Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Reading Tip: This article is about 4600 words, it is recommended to collect it for later reading!

Short circuits, leakage currents, and illegal charging of electric bicycles can easily lead to tripping, fires, and other safety accidents. Therefore, Yu Wuqiang, Ma Xiao, Wang Zhimin, Yang Hongxiang, and Su Zhicheng from the Lijiang Power Supply Bureau of Yunnan Electric Power Company designed a non-intrusive intelligent safety electricity monitoring terminal in the 2024 issue 5 of “Electrical Technology”. The design employs precision resistive voltage division and high-precision zero magnetic flux current sensors to achieve the measurement of voltage and current analog signals, retaining high-order harmonic characteristics through 10kHz high-frequency sampling. By utilizing dual-core Cortex-A9 ARM and Xilinx Artix7 FPGA as the information processing unit, it can simultaneously handle six channels of high-frequency sampling signals. This study investigates the abnormal electricity characteristics of short circuits, leakage currents, and illegal charging of electric bicycles, developing non-intrusive abnormal electricity identification software based on convolutional neural network models. Tests show that the designed intelligent safety electricity monitoring terminal can accurately identify the aforementioned typical abnormal electricity events.

In recent years, the demand for tourism consumption in China has been growing rapidly, leading to a quick increase in electricity load in scenic areas during peak seasons. Issues such as unauthorized wiring, overloaded electrical equipment, and illegal charging of electric bicycles frequently occur, posing safety hazards such as short circuits, leakage currents, and fires in the operation of the distribution network in scenic areas. Traditional electricity inspection methods are unable to timely detect these electricity hazards, placing great pressure on the daily operation and maintenance of the network equipment in scenic areas, while severely impacting users’ electricity experience.

Non-intrusive load monitoring technology can obtain detailed user electricity information without entering homes, allowing real-time monitoring and assessment of the current operating status of various electrical devices, identifying abnormal electricity behaviors and potential electricity hazards. With the development of intelligent distribution network technology, the application of distribution monitoring terminals has become increasingly widespread, significantly enhancing the intelligent perception level of the distribution area. Integrating non-intrusive load monitoring technology with distribution monitoring terminals is of great significance for further improving the safety of electricity use in scenic areas.

Currently, research on electricity safety monitoring mainly focuses on terminal devices, including some abnormal electricity detection algorithms. Some literature has designed intelligent monitoring terminals based on big data for electricity safety, employing a dual CPU control mode to analyze and determine the safety of electricity based on electricity safety standards.

Some literature has researched new methods for abnormal electricity monitoring based on local power benchmarks, where monitoring terminals obtain actual values of voltage, current, and power by monitoring the grid, comparing actual values with benchmark values to assess the operation of the grid, allowing timely detection of abnormal electricity behaviors and statistical measurement of losses caused by measurement errors.

Some literature has developed a distributed low-voltage power distribution remote monitoring system based on the Internet of Things, which integrates electricity parameter collection and metering technology, 4G communication technology, and Internet of Things cloud platform display technology to achieve remote collection and transmission of electrical parameter data. Some literature has researched online monitoring terminals for DC charging piles of electric vehicles, using a modular design approach to transmit collected charging data and environmental data to the online monitoring platform via 4G networks, meeting the basic requirements for online monitoring of charging facility status.

Some literature has developed distribution terminals with dynamic changes in operating power consumption, where power consumption can change in real-time according to terminal operating conditions. Some literature has researched a new type of electric energy monitoring terminal capable of monitoring and statistically analyzing faults such as power loss, phase loss, overvoltage, undervoltage, leakage, and three-phase imbalance, providing precise early warnings for frequent abnormal issues such as power loss or leakage in the distribution area. In summary, safety electricity monitoring has become a widely concerned research field, but there are few studies applying non-intrusive load monitoring technology to safety electricity monitoring.

This article studies the electrical characteristics of common abnormal electricity scenarios, laying the foundation for detecting and identifying abnormal situations such as indoor short circuits, leakage currents, and illegal charging of electric bicycles. Considering factors such as collection accuracy requirements, information transmission, functional upgrades, and system practicality, a non-intrusive abnormal electricity identification algorithm is developed, and an intelligent safety electricity monitoring terminal is created. Tests show that the designed intelligent safety electricity monitoring terminal can accurately identify the aforementioned typical abnormal electricity events.

The developed terminal has already been pilot tested on-site, achieving effective management of electricity loads for residents in scenic areas, thereby reducing safety risks of electrical equipment and fire hazards in scenic areas, and improving the reliability of power supply and user satisfaction.

