger, and then can using user large for this month stealing suspicion coefficient as investigating object.

Preferably, described intelligent algorithm comprises any one algorithm in BP neural network, data mining technology, support vector machine.

Preferably, adopt the step of the anti-electricity-theft model of BP neural network as follows:

A. according to photovoltaic plant Historical Meteorological Information and the same period photovoltaic generation power data set up distributed photovoltaic power generation database, and sample data is divided into training sample and test sample book;

B. set up BP neural network, determine precision of prediction and network structure, namely determine nodes and output node number (m) of input number of nodes (n), the implicit number of plies, each hidden layer;

C. select suitable Algorithm for Training network, make it matching training sample as far as possible;

D. input training sample, calculate each layer and export and each layer error signal, if the error calculated meets the demands, verify this neural network by test sample book;

If e. error does not meet the demands, each layer weights of neural network need be adjusted, recalculate each layer error signal, until error meets the demands;

F. with the network that test sample book data detection trains, if test effect is fine, the weather information that the network integration that trains just can be utilized new is predicted photovoltaic power, if test effect is bad, the each layer weights of adjustment network, repeat step e, until obtain good assay.

The beneficial effect that the present invention reaches: the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm of the present invention adopts the statistical method based on neural network, photovoltaic power generation power prediction is accurate, avoids the concrete modeling to photovoltaic system inversion model and opto-electronic conversion model; The anti-electricity-theft monitoring method of the described distributed photovoltaic power generation based on intelligent algorithm adopts the photovoltaic power generation quantity with accumulated time effect as stealing decision condition, and introduces fiducial interval, and photovoltaic electricity abnormality juding accuracy is high, and stealing judges with a high credibility; The prediction generated energy that the anti-electricity-theft monitoring method of the described distributed photovoltaic power generation based on intelligent algorithm exports according to distributed photovoltaic system inverter and the generated energy that electric energy meter is uploaded carry out contrast and judge, can accurately judge whether electricity filching behavior occurred, and the reference data that power-steeling quantity is how many; The anti-electricity-theft monitoring method of the described distributed photovoltaic power generation based on intelligent algorithm provides stealing suspicion coefficient according to stealing electrodiagnostic result, reduces suspicion scope, improves inspecting efficiency, for effective supervision of distributed photovoltaic power generation is produced evidence.

Accompanying drawing explanation

Fig. 1 is the anti-electricity-theft model flow figure that the present invention is based on BP neural network.

Fig. 2 is schematic diagram of the present invention.

Fig. 3 is stealing suspicion decision flowchart of the present invention.

Embodiment

Below in conjunction with accompanying drawing, the invention will be further described.Following examples only for technical scheme of the present invention is clearly described, and can not limit the scope of the invention with this.

As shown in Figure 1 to Figure 3, the anti-electricity-theft monitoring method of a kind of distributed photovoltaic power generation based on intelligent algorithm, is characterized in that, comprise the following steps:

Step one: set up distributed photovoltaic power generation database, database information comprise photovoltaic plant Historical Meteorological Information and the same period photovoltaic generation power data, the sample data of described database is classified according to fine day, cloudy, cloudy, rainy day four kinds of weather conditions, sample data is divided into training sample and test sample book, training sample and test sample book contain above-mentioned four kinds of weather patterns simultaneously, and the ratio that training sample and test sample book account for total sample is respectively 70% and 30%.

Step 2: take meteorological data as input quantity, with photovoltaic generation power value for exporting, adopts the anti-electricity-theft model of BP neural network, adopts the step of the anti-electricity-theft model of BP neural network as follows:

A. according to photovoltaic plant Historical Meteorological Information and the same period photovoltaic generation power data set up distributed photovoltaic power generation database, and sample data is divided into training sample and test sample book;

B. set up BP neural network, determine precision of prediction and network structure, namely determine nodes and output node number (m) of input number of nodes (n), the implicit number of plies, each hidden layer;

C. select suitable Algorithm for Training network, make it matching training sample as far as possible;

D. input training sample, calculate each layer and export and each layer error signal, if the error calculated meets the demands, verify this neural network by test sample book;

If e. error does not meet the demands, each layer weights of neural network need be adjusted, recalculate each layer error signal, until error meets the demands;

F. with the network that test sample book data detection trains, if test effect is fine, the weather information that the network integration that trains just can be utilized new is predicted photovoltaic power, if test effect is bad, the each layer weights of adjustment network, repeat step e, until obtain good assay.

