A power consumption safety diagnosis method based on a convolutional neural network

Abstract

The application provides a power consumption safety diagnosis method based on a convolutional neural network, a denoising correction module is constructed to perform denoising correction processing on a temperature value at a current diagnosis moment to obtain a corrected temperature value at the current diagnosis moment; a comparison module is constructed to compare the corrected temperature value at the current diagnosis moment with a preset temperature threshold to generate a selection signal; a preprocessing module is constructed to perform contrast enhancement on a target region image to generate a target region enhanced image; a feature extraction module is constructed to select a high-receptive-field feature extraction channel or a low-receptive-field feature extraction channel according to the selection signal, perform feature extraction on the target region enhanced image, and generate target features; and a classifier is constructed to classify the target features to generate a diagnosis result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of smart power utilization, in particular to a power utilization safety diagnosis method based on a convolutional neural network. BACKGROUND

[0002] In recent years, with the addition of more and more power utilization equipment, the probability of electrical fire occurrence is also increasing, and the problem of safe power utilization is becoming increasingly important. The existing technology mainly uses wireless sensing technology such as Internet of Things and mobile terminal to diagnose the running condition of the monitored line or the environmental condition of the monitored scene in real time, analyzes the changes of electrical parameters, temperature, smoke and combustible gas concentration, and identifies the possible fire or the fire about to occur, so as to improve the level of power utilization safety.

[0003] However, the electrical parameters, temperature, smoke and combustible gas concentration in the existing diagnosis method are easily affected by the surrounding electronic noise and environment, and the false positive rate is extremely high. Compared with the wireless sensing technology, the power utilization safety diagnosis method based on visual information has the advantages of shorter response, wider monitoring range, more information acquisition, etc.

[0004] Therefore, it has become a technical problem to be solved at present to provide a power utilization safety diagnosis method with high accuracy and fast response time. SUMMARY

[0005] In view of the above problems, the present application is proposed, which provides a power utilization safety diagnosis method based on a convolutional neural network. The calibrated temperature value at the current diagnosis time is used to select a high receptive field feature extraction channel or a low receptive field feature extraction channel to extract features from the target region image at the current diagnosis time, prevent the loss or blur of target features, improve the response speed, reduce the response time, and realize accurate and rapid diagnosis of the target region power utilization safety.

[0006] The present application provides a power utilization safety diagnosis method based on a convolutional neural network, which includes steps S1 to S6:

[0007] S1: obtaining the temperature value at the current diagnosis time of the target region and the target region image, and the temperature values at N-1 diagnosis times before the current diagnosis time, wherein N is greater than or equal to 2;

[0008] S2: constructing a denoising correction module, receiving the temperature value at the current diagnosis time and the temperature values at N-1 diagnosis times before the current diagnosis time, and performing denoising correction processing on the temperature value at the current diagnosis time to obtain a calibrated temperature value at the current diagnosis time;

[0009] S3: constructing a comparison module, receiving the calibrated temperature value at the current diagnosis time, comparing the calibrated temperature value at the current diagnosis time with a preset temperature threshold, and generating a selection signal;

[0010] S4: Construct a preprocessing module, receive a target region image of a current diagnosis moment of a target region, perform contrast enhancement on the target region image, and generate a target region enhanced image;

[0011] S5: Construct a feature extraction module, receive a selection signal and the target region enhanced image, select a high-receptive-field feature extraction channel or a low-receptive-field feature extraction channel according to the selection signal, perform feature extraction on the target region enhanced image, and generate a target feature;

[0012] S6: Construct a classifier, receive the target feature, perform classification on the target feature, and generate a diagnosis result.

[0013] Wherein, the denoising correction processing on the temperature value of the current diagnosis moment to obtain the corrected temperature value of the current diagnosis moment includes steps S21 to S23:

[0014] S21: Based on the temperature value of the current diagnosis moment and the temperature values of N-1 diagnosis moments before the current diagnosis moment, the maximum temperature value and the minimum temperature value are compared;

[0015] S22: According to the maximum temperature value, the minimum temperature value and the temperature values of N diagnosis moments, the optimal correction coefficient is calculated;

[0016] S23: According to the optimal correction coefficient α opt , the corrected temperature value of the diagnosis moment before the current diagnosis moment and the temperature value of the current diagnosis moment, the corrected temperature value of the current diagnosis moment is calculated.

[0017] Wherein, the optimal correction coefficient is specifically:

[0018]

[0019] Wherein, α opt is the optimal correction coefficient, X max is the maximum temperature value, X min is the minimum temperature value, and X t is the temperature value at diagnosis moment t.