1 Typical Abnormal Electricity and Its Characteristics

1.1 Overcurrent Short Circuit

A short circuit is an extremely dangerous abnormal state, caused by various factors including the mismatch between the selection and installation of electrical equipment and actual usage requirements, or damage to the insulation layer due to aging. When a short circuit occurs indoors, the current increases sharply, with peak currents reaching hundreds to thousands of amperes, causing the circuit to heat up or produce arcs, potentially leading to electrical accidents such as fires.

When a short circuit occurs, there is a transition period during which the circuit changes from a stable state of normal operation to a stable state of deep short circuit. This transition process is generally referred to as the transient process of short circuit current. A typical short circuit current waveform is shown in Figure 1.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 1 Typical Short Circuit Current Waveform

In Figure 1, i_p is the instantaneous value of the AC component of the short circuit current, i_g is the instantaneous value of the DC component of the short circuit current, i_sh is the instantaneous value of the entire short circuit current process, i_pk is the peak value of the impulse short circuit current, u is the voltage of the distribution network, i is the normal operating current of the distribution network.

1.2 Leakage Current

Leakage current is an abnormal electricity phenomenon caused by wear of the line insulation layer, moisture in insulators, aging of equipment, etc., which is not easily detected in the early stages. Once leakage occurs in the circuit, it frequently triggers the current protector to trip, affecting users’ normal electricity use and even causing electric shock accidents, posing safety hazards to people’s production and life. A typical leakage current waveform is shown in Figure 2.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 2 Typical Leakage Current Waveform

In Figure 2, the leakage current waveform is sampled at a frequency of 1.6kHz, with 32 sampling points per cycle. It can be seen that when leakage occurs, high-order harmonics appear in the current, and the number and amplitude of harmonics are related to the specific leakage situation.

1.3 Electric Bicycle Charging

The charging process of electric bicycles can generally be divided into three stages: constant current, constant voltage, and trickle charging, with distinct load curve characteristics. In the constant current stage, the load curve gently rises; in the constant voltage stage, the load curve gradually declines from high power; in the trickle stage, the load curve slowly declines at low power levels. The power curve for charging electric bicycles is shown in Figure 3.

Charging generally lasts about 4 hours. The electric bicycle charger acts as a rectifier and generates a large amount of harmonics during operation. The instantaneous current waveform during electric bicycle charging is shown in Figure 4.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 3 Power Curve of Electric Bicycle Charging

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 4 Instantaneous Current Waveform During Electric Bicycle Charging

Analyzing the current waveform in Figure 4 reveals that the harmonic content of the electric bicycle charging current is relatively high. Further frequency domain analysis indicates that the harmonics primarily consist of odd harmonics such as the 3rd, 5th, and 7th, which are key features for identifying electric bicycle charging.

2 Abnormal Electricity Identification Based on Convolutional Neural Networks

This article employs convolutional neural networks (CNN) to identify abnormal electricity current curves. Compared to other fully connected neural networks, CNNs can retain spatial features of data during the processing of two-dimensional image data, achieving higher recognition accuracy. CNN consists of an input layer, hidden layers, and an output layer, where the hidden layers comprise convolutional layers, pooling layers, dropout layers, and fully connected layers, with alternating repetitions of convolutional and pooling layers. The structure of CNN is shown in Figure 5.

In Figure 5, the input is a two-dimensional image of the load to be identified. The convolutional layer extracts features from the input data using multiple convolutional kernels. The convolutional kernel, also known as a filter, has internal weights adjusted through multiple backpropagation corrections.The pooling layer compresses the input image, thereby reducing the input dimensions for the next layer and improving computational efficiency.The fully connected layer is responsible for load classification, with the category yielding the highest probability taken as the identification result.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 5 Structure of Convolutional Neural Network

Generally, the more layers a CNN has, the more representative the features extracted, but it also requires more computational storage space and results in slower computation speeds. Considering recognition accuracy and the hardware requirements of the algorithm, the parameters of the convolutional neural network structure used in this article are shown in Table 1.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Table 1 Structure Parameters of Convolutional Neural Network

3 Terminal Hardware Design

3.1 Voltage Signal Sampling Design

The voltage sampling uses the principle of resistive voltage division, which has the characteristics of good linearity and low temperature drift (using the bridge principle), requiring a high input impedance sampling circuit. The input voltage is divided through high-precision, high-stability resistors and then followed by a differential amplifier, achieving high input impedance through a specialized high-precision AD front-end differential driver. The voltage signal sampling scheme is shown in Figure 6.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 6 Voltage Signal Sampling Scheme

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Equation (1)

3.2 Current Signal Sampling Design

Current measurement uses a zero magnetic flux transformer, also employing a high input impedance design. The input current is followed by a differential amplifier after passing through the zero magnetic flux transformer, and then through a specialized high-precision AD front-end differential driver, achieving high impedance input. The current signal sampling scheme is shown in Figure 7.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 7 Current Signal Sampling Scheme

To improve the dynamic measurement performance of the current signal, the zero magnetic flux transformer uses four ranges: 100A, 50A, 10A, and 2A, with an overall range error of less than 0.2%.