Step 3: utilize anti-electricity-theft model to predict photovoltaic power, by the theoretical generated energy computation model of photovoltaic plant, and then calculates the theoretical generated energy of photovoltaic plant.Utilize the BP neural network established, by inputting the weather information such as photovoltaic plant irradiation, temperature collected, just the prediction of photovoltaic plant output power can be realized, because photovoltaic plant output power is random variation amount, and it is larger by the meteorological condition such as irradiation, temperature variable effect, single-point does not have reference value, and for improving the accuracy of anti-electricity-theft monitoring, the present invention adopts photovoltaic power station power generation amount as stealing whether distinguishing rule.By photovoltaic plant output power predicted value, the theoretical generated energy of photovoltaic generating system can be calculated.The theoretical generated energy expression formula of photovoltaic generating system is:

$W={\∫}_0^t\mathrm{Pdt}=\mathrm{Pt}$

Wherein, P is photovoltaic generation power, and W is photovoltaic power generation quantity; Namely photovoltaic power generation quantity is the accumulation of photovoltaic generation power on time t.Consider that photovoltaic power generation quantity is statistical information, error small between photovoltaic electric amount calculated value and actual electrical measurement, under the effect of accumulated time, deviation therebetween also may be caused progressively to increase, and then cause stealing to be judged by accident.Thus, the theoretical generated energy computation model of described photovoltaic plant every ten days is a cycle for setting, in this computation period, judges that whether photovoltaic plant electricity is abnormal, carry out relevant alarm if abnormal by electricity abnormality juding module; If normal, next computation period starting point is photovoltaic plant actual electrical measurement, recalculates the theoretical generated energy of photovoltaic plant on this basis.Adopt the statistical method based on neural network, avoid the concrete modeling to photovoltaic system inversion model and opto-electronic conversion model, can systematic error be reduced, the accuracy that the theoretical generated energy improving photovoltaic generating system calculates.

Step 4: real-time for the photovoltaic plant collected generated energy and the theoretical generated energy of photovoltaic plant are input to electricity abnormality juding module, photovoltaic power station power generation stealing suspicion result of determination can be obtained, and according to different stealing suspicion coefficients, carry out corresponding stealing event alarm.

Utilize distributed photovoltaic power generation electricity computation model, the theoretical generated energy of photovoltaic plant can be calculated.Consider that photovoltaic power generation quantity affects by meteorological condition larger, there is randomness and uncontrollable feature, for improving the accuracy of photovoltaic electricity abnormality juding, need to carry out uncertainty analysis, therefore stealing suspicion coefficient decision is introduced fiducial interval and is judged, the calculating of fiducial interval bound adopts sample average and standard deviation to calculate or adopt monte carlo method to obtain estimated value distribution, namely when 95% fiducial interval providing photovoltaic electricity theoretical value is [a b], can be understood as us and have the mean value of the confidence guarantee sample of 95% between a to b, and the probability made a mistake is 5%.The calculating of fiducial interval bound adopts sample average and standard deviation to calculate, account form is: when known sample average (M) and standard deviation (ST), lower limit of confidence interval: a=M-n*ST, the fiducial interval upper limit: b=M+n*ST; The n=1.645 when asking for 90% fiducial interval; The n=1.96 when asking for 95% fiducial interval; The n=2.576 when asking for 99% fiducial interval.

Photovoltaic generation metering ammeter is mounted in the photovoltaic power generation quantity clearing ammeter of photovoltaic DC-to-AC converter outlet, the photovoltaic power generation quantity that this table is uploaded is exactly the generated energy of this photovoltaic plant of metering, described electricity abnormality juding module is judge compared with the fiducial interval bound of photovoltaic theoretical capacity according to photovoltaic meter volume and electricity, if photovoltaic meter volume and electricity is less than the lower limit of confidence interval of photovoltaic theoretical capacity, then judge that its generating is abnormal, provide generating abnormality alarming; If photovoltaic meter volume and electricity is greater than the fiducial interval upper limit of photovoltaic theoretical capacity, then it is determined that the presence of stealing suspicion.By the collection value of distributed photovoltaic power station generated energy and comparing of this fiducial interval upper limit, obtain different stealing suspicion coefficients, thus improve the specific aim of inspection, improve inspecting efficiency.

Described stealing suspicion coefficient is that to get Δ be the number percent that difference that photovoltaic meter volume and electricity is greater than the fiducial interval upper limit of photovoltaic theoretical capacity accounts for the fiducial interval upper limit of photovoltaic theoretical capacity, namely

As 10%> Δ >0, stealing suspicion coefficient is 0.1; When 20%> Δ >=10%, stealing suspicion coefficient is 0.2; When 30%> Δ >=20%, stealing suspicion coefficient is 0.3; When 40%> Δ >=30%, stealing suspicion coefficient is 0.4; When 50%> Δ >=40%, stealing suspicion coefficient is 0.5; When 60%> Δ >=50%, stealing suspicion coefficient is 0.6; When 70%> Δ >=60%, stealing suspicion coefficient is 0.7; When 80%> Δ >=70%, stealing suspicion coefficient is 0.8; When >=80%, stealing suspicion coefficient is 0.9.According to different stealing suspicion coefficients, carry out corresponding stealing event alarm, simultaneously stealing suspicion coefficient is more close to 1, illustrates that stealing possibility is larger, and then can using user large for this month stealing suspicion coefficient as investigating object.

The above is only the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the prerequisite not departing from the technology of the present invention principle; can also make some improvement and distortion, these improve and distortion also should be considered as protection scope of the present invention.

