KR101398413B1 - Method for learning and calculating the number of received cases of electrical fault using weather data and neural networks - Google Patents

Abstract

The present invention relates to a method for learning and calculating a number of received power failure cases using weather data and a neural network and, more specifically, to a method for automatically calculating daily predictive number of received failure cases of electrical equipment, and performing calculation learning of a predictive number of received failure cases by inputting daily historical and forecast weather data and daily number of received past failures of electrical equipment into a neural network. According to the present invention, an appropriate number of personnel to be dispatched according to the received failure cases can stand-by by calculating the number of received failure cases closely related to the weather data by using the past and forecast weather data.

Description

[0001] METHOD FOR LEARNING AND CALCULATING ELECTRIC FAULT USING WEATHER DATA AND NEURAL NETWORKS [0002]

More particularly, the present invention relates to a method and apparatus for learning and calculating the number of received electrical faults by using weather data and neural networks, The present invention relates to a method capable of automatically calculating the number of forecasted failures per day of an electrical installation.

Related to the apparatus for detecting the state of the electric equipment, there are many applications and disclosures in addition to Korean Patent Laid-Open No. 10-2012-0016423 (hereinafter referred to as "prior art"). In the above-mentioned prior art, an ultrasonic wave is applied from an ultrasonic sensor array remotely, an electric signal is extracted, and an abnormal state of the electric power facility is detected by analyzing a signal component.

On the other hand, faults in electrical equipment have a close relationship with precipitation. For example, in the third quarter, where rainfall is highest in one year, resistive leakage currents frequently occur in electrical equipment, which may cause electrical equipment to fail.

In addition, waiters who are waiting for dispatch by complaints about electrical equipment failure are waiting for one person in each region and dispatching them in case of an emergency electrician accident, thereby preventing electric accidents. However, if a lot of faults occur, one waiting person is dispatched and the accident is safely handled. However, there may be cases where an additional electric accidents complaints occur and the treatment is delayed or the accident can not be handled. In order to solve such problems, it is possible to process complaints by waiting for a large number of dispatched personnel. However, when a large number of daytime workers are arranged at night, the number of people who can perform weekly tasks is small, Manpower management problems arise. Therefore, since one person is selected and operated by the local dispatch work force, the summer heavy rainfall that causes many electric accidents is delayed due to a large number of notifications, the delay of proper accident response, And the like.

Although there is a technology for detecting a failure by detecting a state of an electric facility as in the above-mentioned prior art, there is no technology for predicting and calculating the number of failures received by an electric facility using weather data such as precipitation.

SUMMARY OF THE INVENTION The present invention has been conceived in view of the above problems, and it is an object of the present invention to provide a neural network in which the number of past failures received by each of the past, And a method for learning and calculating the number of received electrical faults using weather data and a neural network that automatically calculates the number of predicted faults received.

According to an aspect of the present invention, there is provided a method for learning and calculating a number of received electrical faults using weather data and a neural network, the method comprising the steps of: (a) acquiring past past and predicted weather data from a weather data server; (b) normalizing the past weather data acquired by the neural network learning and calculation unit through the data acquisition unit and the size of predicted vapor data per day; (c) determining whether the learning signal of the neural network learning and calculating unit is a learning control signal of a user, the control signal input through the input unit; (d) a learning module of the neural network learning and calculating unit inputs the past weather data normalized by the norming module to the input terminal of the neural network, as a result of the determination of step (c) Inputting the number of past failures received by the date stored in the storage unit to the output terminal of the neural network; (e) adjusting a weight for learning by a learning module of the neural network learning and calculating unit; (f) generating a recalculated normalized past meteorological data by recalculating normalized meteorological past meteorological data by applying an adjusted weight to a learning module of the neural network learning and calculation unit; (g) estimating the number of failures received per day based on the recalculated normalized past meteorological data by the learning module of the neural network learning and calculating unit; (h) determining whether the estimated number of failure acceptances per day by the learning module of the neural network learning and calculating unit is all the same as the number of past failure acceptances by the number inputted to the output terminal of the neural network; And (i) ending the learning by the learning module of the neural network learning and calculation unit when the determination result of step (h) is all the same; .