[0020] Wherein, the corrected temperature value of the current diagnosis moment is specifically:

[0021] S N = S N-1 + α opt (X N -S N-1 )

[0022] Wherein, S N is the corrected temperature value of the current diagnosis moment, S N-1 is the corrected temperature value of the diagnosis moment before the current diagnosis moment, and X NThe temperature value at the current diagnosis moment.

[0023] The target region image is subjected to contrast enhancement to generate a target region enhanced image, including steps S41-S42:

[0024] S41: adjusting the gray value of a pixel of the target region image;

[0025] S42: normalizing the adjusted gray value of the pixel to obtain the target region enhanced image.

[0026] The gray value of the pixel of the target region image is adjusted, specifically as follows:

[0027]

[0028] wherein P is the gray value of the pixel in the target region image, P min is a low gray threshold, P max is a high gray threshold, and P1 is the adjusted gray value of the pixel.

[0029] The normalization process is specifically as follows:

[0030]

[0031] wherein P2 is the normalized gray value of the pixel.

[0032] The target region enhanced image is subjected to feature extraction according to the selected high-receptive-field feature extraction channel or low-receptive-field feature extraction channel to generate a target feature, specifically including steps S51-S52:

[0033] S51: when the selection signal indicates that the corrected temperature value at the current diagnosis moment is less than a preset temperature threshold, the low-receptive-field feature extraction channel is selected, wherein the low-receptive-field feature extraction channel is connected by four convolution layers in sequence to extract features from the target region enhanced image to generate the target feature;

[0034] S52: when the selection signal indicates that the corrected temperature value at the current diagnosis moment is not less than the preset temperature threshold, the high-receptive-field feature extraction channel is selected, wherein the high-receptive-field feature extraction channel is connected by eight convolution layers in sequence to extract features from the target region enhanced image to generate the target feature.

[0035] The classifier adopts a Softmax function to generate a diagnosis result.

[0036] The application has the following beneficial effects:

[0037] (1) The application provides a power consumption safety diagnosis method based on a convolutional neural network, which selects a high receptive field feature extraction channel or a low receptive field feature extraction channel by using the calibration temperature value at the current diagnosis time, extracts features of the target region image at the current diagnosis time, prevents loss or blur of target features, improves response speed, reduces response time, and realizes accurate and rapid diagnosis of target region power consumption safety.

[0038] (2) The application also constructs a denoising correction module and designs a denoising calibration method, removes burrs and noise of the temperature value according to the temperature value at the current diagnosis time of the target region and the correction temperature value at the previous diagnosis time, and obtains the correction temperature value at the current diagnosis time through comprehensive calculation, thereby improving the accuracy of temperature value acquisition and improving the diagnosis result of subsequent power consumption safety.

[0039] (3) The application also constructs a preprocessing module, receives the target region image at the current diagnosis time of the target region, and performs contrast enhancement preprocessing on the target region image to enhance target features and improve the accuracy of subsequent feature extraction. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without creative labor under the premise of the drawings.

[0041] Figure 1 The flowchart of the power consumption safety diagnosis method based on the convolutional neural network provided by the application;

[0042] Figure 2 The flowchart of the application for obtaining the correction temperature value;

[0043] Figure 3 The flowchart of the application for contrast enhancement;

[0044] Figure 4 The overall block diagram of the power consumption safety diagnosis method based on the convolutional neural network provided by the application. DETAILED DESCRIPTION

[0045] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0046] The terms used in the embodiments of the present application are merely for the purpose of describing specific embodiments, and are not intended to limit the present application. The singular forms "a", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. "Plural" generally includes at least two, but does not exclude the case of including at least one.

[0047] The present application provides a power consumption safety diagnosis method based on a convolutional neural network, which selects a high receptive field feature extraction channel or a low receptive field feature extraction channel using a calibration temperature value at a current diagnosis moment, extracts features of a target region image at the current diagnosis moment, prevents loss or blur of target features, improves response speed, reduces response time, and realizes accurate and rapid diagnosis of target region power consumption safety.

[0048] Meanwhile, the present application also constructs a denoising correction module and designs a denoising calibration method, removes burrs and noise of the temperature value according to the temperature value at the current diagnosis moment of the target region and the correction temperature value at the previous diagnosis moment, and comprehensively calculates the correction temperature value at the current diagnosis moment, thereby improving the accuracy of temperature value acquisition and improving the diagnosis result of subsequent power consumption safety.

[0049] Finally, the present application also constructs a preprocessing module, receives a target region image at a current diagnosis moment of a target region, performs contrast enhancement preprocessing on the target region image, enhances target features, and improves the accuracy of subsequent feature extraction.