3.3 Main Control Unit Design

The design employs a dual-core Cortex-A9 advanced RISC machine (ARM) and Xilinx Artix7 FPGA. The Cortex-A9 processor is compatible with other Cortex series processors and ARM MPCore technology, allowing for efficient reuse of a rich ecosystem including operating systems (OS), real-time operating systems (RTOS), middleware, and applications, thereby reducing costs associated with adopting a completely new processor. Compared to similar FPGAs, Artix7 has lower power consumption and cost, is 30% faster than Spartan-7, and 50% lower in power consumption, with a price reduction of 35%. The multi-core framework of the intelligent safety electricity monitoring terminal is shown in Figure 8.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 8 Multi-Core Framework of Intelligent Safety Electricity Monitoring Terminal

In Figure 8, PL (programmable logic) core is the FPGA, and PS (processing system) core is the dual-core ARM. The MicroBlaze soft core is used within PL, while the ARM dual hard core is within PS, with general peripheral controllers and DDR (double data rate) memory mounted on PS, sharing DDR between PL and PS. The PS program runs the application in DDR, while MicroBlaze runs in the internal BRAM (block RAM).

4 Terminal Software Design

The program for the intelligent safety electricity monitoring terminal is developed in embedded C language, with the main program structure divided into board-level driver layer, chip configuration layer, system service layer, and business logic layer. The business logic layer mainly implements the logical coordination and service combination of upper-level functional applications.

During the algorithm upgrade process, system service calls are implemented according to the algorithm architecture’s requirements, adding or removing data operations and processing, maintaining independence from the underlying code while keeping the chip configuration layer and board-level driver layer stable and secure. In the implementation of the abnormal electricity detection algorithm, interface functions of the business logic layer and system service layer are called to achieve functions such as data computation, record storage, data communication, and remote upgrades. The code layering is reasonable and clear, allowing for quick and convenient development of new business functions, flexibly cooperating with object-oriented protocols for data processing and uploading.

The main program flow is shown in Figure 9. After starting the terminal, necessary registers are maintained first, followed by the initiation of system services. The main control unit analyzes the voltage and current sampling data, detecting and identifying abnormal electricity events based on abnormal electricity characteristics, and issues alarm signals upon detection of such events.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 9 Main Program Flow

5 Simulation and Testing

5.1 Simulation Testing

To verify the accuracy of the CNN-based abnormal electricity identification model, simulation tests are conducted using the electric bicycle charging identification model as an example. First, the collected data samples are preprocessed, identifying that electric bicycle charging mainly occurs in the steady-state working segment, capturing the voltage and current waveforms during the steady-state segment, and plotting the V–I trajectory as shown in Figure 10, generating a 32×32 pixel grayscale image.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 10 V-I Trajectory of Electric Bicycle Charging

Then, the dataset is proportionally divided into training and testing sets, and the CNN model is trained on a computer, with training results shown in Figure 11.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Figure 11 Training Results of Electric Bicycle Charging Identification Model

As seen in Figure 11, the model training is highly efficient, with an accuracy of about 95% on the training set and about 99% on the testing set.

5.2 Actual Testing

The developed intelligent safety electricity monitoring terminal undergoes actual testing to verify its ability to accurately identify abnormal electricity events. Multiple tests for short circuits, leakage currents, and electric bicycle charging are designed, with identification results shown in Table 2. The short circuits and leakage tests in Table 2 are simulated using a circuit fault simulator, with 15 operations each. Electric bicycle charging types vary, such as 48V/3A and 60V/2A based on charging parameters, and lead-acid and lithium batteries based on battery types. This article selects 50 real electric bicycles for testing.

Design of Non-Intrusive Intelligent Safety Electricity Monitoring Terminal

Table 2 Abnormal Electricity Identification Results

6 Conclusion

This article studies the characteristics of typical abnormal electricity scenarios such as short circuits, leakage currents, and electric bicycle charging, combining the safety electricity monitoring terminal with non-intrusive load monitoring technology to develop an intelligent safety electricity monitoring terminal. By employing an advanced multi-core processing architecture, the terminal’s data processing and analysis performance is greatly enhanced, ensuring a short detection time for abnormal electricity, thereby ensuring real-time detection.

By applying the intelligent safety electricity monitoring terminal, monitoring and timely alarms for abnormal electricity behaviors in key areas are achieved, which is of great significance for improving the safety control level of electricity loads, reducing unauthorized wiring, enhancing the safety of electrical equipment, and lowering fire risks in scenic areas.