(J) if the learning module of the neural network learning and calculating unit does not perform the steps of (e) as a result of the determination of step (h); And a control unit.

(K) determining whether the control signal input by the learning module of the neural network learning and calculating unit is a calculation control signal of the user, if the control signal is not a learning control signal as a result of the determining in the step (c); (l) if the calculation control signal is a result of the determination in step (k), the calculating module of the neural network learning and calculating unit calculates the predicted vapor data normalized by the normizing module in step (b) ; (m) a calculation module of the neural network learning and calculating unit recalculates the normalized forecasted vapor data by applying the weights corresponding to the inputted standardized predicted vapor data of each day, ≪ / RTI > And (n) calculating and outputting the number of predicted failure acceptances per day based on the recalculated normalized predicted day-by-day weather data of the neural network learning and calculating unit; And a control unit.

And the past weather data acquired through the data acquisition unit and the number of past failure acceptances by date stored in the storage unit are data for the same period.

According to the present invention as described above, since the number of trouble receipts closely related to the weather data is calculated by using past and predicted vapor data, , It is possible to appropriately wait for the dispatching personnel according to the fault reception.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic view showing conceptually a device for learning and calculating the number of received electric faults using weather data and a neural network according to the present invention; FIG.
2 is a schematic diagram of a neural network having an input layer unit, a hidden layer unit and an output layer according to the present invention.
3 is an overall flowchart of a method for learning and calculating the number of received electrical failures using weather data and a neural network according to the present invention.

Specific features and advantages of the present invention will become more apparent from the following detailed description based on the accompanying drawings. It is to be noted that the detailed description of known functions and constructions related to the present invention is omitted when it is determined that the gist of the present invention may be unnecessarily blurred.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will now be described in detail with reference to the accompanying drawings.

1 and 2, a description will be made of an apparatus for learning and calculating the number of received electric faults using the weather data and the neural network according to the present invention.

1 is a block diagram conceptually showing an apparatus for learning and calculating the number of received electrical faults using the weather data and the neural network according to the present invention. As shown in the figure, the data acquisition unit 100, the storage unit 200, A network learning and calculating unit 300 and an input unit 400.

The data acquisition unit 100 acquires past past and predicted weather data from the weather station server 10.

At this time, the past weather data includes data of the immediately preceding one year, and the predicted weather data includes data such as the predicted precipitation amount by day for one month from now.

In the present embodiment, the past weather data is set as the data of the immediately preceding one year, and the predicted weather data is set as the data of one month from now. However, the present invention is not limited to this period, May be received and used. The past and predicted weather data may also include data such as rain probability and temperature.

The storage unit 200 stores the number of past failures received for each electric facility. At this time, the number of failure acceptances stored in the storage unit 200 may be the number of failures received per day of the immediately preceding one year like the data acquisition unit 100.

In addition, the storage unit 200 stores past past and predicted vapor data acquired through the data acquisition unit 100, data calculated through the neural network learning and calculation unit 300, and the number of calculated predictive failure acceptances. Accordingly, the neural network learning and calculation unit 300 may refer to the data stored in the storage unit 200. [

The neural network learning and calculation unit 300 normalizes the past and predicted vapor data acquired by the data acquisition unit 100 and inputs the past and predicted vapor data to the neural network 20 to calculate the number of failure acceptance And performs a function to calculate the number of forecasted failure acceptance by day, and includes a normalization module 310, a learning module 320, and a calculation module 330 as shown in FIG.

Specifically, the normalization module 310 calculates the size of the predicted weather data including the past weather data including the immediately preceding one-year precipitation amount acquired through the data acquisition unit 100 and the one-month predicted precipitation amount from the present Normalize.