[0050] The present application will be further described below with reference to the drawings and specific embodiments.

[0051] As shown in Figure 1 In one embodiment, the present application provides a power consumption safety diagnosis method based on a convolutional neural network. The power consumption safety diagnosis method specifically includes the following steps S1 to S6:

[0052] S1: obtaining a temperature value at a current diagnosis moment of a target region and a target region image, and temperature values at N-1 diagnosis moments before the current diagnosis moment, wherein N is greater than or equal to 2;

[0053] Specifically, a temperature sensor is selected to monitor the temperature of the target region, and temperature values at the current diagnosis moment and N-1 diagnosis moments before the current diagnosis moment are obtained.

[0054] A visual sensor is selected to monitor the image of the target region, and a target region image at the current diagnosis moment is obtained.

[0055] S2: A denoising correction module is constructed to receive the temperature value at the current diagnosis moment and the temperature values at the N-1 diagnosis moments before the current diagnosis moment, and to perform denoising correction processing on the temperature value at the current diagnosis moment to obtain a corrected temperature value at the current diagnosis moment.

[0056] In the temperature acquisition process of the target region, due to the interference of various environmental factors such as electromagnetic noise and temperature sensor instability, the directly acquired temperature value often contains many burrs and noises. These burrs and noises not only lead to inaccurate temperature values, but also lead to inappropriate selection of subsequent feature extraction channels, affecting the diagnosis results of power safety. By constructing a denoising correction processing module, the current diagnosis moment temperature value of the target region is received, the burrs and noises of the temperature value are removed, and the corrected temperature value at the current diagnosis moment is obtained.

[0057] Specifically, as shown in Figure 2 , the temperature value at the current diagnosis moment is denoised and corrected to obtain the corrected temperature value at the current diagnosis moment, including the following steps:

[0058] S21: Based on the temperature value at the current diagnosis moment and the temperature values at the N-1 diagnosis moments before the current diagnosis moment, the maximum temperature value and the minimum temperature value are compared.

[0059] S22: According to the maximum temperature value, the minimum temperature value, and the temperature values at the N diagnosis moments, the optimal correction coefficient a is calculated. opt , the optimal correction coefficient a opt Specifically:

[0060]

[0061] Where X max is the maximum temperature value, X min is the minimum temperature value, and X t is the temperature value at diagnosis moment t. By this formula, the optimal correction coefficient a opt is calculated, which can not only ensure that the temperature average value is within the temperature change interval X max -X min , but also ensure that 0≤a opt ≤1.

[0062] S23: According to the optimal correction coefficient a opt, the correction temperature value of the previous diagnosis moment of the current diagnosis moment and the temperature value of the current diagnosis moment, the correction temperature value of the current diagnosis moment is calculated, and the correction temperature value of the current diagnosis moment is specifically:

[0063] S N = S N-1 + α opt (X N - S N-1 )

[0064] Wherein, S N is the correction temperature value of the current diagnosis moment, S N-1 is the correction temperature value of the previous diagnosis moment of the current diagnosis moment, and X N is the temperature value of the current diagnosis moment.

[0065] Specifically, the present application constructs the correction temperature value S N of the current diagnosis moment by the temperature value X N-1 of the current diagnosis moment and the correction temperature value S N of the previous diagnosis moment, sets the correction coefficient α to adjust the weight of the temperature value X N of the current diagnosis moment and the correction temperature value S N-1 of the previous diagnosis moment, that is

[0066] S N = αX N + (1-α)S N-1

[0067]

[0068] Wherein, the correction coefficient α of each diagnosis moment is different, 0≤α≤1, when the correction coefficient α is closer to 1, the correction temperature value S N is equal to the temperature value X N of the current diagnosis moment, and the correction temperature value S N is mainly determined by the temperature value X N of the current diagnosis moment. When the correction coefficient α is closer to 0, the correction temperature value S N is equal to the correction temperature value S N-1 of the previous diagnosis moment, and the correction temperature value S N is mainly determined by the correction temperature value S N-1 of the previous diagnosis moment.

[0069] Further, in step S22 of the present application, the temperature change interval is defined by the temperature difference between the maximum temperature value and the minimum temperature value, and the deviation degree of the temperature average value from the maximum temperature value of the temperature change interval is determined according to , and then the real-time optimal correction coefficient α optUsed to correct the temperature value X at the current diagnostic time. N When the average temperature approaches the maximum temperature value within the temperature variation range, the indication deviation is significant, and the temperature value exhibits considerable spikes and noise. Therefore, the optimal correction factor α is... opt When the temperature is close to 0, the corrected temperature value at the current diagnostic time is primarily determined by the corrected temperature value at the previous diagnostic time. Conversely, when the average temperature is far from the maximum temperature value within the temperature variation range, the deviation is smaller, and the noise and glitches in the temperature value are weaker, thus the optimal correction coefficient α is [value missing]. opt The corrected temperature value at the current diagnostic time is close to 1, and is mainly based on the temperature value X at the current diagnostic time. N Decide.

