Portable photovoltaic energy storage system fault arc fire multi-mode early warning device

Abstract

The invention relates to a portable photovoltaic energy storage system fault arc fire multi-mode early warning device. According to the device, a portable automatic fault arc simulation generation device is built through a Raspberry Pi, a stepping motor, an arc generator and a mobile power supply; a photovoltaic energy storage system fault arc multi-mode information acquisition system is established based on a Raspberry Pi, a camera module, a smog sensor, a flame sensor, a temperature sensor, a current sensor, a voltage sensor and the like, and a DC fault arc fire multi-mode information database is established. The photovoltaic energy storage system fault arc multi-mode early warning method is designed to solve the problems that information is incomplete and context is missing due to the fact that single-mode text data is combined with a machine learning classification algorithm and image data is combined with a deep neural network classification algorithm. The multi-modal fusion has the advantage of information complementation, and especially has stronger anti-interference capability for multi-modal early warning in a monitoring environment with strong electromagnetic interference and frequent illumination sudden change interference.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical fire early warning, and in particular to a portable photovoltaic energy storage system fault arc fire multi-modal early warning device. Background Art

[0002] As one of the countries' vigorous efforts to develop clean energy, the photovoltaic energy storage industry is expanding in scale. At the same time, related fault problems are becoming increasingly prominent. DC fault arc fire accidents frequently occur in photovoltaic energy storage systems due to aging lines, poor contact, animal gnawing and other problems.

[0003] Due to the randomness, complexity and suddenness of fault arcs, it is difficult to obtain fault arc data in photovoltaic energy storage systems. The research on fault arc detection, positioning and early warning technology requires a large amount of experimental data support. More importantly, the complex indoor and outdoor experimental environment of photovoltaic energy storage systems requires that the experimental equipment must be lightweight and portable. The differences in fault arc characteristics under different working conditions require the automatic control of the experimental equipment to avoid experimental errors and uncontrollability of experimental parameters caused by manual operation of the arc generating device.

[0004] Traditional arc fault detection and early warning technologies often rely on single-mode text-type data such as (voltage, current, smoke, temperature, flame, etc.), and realize arc fault detection and early warning functions by performing data processing, feature extraction, model training, fault classification and other operations on electrical signals or environmental signals. The present invention combines image modal data and text modal data to propose a DC arc fault fire multimodal information acquisition system to solve the problem of missing information in single-mode data. The multimodal information acquisition system is characterized by information complementation and feature enhancement. Images provide visual information and text provides detailed descriptions. It has a more accurate and comprehensive description of the early characteristics of DC arc faults in photovoltaic energy storage systems.

[0005] Photovoltaic energy storage systems have diverse environments and complex lines. When a DC fault arc occurs in the system, risk assessment and fire warning of the fault arc are required based on factors such as the laying of surrounding lines, the location of flammable materials, the degree of insulation aging, and the presence of key equipment. Current fire warning technologies mainly rely on environmental monitoring equipment such as smoke and sudden temperature rise, which are effective in small indoor environments, but difficult to quickly and accurately achieve fire warnings in complex and wide outdoor environments. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a portable photovoltaic energy storage system fault arc fire multimodal warning device. Multimodal fusion has the advantage of information complementarity, especially in the monitoring environment with strong electromagnetic interference and frequent light mutation interference, the multimodal warning has stronger anti-interference ability.

[0007] To achieve the above-mentioned purpose, the present invention adopts the following technical solutions: a portable photovoltaic energy storage system fault arc fire multi-modal early warning device, including a photovoltaic energy storage system portable automatic precision control fault arc simulation generation device and a photovoltaic energy storage system fault arc fire multi-modal data acquisition system based on Raspberry Pi;

[0008] The portable automatic precise control fault arc simulation device for photovoltaic energy storage system comprises an embedded device Raspberry Pi, a stepper motor, a driver, and a mobile power supply; the mobile power supply is connected to the embedded device Raspberry Pi and the driver; the embedded device Raspberry Pi and the driver are connected to the stepper motor;

[0009] The multimodal data acquisition system for arc fault fires in photovoltaic energy storage systems based on Raspberry Pi collects image type data of arc faults based on the Raspberry Pi camera module, and collects text type data of arc faults based on current, voltage, smoke, flame, and temperature sensors, so as to establish a multimodal arc fault database.

[0010] In a preferred embodiment, the mobile power supply provides working power to the Raspberry Pi and the driver through the power conversion module; the driver is connected to the stepper motor A and B phases to enable the motor, and the pulse frequency f sent by the Raspberry Pi controls the electrode movement speed v, and the relationship is shown in formula (1):

[0011]

[0012] The step angle of the stepper motor is α, the number of driver subdivisions is N, the lead of the screw is d, and the sending frequency of the Raspberry Pi is f.

[0013] In a preferred embodiment, the multi-modal information acquisition of fault arc on Raspberry Pi includes: 1) Hardware preparation: high-performance Raspberry Pi, camera module for image acquisition, current and voltage sensor ADS1115, smoke sensor MQ-2, temperature sensor DHT22, flame sensor; 2) Software installation: Raspberry Pi OS, Python programming environment, sensor driver and library; 3) Sensor connection with Raspberry Pi: camera module connected to CSI interface, current and voltage sensor connected to I2C interface, smoke sensor connected to ADC interface, temperature and flame sensor connected to GPIO interface; 4) Image acquisition: configure the camera module to perform image acquisition through OpenCV; 5) Multi-modal data integration and synchronous acquisition: use multi-threaded synchronous acquisition technology to ensure the real-time and synchronization of data; 6) Data storage: data is stored in Raspberry Pi and cloud.

