Research and Application of Anti-electric Shock Communication Terminal Based on Internet of Things Technology

Leakage protection cannot effectively identify the voltage and current reception signals of electric shock grounding points that may pose a danger to personal safety through analysis of electric shock parameters, which is inherently inadequate. When a communication electric shock safety accident occurs, the accident site can be visually displayed in the safety prevention and control center, and an alarm is sent out. Meanwhile, relevant control instructions are issued to cut off the accident communication phase line to prevent secondary injury. In this paper, artificial intelligence and Internet of Things technology are applied to the communication system. The tag sensor is used in the monitoring system and is also used as the perception layer. The electric shock receiving signal obtained under the control circuit structure of the Internet of Things is determined by means of S transformation. The physical characteristics of wave frequency sensitivity are analyzed. Then, the external characteristics of biological shock are extracted with the help of the working frequency of each band of wavelet multi-resolution analysis, so that the electric shock model can deal with the hazards that cannot be identified by the traditional perception layer. The deep neural network is optimized to improve the recognition probability. Experiments verify that the recognition rate of the biological electric shock is as high as 99%, which further improves the stability of the leakage protection device and provides a guarantee for the normal operation of the communication system.


Introduction
With the construction and development of the country, the infrastructure construction of the power system is becoming more and more perfect.The coverage of power communication is also expanding, which plays a more important role in national economic construction and livelihood security.At present, most low-voltage power grid systems adopt voltage and current leakage protection devices to prevent electric shock accidents in human bodies and abnormal work of household appliances' cable insulation.However, the current operating procedures show that the original leakage protection device of the leakage circuit breaker is often mistakenly triggered.The protection's physical characteristics are hindered by normal leakage voltage and current and the leakage current of an alternating current signal angle is abnormal.There is a protection dead zone.When there is abnormal weather and large load capacity, there will be abnormal phenomena such as disconnector failure [1].Relevant data show that there are serious defects in the posture working principle of the leakage protector, and it is imperative to develop a high-efficiency leakage protection device for the leakage circuit breaker based on the impulse voltage and current situation [2].
[3] distinguishes the load impedance signal using the positive and negative characteristics of leakage protection and calculates the loss parameters of impedance by combining the two-point trigonometric function.In [4], the eigenvector is adopted to simulate the characteristics of the human body's electric shock signal.The neural network is used to analyze it, and the protection model is designed to effectively identify the fault type of the electric shock signal.[5] proposes a detection method of electric shock characteristics based on the analysis of currency fluctuation by the wavelet algorithm.[6] adopts the minimum forest algorithm to process the signal and detect the electric shock current, which can extract the electric shock current model of the contact object and determine whether the threshold is triggered.[7] proposes an anomaly diagnosis method based on the eigenmode vector integrated with the Internet of Things technology.It studied the practical application in the low-voltage communication network, monitored the leakage trigger point, and solved various problems from operation management to subsequent maintenance.[8] designs a single-phase short circuit trigger identification of the low-voltage distribution communication network.It solves the problem of line information acquisition and judgment and promotes the development of real-time monitoring technology in the communication platform signal receiving alarm technology.
It is clear that the leakage of the power grid communication terminal will cause the abnormal phenomenon of network disconnection, thus hindering the safe and stable operation of the power system.Moreover, the abnormal phenomenon of electric shock caused by electric leakage also endangers the safety of operators.To ensure the long-term operation of the power system, the closing function of the leakage protector must be turned off, which may eventually affect the forced closing of the communication terminal protection switch.In this paper, the rules of the leakage protection switch are divided according to the specific classification of the electrocution organisms.When the communication terminal starts the program due to the superior protection, the closing rules of the leakage circuit breaker are judged by extracting the spectrum characteristics of the biological electric shock, which is to protect the normal operation of the electric power communication system and ensure the safety of operators.

The shock current signal
To extract the leakage current in the total residual current line of the low-voltage power grid, it is necessary to build a biological electric shock physical test platform.It also needs to obtain a large number of relevant electric shock test signals and characterize the characteristics of electric shock current signals by collecting the biological electric shock current signal data.
Referring to the electronics foundation of [9] and the structural characteristics of the low-voltage power communication network, a physical experimental platform for biological electric shock is built, as shown in Figure 1.The electric shock test bench is mainly composed of a three-phase power supply voltage, fault recorder, current protection device, and current transformer.The magnitude of the electric shock current is closely related to the path of the current flowing through the human body.Therefore, when testing, the left arm of the test organism is selected for electric shock, and the right leg is grounded.
In Figure 1, we adjust the voltage between the two ends and the value of the automatic balancing filter capacitor box to a reasonable range.Before the electric shock test, all the participants are generally divided into four groups wave recording, four groups of database data recording, four groups of biological electric shock, and four groups for line switching.To achieve this effect, it is necessary to prevent biological injury or disability.If the voltage and current detected by the logger software are greater than 45 mA, the output signal is clipped.If we want to achieve this effect, we must control the electric shock voltage and current.When the voltage and current are relatively reasonable, we should use the parallel circuit voltage and current to limit the current of the electric shock grounding point.When the external environment changes or the grid voltage of the grid system is large, the three-phase voltage and current can be adjusted accordingly and remedial measures can be taken.