[0070] This application constructs a noise reduction and correction module, based on the temperature value X of the target area at the current diagnostic time. N By combining the corrected temperature value from the previous diagnostic moment with the corrected temperature value, removing noise and glitches in the temperature value, and comprehensively calculating the corrected temperature value for the current diagnostic moment, the accuracy of temperature value acquisition is improved, thus enhancing the diagnostic results for subsequent electrical safety.

[0071] In a preferred embodiment, when the current diagnostic time is the initial diagnostic time, the correction temperature value is set to the temperature value of the target area at the current diagnostic time.

[0072] S3: Construct a comparison module, receive the corrected temperature value at the current diagnostic time, compare the corrected temperature value at the current diagnostic time with the preset temperature threshold, and generate a selection signal.

[0073] If the corrected temperature value at the current diagnostic time is less than the preset temperature threshold, it indicates that the probability of a fire occurring in the target area at the current diagnostic time is low; conversely, if the corrected temperature value at the current diagnostic time is not less than the preset temperature threshold, it indicates that the probability of a fire occurring in the target area at the current diagnostic time is high.

[0074] S4: Construct a preprocessing module, receive the target region image at the current diagnostic time, enhance the contrast of the target region image, and generate an enhanced target region image.

[0075] In the early stages of a fire, the fire targets in the images of the target area obtained by the visual sensor are relatively small and their features are relatively blurry. Therefore, this application performs contrast enhancement preprocessing on the target area images to enhance the target features and improve the accuracy of subsequent feature extraction.

[0076] Specifically, such as Figure 3 As shown, contrast enhancement is performed on the target region image to produce an enhanced target region image, specifically including the following steps:

[0077] S41: adjust the gray value of the pixel of the target region image, specifically:

[0078]

[0079] Wherein, P is the gray value of the pixel in the target region image, P min is a low gray threshold, P max is a high gray threshold, P1 is the gray value of the adjusted pixel.

[0080] S42: normalize the gray value of the adjusted pixel to obtain a target region enhanced image, and the normalization process is specifically:

[0081]

[0082] Wherein, P2 is the gray value of the normalized pixel.

[0083] S5: Construct a feature extraction module, receive a selection signal and a target region enhanced image, select a high receptive field feature extraction channel or a low receptive field feature extraction channel according to the selection signal, and perform feature extraction on the target region enhanced image to generate a target feature.

[0084] The temperature value of the target region at the current diagnosis moment is directly related to whether a fire occurs. When a fire is about to occur or after a fire occurs, the temperature value of the target region will be relatively large. At this time, since the fire just occurred, the fire target is relatively small, and the fire target feature is relatively fuzzy, the feature extraction capability of the feature extraction module at the current diagnosis moment needs to be enhanced, therefore, the high receptive field feature extraction channel is selected through the high temperature value at the current diagnosis moment, and the feature extraction capability is enhanced by using relatively more convolution layers, to prevent missed diagnosis and improve the accuracy of the target region power safety diagnosis. Conversely, when there is no fire or under normal circumstances, the temperature value of the target region will be relatively small, and the possibility of fire occurrence is also relatively small, therefore, the low receptive field feature extraction channel is selected through the low temperature value at the current diagnosis moment, and the response speed of the diagnosis is improved by using relatively fewer convolution layers, to reduce the response time and realize the rapid diagnosis of the target region power safety.

[0085] Specifically, the overall block diagram of the power safety diagnosis method based on the convolutional neural network is as shown in Figure 4 , wherein the high receptive field feature extraction channel or the low receptive field feature extraction channel is selected according to the selection signal, the target region enhanced image is subjected to feature extraction to generate a target feature, and the specific steps include the following steps:

[0086] S51: when the selection signal indicates that the corrected temperature value at the current diagnosis moment is less than a preset temperature threshold, the low receptive field feature extraction channel is selected, wherein the low receptive field feature extraction channel is connected by 4 convolution layers in turn, the target region enhanced image is subjected to feature extraction, and a target feature is generated.

[0087] S52: When the selection signal indicates that the correction temperature value at the current diagnosis moment is not less than the preset temperature threshold, a high receptive field feature extraction channel is selected, wherein the high receptive field feature extraction channel is sequentially connected by 8 convolutional layers, and the target region enhanced image is subjected to feature extraction to generate target features.

