A voltage sag event prediction method, medium and system for sensitive users

Abstract

The present invention discloses a method, medium and system for predicting voltage sag events for sensitive users, including: obtaining a power supply bus of sensitive users; obtaining a first historical parameter of each voltage sag event occurring on the power supply bus within a preset time period, and preprocessing to obtain a first training sample; obtaining a second historical parameter of the normal daily operation of the power supply bus within the preset time period, and preprocessing to obtain a second training sample; using the first training sample and the second training sample to train a deep neural network DNN to obtain a trained prediction network; preprocessing the collected prediction parameters and inputting them into the prediction network to obtain a voltage prediction amplitude; if the voltage prediction amplitude is less than a first preset threshold, predicting the occurrence of a voltage sag event to warn the operation and maintenance personnel. The present invention can more accurately predict whether a voltage sag event occurs, so that a warning can be given in a timely manner.

Description

Technical Field

[0001] The present invention relates to the field of voltage sag technology, and in particular to a voltage sag event prediction method, medium and system for sensitive users. Background Art

[0002] With the increasing number of precision equipment and sensitive users in modern power systems, higher requirements are placed on power quality. Among various power quality problems, accidents caused by voltage sags account for more than 80% of the total accidents, and almost every sag event will cause huge economic losses. Voltage sag risk warning is one of the important tasks to mitigate voltage sags. Accurately predicting the amplitude of voltage sags has important theoretical value and practical significance for reducing losses; at the same time, it is conducive to mitigating the impact of voltage sags and reasonably arranging production plans. Summary of the invention

[0003] The embodiments of the present invention provide a voltage sag event prediction method, medium and system for sensitive users, so as to solve the problem that the prior art cannot accurately predict voltage sag events.

[0004] In a first aspect, a voltage sag event prediction method for sensitive users is provided, comprising:

[0005] Obtain the power supply busbar of sensitive users;

[0006] Obtaining first historical parameters of each voltage sag event on the power supply bus within a preset time period, and performing preprocessing to obtain a first training sample, wherein the first historical parameters include: a first historical time, a first historical date, a historical voltage sag amplitude, a first historical weather, and an average economic loss value of sensitive users caused by the voltage sag event within the preset time period;

[0007] Acquire the second historical parameters of the normal daily operation of the power supply bus within the preset time period, and perform preprocessing to obtain a second training sample, wherein the second historical parameters include: a second historical time, a second historical date, a historical voltage amplitude, a second historical weather, and an economic loss value;

[0008] Using the first training sample and the second training sample to train a deep neural network DNN to obtain a trained prediction network;

[0009] Preprocessing the collected prediction parameters and inputting them into the prediction network to obtain the voltage prediction amplitude, wherein the prediction parameters include: prediction time, prediction date, and prediction weather;

[0010] If the voltage prediction amplitude is less than the first preset threshold, a voltage sag event is predicted to occur, so as to warn the operation and maintenance personnel.

[0011] In a second aspect, a computer-readable storage medium is provided, on which computer program instructions are stored; when the computer program instructions are executed by a processor, the voltage sag event prediction method for sensitive users as described in the embodiment of the first aspect above is implemented.

[0012] In a third aspect, a voltage sag event prediction system for sensitive users is provided, comprising: a computer-readable storage medium as described in the above-mentioned second aspect embodiment.

[0013] In this way, the embodiment of the present invention uses appropriate parameters to train the deep neural network, so that the trained deep neural network can more accurately predict whether a voltage sag event occurs, so that an early warning can be issued in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0015] Figure 1 It is a flow chart of a voltage sag event prediction method for sensitive users according to an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] The embodiment of the present invention discloses a voltage sag event prediction method for sensitive users. The sensitive users in the embodiment of the present invention refer to users who are easily affected by voltage changes. Figure 1 As shown, the method comprises the following steps:

[0018] Step S1: Obtain the power supply bus of sensitive users.

[0019] Step S2: obtaining first historical parameters of each voltage sag event occurring on the power supply bus within a preset time period, and performing preprocessing to obtain a first training sample.

[0020] The first historical parameter includes: a first historical time, a first historical date, a historical voltage sag amplitude, a first historical weather, and an average economic loss value of sensitive users caused by a voltage sag event within a preset time period.

[0021] The first historical time, the first historical date, and the historical voltage sag amplitude can be obtained from the power quality monitoring system. The first historical weather can be queried from the weather system. The first historical weather includes two types of air humidity and cloudy and sunny data, among which the cloudy and sunny data includes three types: sunny, cloudy, and thunderstorm. The average economic loss value caused by the voltage sag event in the preset time period to sensitive users is obtained by calculating the average economic loss caused by all voltage sag events in the preset time period. The economic loss caused by each voltage sag event is obtained by statistics after the voltage sag event occurs.

[0022] Depending on the type of data, the preprocessing method is different, as follows:

[0023] (1) For the first historical time, the preprocessing includes: converting the time into minutes, and if the seconds corresponding to the time exceed 30 seconds, adding 1 to the minutes.