The learning module 320 inputs the past weather data normalized by the norming module 310 to the input terminal of the neural network 20 according to the learning control signal of the user inputted through the input unit 400, The number of past failures received by the date stored in the storage unit 200 is input to the output terminal of the neural network 20. [

Thereafter, the learning module 320 generates the recalculated standardized past meteorological data by adjusting the weights for learning and recalculating the standardized meteorological data for each standard by applying the adjusted weights, Based on the recalculated normalized day - to - day meteorological data, we estimate the number of breakdowns by day.

delete

delete

delete

delete

delete

delete

Then, the learning module 320 determines whether the estimated number of received failures per day is equal to the number of received past failures per day inputted to the output terminal of the neural network 20.

If the results of the judgment are all the same, the learning is terminated. If not identical, the weight adjustment, the recalculation of the normized data, and the estimation of the number of failures are repeated until the same, The corresponding weight value is recognized.

Through such learning, when the standardized data is input, the number of predicted failure acceptances per one day of the electric equipment can be automatically calculated and output by using the learned and recognized normalized data and corresponding weights.

That is, in accordance with the user's control signal inputted through the input unit 400, the calculation module 330 inputs the predicted vapor-phase data normalized through the normalization module 310 to the input terminal of the neural network 20 And recalculates the normalized forecasted weather data by applying the weights corresponding to the inputted standardized forecasted weather data to generate the recalculated standardized forecasted weather data for each day to generate the recalculated standardized daily data Based on the predicted meteorological data, the number of predicted failure receipts per day of the electrical equipment is calculated and output.

delete

delete

delete

delete

delete

delete

)가

를 거쳐

를 출력하게 되며, 학습을 수행할 수 있다.
FIG. 2 shows a structure of a neural network 20 having an input layer unit, a hidden layer unit and an output layer according to the present invention. As described above, signal(

)end

Through

And the learning can be performed.

The input unit 400 receives the learning control signal and the calculation control signal of the user.

The input unit 400 may receive various control signals of the user regarding calculation, storage, output, and the like.

Hereinafter, a method of learning and calculating the number of received electrical failures using the weather data and the neural network using the above-described apparatus will be described with reference to FIG.

FIG. 3 is an overall flowchart of a method for learning and calculating the number of received electrical failures using the weather data and the neural network according to the present invention. As shown in FIG. 3, the data acquisition unit 100 receives data from the meteorological office server 10, Data is acquired (S10). At this time, the past weather data includes data of the immediately preceding one year, and the predicted weather data includes data such as the predicted precipitation amount by day for one month from now.

Thereafter, the normalization module 310 of the neural network learning and calculation unit 300 acquires the data acquired by the data acquisition unit 100 (S20) the past meteorological data including the immediately preceding one-year precipitation amount and the predicted meteorological data including the one-month forecasted precipitation amount from the present one.

In addition, the learning module 320 of the neural network learning and calculation unit 300 determines whether the control signal input through the input unit 400 is a learning control signal of the user (S30).

As a result of the determination in step S30, if the learning control signal is a learning control signal, the learning module 320 inputs the past weather data normalized by the normizing module 310 to the input terminal of the neural network 20, 200 to the output terminal of the neural network 20 (S40).

Thereafter, the learning module 320 adjusts a weight for learning (S50), and recalculates the recalculated past meteorological data based on the standardized meteorological data by applying the adjusted weight to the data (S60), and estimates the number of failures received per day based on the recalculated normalized past weather data (S70).

In step S80, the learning module 320 determines whether the estimated number of received failures per day is equal to the number of received past failures per day input to the output terminal of the neural network 20 (S80).

As a result of the determination in step S80, if all are the same, the storage unit 200 stores past and predicted vapor data acquired by the data acquisition unit 100, data calculated through the neural network learning and calculation unit 300, and the like (S90), and the learning module 320 ends the learning.

As a result of the determination in step S80, if not, the learning module 320 proceeds to step S50. That is, by performing the weight adjustment, the recalculation of the normalized data, and the estimation of the number of received failures repeatedly until all are the same, the inputted normative data and the corresponding weight are recognized according to the number of received failures.

Through such learning, when the standardized data is input, the number of predicted failure acceptances per one day of the electric equipment can be automatically calculated and output by using the learned and recognized normalized data and corresponding weights.