Spectrum analysis of the electric shock signal
When the working frequency of the composite function's main lobe is not fixed, the actual length of the windowing is also virtually determined, so that the frequency domain of the received signal cannot be taken into account.The Fourier transform is a low decomposition.S-transform inherits the advantages and disadvantages of the neural network and uses variable Gaussian window composite function to overcome the disadvantages effectively.The S-transform formula is as follows: where , is a composite function whose relevant parameters are the height and width of the window.It is obtained from the high-wide composite function of the t i  window.Its low resolution is determined by the operating frequency component to be analyzed.
When a relatively narrow Gaussian window is chosen to open the signal, the overall inference of the received signal can be made.When analyzing the high-frequency components, it may be impossible to identify the micro-motion parameters of the received signal by using a relatively narrow Gaussian window.
Similar to neural networks, the S-transform simplifies the calculation steps and greatly reduces the complexity, which can be rewritten into a specific form of Fourier transform.The calculation time can be further reduced by means of a fast Fourier transform.The S-transform formula is like this: The corresponding convolutional network composite function is shown in Formula (4).
The simulation experiment analysis

The simulation environment settings
This article establishes a simulation model for the anti-electric shock communication terminal system based on Matlab/Simulink.The power module settings are shown in Table 1: The database data of the neural network model uniformly select the database of [10] as the external feature to facilitate the comparison of the database data.In addition, 200 groups of data in the database were chosen for intensive training, and 30 groups of test data were used.
Two node tests of the deep neural network's direct input layer are set.The loss function of the threelayer structure is Relu, and the loss function of the output layer is Softmax.Finally, 0  stands for biological shock and 1  stands for non-biological shock.
We take the highest peak-valley method as an example.The relevant parameters are displayed in Table 2. 20 / S -After the electric shock identification is completed, the highest peak-valley value is still selected for the database after the electric shock.The SVD singular values from the encoder are decomposed, and the least square method is applied to reduce the dimension of the AE and other database data.The SVD singular value extrinsic feature distribution is shown in Figure 2.

Analysis of test results (1) Accuracy test
The high accuracy of the least-squares test varies with the setting of the height-width parameter, as shown in Figure 3. (2) Data processing comparison test In this paper, SVD (Singular Value Decomposition) and KNN (K-Nearest Neighbors) algorithms are taken to compare with the method.With the help of data processing and ways for classifying the least squares of electric shock, intensive training and testing are done.The final results of all-electric shock data processing and the classification method for the electric shock least square method are presented in Table 3.

correct
In Table 3, the accuracy of the SVD algorithm is low when the data is processed as the highest peak and valley value, and some labels are identified incorrectly.The average classification accuracy of the KNN algorithm is 83.22%, and the accuracy is relatively high.The data processing efficiency of the algorithm in this paper reaches 99.99% after several iterations, which proves that it is the most suitable for the leakage judgment and prediction of communication terminals.

Conclusion
In this paper, a sensor monitoring system of the Internet of Things based on artificial intelligence is proposed.The system generates historical reports from the collected data, achieves real-time analysis of the operation status, and optimizes and adjusts the working status of communication terminals to prevent the risk of electric shock.After testing the prediction accuracy of the power grid communication terminal leakage, it is found that the electric shock judgment accuracy is as high as 99.99%.Besides, the nonliving animal electric shock accuracy is as high as 96.13%, and the living animal electric shock classification accuracy is 100%.
In this paper, the dangerous area of the electric power communication system is simulated and tested, and the test object is single.The test of the simulated dangerous area cannot reflect the accurate effect of the system on the calculation of the dangerous area.With the continuous change in the working face environment, even in the same area, there are different environmental indicators.In the face of uncertain environmental changes, in future work, we need to base on the actual environment of the power communication system.Besides, we should be able to carry out dangerous area tests based on ensuring safety and stability and compare the test results of different environments.

Figure 1
Figure 1 Bioelectric shock test

Figure 2
Figure 2 The distribution of the SVD singular value external feature

Figure 3
Figure 3 Accuracy test of the window parameter change In the neural network, we select the penalty compound function C = 1.The kernel compound function chooses the quadratic equation kernel compound function poly, the quadratic equation kernel compound function, the two-dimensional degree = 6, and the parameter gamma = 20.1.How to classify a 1 classifier is considered.For the 200 groups of the electric shock database data, we set the intensive training database data as 140 groups and the test data as 25 groups.(2)Data processing comparison test In this paper, SVD (Singular Value Decomposition) and KNN (K-Nearest Neighbors) algorithms are taken to compare with the method.With the help of data processing and ways for classifying the least squares of electric shock, intensive training and testing are done.The final results of all-electric shock data processing and the classification method for the electric shock least square method are presented in Table3.Table3Comparison results of data processing Comparison method Labels Predicted value Recognition result

Table 2
Settings of the simulation environment parameter