[0088] S6: A classifier is constructed, target features are received, the target features are classified, and a diagnosis result is generated.

[0089] Specifically, the classifier adopts a Softmax function to generate the diagnosis result.

[0090] The above description shows and describes the preferred embodiments of the present application, but as previously described, it should be understood that the present application is not limited to the forms disclosed herein, should not be considered as excluding other embodiments, and can be used in various other combinations, modifications and environments, and can be modified within the scope of the inventive concept described herein by the above teaching or related technical or knowledge. Any modification and change made by those skilled in the art without departing from the spirit and scope of the present application shall be within the protection scope of the claims appended to the present application.

Claims

1. A power consumption safety diagnosis method based on a convolutional neural network, characterized in that, The method comprises steps S1 to S6: S1: obtaining temperature values at a current diagnosis time and target region images of a target region, and temperature values at N-1 diagnosis times before the current diagnosis time, wherein N is greater than or equal to 2; S2: constructing a denoising correction module, receiving the temperature values at the current diagnosis time and the temperature values at the N-1 diagnosis times before the current diagnosis time, and performing denoising correction processing on the temperature values at the current diagnosis time to obtain corrected temperature values at the current diagnosis time; S3: constructing a comparison module, receiving the corrected temperature values at the current diagnosis time, comparing the corrected temperature values at the current diagnosis time with a preset temperature threshold, and generating a selection signal; S4: constructing a preprocessing module, receiving the target region images of the target region at the current diagnosis time, performing contrast enhancement on the target region images to generate target region enhanced images; S5: constructing a feature extraction module, receiving the selection signal and the target region enhanced images, selecting a high-receptive-field feature extraction channel or a low-receptive-field feature extraction channel according to the selection signal, and performing feature extraction on the target region enhanced images to generate target features, specifically comprising steps S51 to S52: S51: when the selection signal indicates that the corrected temperature values at the current diagnosis time are less than the preset temperature threshold, selecting the low-receptive-field feature extraction channel, wherein the low-receptive-field feature extraction channel is connected by four convolution layers in sequence, and the target region enhanced images are subjected to feature extraction to generate the target features; S52: when the selection signal indicates that the corrected temperature values at the current diagnosis time are not less than the preset temperature threshold, selecting the high-receptive-field feature extraction channel, wherein the high-receptive-field feature extraction channel is connected by eight convolution layers in sequence, and the target region enhanced images are subjected to feature extraction to generate the target features; S6: constructing a classifier, receiving the target features, classifying the target features, and generating a diagnosis result.

2. The power safety diagnostic method of claim 1, wherein, The denoising correction processing on the temperature values at the current diagnosis time to obtain the corrected temperature values at the current diagnosis time comprises steps S21 to S23: S21: comparing the temperature values at the current diagnosis time and the temperature values at the N-1 diagnosis times before the current diagnosis time to obtain a maximum temperature value and a minimum temperature value; S22: calculating an optimal correction coefficient according to the maximum temperature value, the minimum temperature value, and the temperature values at the N diagnosis times; S23: Calculate the correction temperature value at the current diagnosis time according to the optimal correction coefficient a opt , the correction temperature value at the previous diagnosis time of the current diagnosis time, and the temperature value at the current diagnosis time.

3. The power safety diagnostic method of claim 2, wherein, The optimal correction coefficient is specifically: wherein α opt is the optimal correction factor, is the maximum temperature value, is the minimum temperature value, is the temperature value at the diagnosis time t.

4. The power safety diagnostic method of claim 3, wherein, The corrected temperature values at the current diagnosis time are specifically: wherein, is a corrected temperature value at a current diagnosis time point, is a corrected temperature value at a previous diagnosis time point of the current diagnosis time point, is a temperature value at the current diagnosis time point.

5. The power safety diagnostic method of claim 1, wherein, The contrast enhancement on the target region images to generate the target region enhanced images comprises steps S41 to S42: S41: adjusting the gray values of pixels of the target region images; S42: normalizing the adjusted gray values of the pixels to obtain the target region enhanced images.

6. The power safety diagnostic method of claim 5, wherein, The adjustment of the gray values of the pixels of the target region images is specifically: wherein P is a gray value of a pixel in the target region image, is a low gray threshold value, is a high gray threshold value, is a gray value of the adjusted pixel.

7. The power safety diagnostic method of claim 6, wherein, The normalization processing is specifically: wherein, is the normalized pixel gray value.

8. The power safety diagnostic method of claim 1, wherein, The classifier adopts a Softmax function to generate the diagnosis result.

Images