On the other hand, if it is determined in operation S30 that the control signal is not a learning control signal, the learning module 320 determines whether the input control signal is a user's control signal (S100).

As a result of the determination in step S100, in the case of the calculation control signal, the calculation module 330 of the neural network learning and calculation unit 300 calculates the predicted vapor-phase data normalized through the normalization module 310 in step S20 (S110), and the weighted value corresponding to the inputted standardized predicted vapor data for each day is applied to recalculate the normalized predicted vapor data for each day to thereby calculate the recalculated standardized day-by- (S120), calculates and outputs the number of predicted failure acceptance by day based on the recalculated normalized predicted day-by-day meteorological data (S130), and executes the procedure to step S90.

On the other hand, if the result of the determination in step S100 is negative, the learning module 320 ends the process.

While the present invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. It will be appreciated by those skilled in the art that numerous changes and modifications may be made without departing from the invention. Accordingly, all such appropriate modifications and changes, and equivalents thereof, should be regarded as within the scope of the present invention.

100: Data acquisition unit 200:
300: Neural network learning and calculation part 400: Input part
310: Normalizing module 320: Learning module
330: Output module 10: weather station server
20: Neural network

Claims (4)

In the method of learning and calculating the number of electrical faults received using weather data and neural networks,
(a) the data acquisition unit (100) acquiring historical past and predicted weather data from the weather station server (10);
(b) normalizing the size of the past weather data obtained by the normalization module 310 of the neural network learning and calculation unit 300 through the data acquisition unit 100 and the predicted vapor data by day;
(c) determining whether the control signal input through the input unit 400 of the learning module 320 of the neural network learning and calculation unit 300 is a learning control signal of the user;
(d) If it is determined in the step (c) that the learning module 320 of the neural network learning and calculating unit 300 receives the normalized past meteorological data through the norming module 310, Inputting the data to the input of the neural network 20 and inputting the number of past failure acceptances stored in the storage unit 200 to the output terminal of the neural network 20;
(e) adjusting a weight for learning by the learning module 320 of the neural network learning and calculating unit 300;
(f) The learning module 320 of the neural network learning and calculation unit 300 recalculates the normalized past meteorological data by applying the adjusted weights to generate the recalculated metricized past meteorological data of the past step;
(g) estimating the number of failures received per day based on the recalculated standardized past weather data of the neural network learning and calculating unit 300;
(h) The learning module 320 of the neural network learning and calculation unit 300 determines whether the estimated number of received failures per day is equal to the number of received past failures per day input to the output terminal of the neural network 20 ; And
(i) ending the learning by the learning module 320 of the neural network learning and calculation unit 300 if all of the results are the same as the result of the determination of step (h); And the method of learning and calculating the number of received electrical failures using the neural network.
The method according to claim 1,
(j) if the learning module 320 of the neural network learning and calculation unit 300 does not perform the steps of (e) as a result of the determination in step (h); And calculating and counting the number of received electrical failures using the weather data and the neural network.
The method according to claim 1,
(k) If the learning module 320 of the neural network learning and calculation unit 300 determines that the input control signal is not a learning control signal as a result of the determination in step (c) ;
(1) If the calculation control signal is a result of the determination in step (k), the calculation module 330 of the neural network learning and calculation unit 300 performs normalization Inputting the predicted day-by-day weather data to the input of the neural network 20;
(m) The calculation module 330 of the neural network learning and calculation unit 300 recalculates the normalized forecasted gas-phase data by recalculating the normalized predicted gas-phase data by applying the weights corresponding to the inputted standardized predicted day-by-day gas- Generating normalized predicted day-to-day meteorological data; And
(n) calculating and outputting the number of predicted failure acceptance by day based on the recalculated normalized predicted day-by-day weather data of the neural network learning and calculation unit 300; And calculating and counting the number of received electrical failures using the weather data and the neural network.
The method according to claim 1,
Wherein the past weather data acquired by the data acquisition unit (100) and the number of past failures received by each store stored in the storage unit (200) are data for the same period. Learning method and calculation method.

Images