[0024] For example, let the hour hand scale value be m, the minute hand scale value be n, and the preprocessed time value be x, then x=m*60+n, and if the second hand scale value is greater than 30, then x=m*60+n+1.

[0025] (2) For the first historical date, the preprocessing includes: calculating the number of days between the date and January 1 of the current year.

[0026] For example, if the date is March 1 of the current year, the number of days between March 1 and January 1 is calculated as the preprocessed date.

[0027] (3) For the historical voltage sag amplitude, the preprocessing includes: standardizing the amplitude to a per-unit value.

[0028] Specifically, the calculation formula for per-unit value processing is:

[0029] Among them, U * It represents the per unit value, U represents the historical voltage sag amplitude, that is, the effective value of the voltage after the voltage sag event occurs on the power supply bus, and U B Indicates the supply voltage level of the power supply bus.

[0030] (5) For the first historical weather, the preprocessing includes: normalizing the air humidity in the weather so that the air humidity is in the interval [0, 1]; and dividing the cloudy and sunny data in the weather into three types of weather: sunny, cloudy, and thunderstorm, and assigning a value of 1 to sunny, 2 to cloudy, and 3 to thunderstorm.

[0031] Therefore, the first training sample S1 is a five-dimensional vector consisting of the first historical time, the first historical date, the historical voltage sag amplitude, the first historical weather, and the average economic loss value of sensitive users caused by the voltage sag event within a preset time period after preprocessing, and the label value L1 of the first training sample is the historical voltage sag amplitude.

[0032] Step S3: obtaining a second historical parameter of the normal daily operation of the power supply bus within a preset time period, and performing preprocessing to obtain a second training sample.

[0033] The second historical parameters include: second historical time, second historical date, historical voltage amplitude, second historical weather, and economic loss value.

[0034] The second historical time, the second historical date, and the historical voltage amplitude can be obtained from the power quality monitoring system. The second historical weather can be obtained from the weather system. The second historical weather also includes two types of air humidity and cloudy and sunny data. Among them, cloudy and sunny data includes three types: sunny, cloudy, and thunderstorm. Since it is normal operation, the economic loss value is 0.

[0035] For the second historical time, the second historical date, and the second historical weather, the preprocessing methods are the same as those for the first historical time, the first historical date, and the first historical weather, and are not described in detail here. For the historical voltage amplitude, the same calculation formula as the historical voltage sag amplitude is also used. The difference is that, at this time, U represents the historical voltage amplitude, that is, the effective value of the voltage of the power supply bus in normal daily operation.

[0036] Therefore, the second training sample S2 is a five-dimensional vector consisting of the second historical time, the second historical date, the historical voltage amplitude, the second historical weather, and the economic loss value (value is 0) after preprocessing, and the label value L2 of the second training sample is the historical voltage sag amplitude.

[0037] Step S4: Use the first training sample and the second training sample to train the deep neural network DNN to obtain a trained prediction network.

[0038] The deep neural network DNN is an existing technology, which is a four-layer neural network structure with a five-dimensional vector as input. During training, the input value of the deep neural network is the first training sample S1 and the second training sample S2, and the output value is the voltage prediction amplitude of the power supply bus connected to the sensitive user. The intermediate output result of the neural network is expressed as follows:

[0039]

[0040] Among them, S n-1 represents the number of neurons in the n-1th layer, Wij n represents the weight value between the i-th neuron in the n-1th layer and the j-th neuron in the nth layer, a i n-1 represents the response value of the i-th neuron in the n-1th layer, b i n Represents the bias value of the i-th neuron in the n-th layer.

[0041] The neuron activation function of this neural network is as follows:

[0042]

[0043] In this step, the first training samples and the second training samples are randomly divided into a training set and a validation set with a preset sample ratio. The preset sample ratio can be set based on experience, for example, the preset sample ratio is 8:2, that is, 80% of all the first training samples and the second training samples are used as training sets and 20% as validation sets.

[0044] In this step, when the loss function value calculated by the prediction value output by the deep neural network DNN for a consecutive preset number of times is not less than the second preset threshold, the training is completed.

[0045] The loss function value is calculated as follows:

[0046]

[0047] Among them, loss represents the loss function value, h i represents the predicted value of the i-th training sample, G i represents the true value of the i-th training sample. The preset number and the second preset threshold can be set based on experience.

[0048] In this way, when the loss function value gradually stabilizes, the training can be considered completed.

[0049] Step S5: pre-process the collected prediction parameters and input them into the prediction network to obtain the voltage prediction amplitude.

[0050] Among them, the prediction parameters include: prediction time, prediction date, and prediction weather. The prediction parameters can be currently collected or in the future, and the predicted weather can be the weather in the weather forecast. The predicted weather also includes two types of air humidity and cloudy and sunny data, among which the cloudy and sunny data includes three types: sunny, cloudy, and thunderstorm. It should be understood that for the prediction time, prediction date, and prediction weather, the preprocessing methods are the same as the aforementioned first historical time, first historical date, and first historical weather, respectively, and will not be repeated here.