Claims (9)

1., based on the anti-electricity-theft monitoring method of distributed photovoltaic power generation of intelligent algorithm, it is characterized in that, comprise the following steps:

Step one: set up distributed photovoltaic power generation database, database information comprise photovoltaic plant Historical Meteorological Information and the same period photovoltaic generation power data.

Step 2: take meteorological data as input quantity, with photovoltaic generation power value for exporting, adopts intelligent algorithm to set up anti-electricity-theft model.

Step 3: utilize anti-electricity-theft model to predict photovoltaic power, by the theoretical generated energy computation model of photovoltaic plant, and then calculates the theoretical generated energy of photovoltaic plant.

Step 4: real-time for the photovoltaic plant collected generated energy and the theoretical generated energy of photovoltaic plant are input to electricity abnormality juding module, photovoltaic power station power generation stealing suspicion result of determination can be obtained, and according to different stealing suspicion coefficients, carry out corresponding stealing event alarm.

2. the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm according to claim 1, it is characterized in that, the sample data of described database is classified according to fine day, cloudy, cloudy, rainy day four kinds of weather conditions, sample data is divided into training sample and test sample book, training sample and test sample book contain above-mentioned four kinds of weather patterns simultaneously, and the ratio that training sample and test sample book account for total sample is respectively 70% and 30%.

3. the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm according to claim 1, it is characterized in that, the theoretical generated energy computation model of described photovoltaic plant every ten days is a cycle for setting, in this computation period, judge that whether photovoltaic plant electricity is abnormal by electricity abnormality juding module, carry out relevant alarm if abnormal; If normal, next computation period starting point is photovoltaic plant actual electrical measurement, recalculates the theoretical generated energy of photovoltaic plant on this basis.

4. the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm according to claim 1, it is characterized in that, stealing suspicion coefficient decision introduces fiducial interval, and the calculating of fiducial interval bound adopts sample average and standard deviation to calculate or adopt monte carlo method to obtain estimated value distribution.

5. the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm according to claim 4, it is characterized in that, the account form that fiducial interval adopts sample average and standard deviation to calculate is: when known sample average and standard deviation, lower limit of confidence interval: a=M-n*ST, the fiducial interval upper limit: b=M+n*ST; The n=1.645 when asking for 90% fiducial interval; The n=1.96 when asking for 95% fiducial interval; The n=2.576 when asking for 99% fiducial interval.

6. the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm according to claim 1, it is characterized in that, described electricity abnormality juding module is judge compared with the fiducial interval bound of photovoltaic theoretical capacity according to photovoltaic meter volume and electricity, if photovoltaic meter volume and electricity is less than the lower limit of confidence interval of photovoltaic theoretical capacity, then judge that its generating is abnormal, provide generating abnormality alarming; If photovoltaic meter volume and electricity is greater than the fiducial interval upper limit of photovoltaic theoretical capacity, then it is determined that the presence of stealing suspicion.

7. the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm according to claim 1, it is characterized in that, described stealing suspicion coefficient is that to get Δ be the number percent that difference that photovoltaic meter volume and electricity is greater than the fiducial interval upper limit of photovoltaic theoretical capacity accounts for the fiducial interval upper limit of photovoltaic theoretical capacity, namely

As 10%> Δ >0, stealing suspicion coefficient is 0.1; When 20%> Δ >=10%, stealing suspicion coefficient is 0.2; When 30%> Δ >=20%, stealing suspicion coefficient is 0.3; When 40%> Δ >=30%, stealing suspicion coefficient is 0.4; When 50%> Δ >=40%, stealing suspicion coefficient is 0.5; When 60%> Δ >=50%, stealing suspicion coefficient is 0.6; When 70%> Δ >=60%, stealing suspicion coefficient is 0.7; When 80%> Δ >=70%, stealing suspicion coefficient is 0.8; When >=80%, stealing suspicion coefficient is 0.9.

8. the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm according to claim 1, is characterized in that, described intelligent algorithm comprises any one algorithm in BP neural network, data mining technology, support vector machine.

9. the anti-electricity-theft monitoring method of the distributed photovoltaic power generation based on intelligent algorithm according to claim 8, is characterized in that, adopts the step of the anti-electricity-theft model of BP neural network as follows:

A. according to photovoltaic plant Historical Meteorological Information and the same period photovoltaic generation power data set up distributed photovoltaic power generation database, and sample data is divided into training sample and test sample book;

B. set up BP neural network, determine precision of prediction and network structure, namely determine nodes and the output node number of input number of nodes, the implicit number of plies, each hidden layer;

C. select suitable Algorithm for Training network, make it matching training sample as far as possible;

D. input training sample, calculate each layer and export and each layer error signal, if the error calculated meets the demands, verify this neural network by test sample book;

If e. error does not meet the demands, each layer weights of neural network need be adjusted, recalculate each layer error signal, until error meets the demands;

F. with the network that test sample book data detection trains, if test effect is fine, the weather information that the network integration that trains just can be utilized new is predicted photovoltaic power, if test effect is bad, the each layer weights of adjustment network, repeat step e, until obtain good assay.