[0051] Step S6: If the predicted voltage amplitude is less than the first preset threshold, a voltage sag event is predicted to occur, so as to warn the operation and maintenance personnel.

[0052] The first preset threshold value can be set based on experience. The first preset threshold value of the embodiment of the present invention is 0.9 times the rated voltage amplitude of the power supply bus. If the voltage prediction amplitude is less than the first preset threshold value, a voltage sag event will occur, which will affect the sensitive user.

[0053] An embodiment of the present invention further discloses a computer-readable storage medium having computer program instructions stored thereon; when the computer program instructions are executed by a processor, the voltage sag event prediction method for sensitive users as described in the above embodiment is implemented.

[0054] The embodiment of the present invention further discloses a voltage sag event prediction system for sensitive users, comprising: a computer-readable storage medium as described in the above embodiment.

[0055] In summary, the embodiments of the present invention use appropriate parameters to train a deep neural network, so that the trained deep neural network can more accurately predict whether a voltage sag event occurs, so that an early warning can be issued in a timely manner.

[0056] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A voltage sag event prediction method for sensitive users, characterized in that: include: Obtain the power supply busbar of sensitive users; Obtaining first historical parameters of each voltage sag event on the power supply bus within a preset time period, and performing preprocessing to obtain a first training sample, wherein the first historical parameters include: a first historical time, a first historical date, a historical voltage sag amplitude, a first historical weather, and an average economic loss value of sensitive users caused by the voltage sag event within the preset time period; Acquire the second historical parameters of the normal daily operation of the power supply bus within the preset time period, and perform preprocessing to obtain a second training sample, wherein the second historical parameters include: a second historical time, a second historical date, a historical voltage amplitude, a second historical weather, and an economic loss value; Using the first training sample and the second training sample to train a deep neural network DNN to obtain a trained prediction network; Preprocessing the collected prediction parameters and inputting them into the prediction network to obtain the voltage prediction amplitude, wherein the prediction parameters include: prediction time, prediction date, and prediction weather; If the voltage prediction amplitude is less than the first preset threshold, a voltage sag event is predicted to occur, so as to warn the operation and maintenance personnel; For the historical voltage sag amplitude and the historical voltage amplitude, the preprocessing includes: The amplitude is normalized to per unit value; The calculation formula for the per-unit value processing is: ; Indicates per unit value, Indicates the power supply voltage level of the power supply bus; When the historical voltage sag amplitude is preprocessed, U Indicates the historical voltage sag amplitude; When the historical voltage amplitude is preprocessed, U Indicates the historical voltage amplitude; In the step of training the deep neural network DNN, when the loss function value calculated by the prediction value outputted by the deep neural network DNN for a consecutive preset number of times is not less than a second preset threshold, the training is completed; Among them, the calculation formula of the loss function value is: , loss represents the loss function value, h i Indicates i The predicted value of the training samples, G i Indicates i The true value of the training samples.

2. The voltage sag event prediction method for sensitive users according to claim 1, characterized in that: For the first historical time, the second historical time, and the predicted time, the preprocessing includes: Convert the time to minutes and if the seconds of the time exceed 30s, add 1 to the minutes.

3. The voltage sag event prediction method for sensitive users according to claim 1, characterized in that: For the first historical date, the second historical date and the predicted date, the preprocessing includes: Calculates the number of days between a date and January 1 of the current year.

4. The voltage sag event prediction method for sensitive users according to claim 1, characterized in that: For the first historical weather, the second historical weather and the predicted weather, the preprocessing includes: Normalize the air humidity in the weather so that the air humidity is in the interval [0, 1]; The weather data is divided into three types: sunny, cloudy and thunderstorm. The sunny day is assigned a value of 1, the cloudy day is assigned a value of 2, and the thunderstorm is assigned a value of 3.

5. The voltage sag event prediction method for sensitive users according to claim 1, characterized in that: The first training sample is a five-dimensional vector consisting of the first historical time, the first historical date, the historical voltage sag amplitude, the first historical weather, and the average economic loss value of sensitive users caused by the voltage sag event in the preset time period after preprocessing, and the label value of the first training sample is the historical voltage sag amplitude; The second training sample is a five-dimensional vector consisting of the second historical time, the second historical date, the historical voltage amplitude, the second historical weather, and the economic loss value after preprocessing, and the label value of the second training sample is the historical voltage sag amplitude, and the economic loss value is 0.

6. The voltage sag event prediction method for sensitive users according to claim 1, characterized in that: In the step of training the deep neural network DNN, the first training sample and the second training sample are randomly divided into a training set and a validation set with a preset sample ratio.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by the processor, the voltage sag event prediction method for sensitive users according to any one of claims 1 to 6 is implemented.

8. A voltage sag event prediction system for sensitive users, characterized in that: include: The computer readable storage medium of claim 7.

Images