[0014] In a preferred embodiment, various sensors and camera modules are used to collect multimodal data of fault arcs, and the electrical parameters and environmental parameters of photovoltaic systems and energy storage equipment are monitored in real time. Then, the collected data is preprocessed by filtering and denoising. Then, feature extraction is performed on text type data, and features such as arc area, flame color, background change, cremation and arc recognition are extracted from image data. After the image and text modal features are Transformer encoded, an attention mechanism is applied to different modal data based on the Co-attention model: the mutual attention of the two modal features is calculated, that is, an attention matrix of the correlation between the modalities is generated, and the features of each modality are weighted and synthesized into a new feature set according to the matrix, and the weighted current, voltage, smoke, temperature, flame, and image features are fused. Furthermore, the multimodal fusion features are input into a deep neural network for training and data enhancement and regularization techniques are used to improve the generalization ability of the model. Finally, the fault arc detection model is deployed to the Raspberry Pi for real-time monitoring and fire warning of the photovoltaic energy storage system.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] 1. Design a portable automatic and precise control device for simulating fault arc generation in photovoltaic energy storage system. The device is mainly composed of Raspberry Pi, an embedded device, a stepper motor, an electrode module, and a mobile power supply. The Raspberry Pi program is used to precisely control the arc speed and arc spacing of the electrodes to simulate the fault arc under different working conditions. Compared with manual arc drawing, arc drawing controlled by a stepper motor has the advantages of strong safety, high reliability, and high precision. The device constructed with lightweight equipment and a mobile power supply can meet the experimental conditions of photovoltaic systems and energy storage equipment in complex indoor and outdoor environments.

[0017] 2. Build a multi-modal data acquisition system for arc fault fires in photovoltaic energy storage systems based on Raspberry Pi. The system collects image data of arc faults based on the Raspberry Pi camera module, and collects text data of arc faults using sensors such as current, voltage, smoke, flame, and temperature, to establish a multi-modal arc fault database. Data from different modes complement each other to provide more comprehensive and rich information for early warning of electrical fires caused by arc faults in photovoltaic energy storage systems. By combining image signals with electrical signals and environmental signals, multi-dimensional and intelligent fire early warnings for photovoltaic energy storage systems can be achieved.

[0018] 3. Design a fault arc fire warning method based on multimodal information fusion technology. Install Raspberry Pi devices in the photovoltaic energy storage system where fault arcs are prone to occur, high-risk, and critical equipment areas, and synchronously collect and preprocess multimodal data, extract features, fuse features, train fault arc fire models, and conduct real-time monitoring and warning. Deploy the trained model to the Raspberry Pi, monitor the real-time data, and use the model to determine whether there is a fault arc. The fault arc detection results are combined with smoke concentration and image information to trigger the warning mechanism, notifying maintenance personnel to respond in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic diagram of the overall design concept of fire warning according to a preferred embodiment of the present invention;

[0020] Figure 2 A schematic diagram of a fault arc simulation generating device according to a preferred embodiment of the present invention;

[0021] Figure 3 A schematic diagram of a multimodal data collection process according to a preferred embodiment of the present invention;

[0022] Figure 4 The figure is a flow chart of a fire warning method according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0025] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0026] refer to Figure 1-4, a portable multi-modal early warning device for DC fault arc fire in photovoltaic energy storage system. This method builds a portable and automated fault arc simulation device based on Raspberry Pi. The portable design is designed to cope with the complex and diverse application environment of photovoltaic energy storage system. The automated design customizes the arc pulling speed and electrode spacing according to the fault simulation requirements under different working conditions, solving the problem of difficulty in obtaining fault arc signals in photovoltaic energy storage system; the fault arc image modal information is obtained through the Raspberry Pi camera, and the fault arc text modal information is collected by current, voltage, smoke, flame, and temperature sensors. A multi-modal acquisition system for DC fault arc in photovoltaic energy storage system is built, and a multi-modal database for fault arc is established; the multi-modal database of fault arc is combined with multi-modal information fusion technology to construct a fault arc risk assessment system and a fire warning system, providing theoretical guidance for the research on early warning technology of fault arc fire in photovoltaic energy storage system.

[0027] (1) Portable automatic and precise control device for simulating fault arc generation in photovoltaic energy storage system

[0028] Due to the randomness, complexity, and suddenness of fault arcs, there are problems such as difficulty in data acquisition, low data authenticity, and single data type. The fault arc simulation device has the characteristics of lightweight portability and automatic and precise control. The design ideas are as follows:

[0029] The fault arc simulation device is mainly composed of four parts: arc generator, Raspberry Pi, stepper motor controller, and stepper motor. It is used to simulate the fault arc caused by insulation aging, poor contact, animal gnawing, etc. Figure 2 As shown, it is divided into two units, the control unit in the left box diagram and the actuator (arc generator) on the right. The arc generator is designed according to "GB / T39750-2021 Technical Requirements for DC Arc Protection of Photovoltaic Power Generation Systems".

[0030] The mobile power supply provides working power for the Raspberry Pi (5V, 3A) and the driver (24V, 3A) through the power conversion module. The driver is connected to the stepper motor A and B phases to enable the motor, and the pulse frequency T sent by the Raspberry Pi controls the electrode movement speed v. The relationship is shown in formula (1):

[0031]

[0032] The step angle of the stepper motor is α, the subdivision number of the driver is N, the lead of the screw is d, and the sending frequency of the Raspberry Pi is T. The fault simulation device can meet all the arc test conditions listed in the standard as shown in Table 1 below.

[0033] Table 1 Arc test conditions

[0034]

[0035]

[0036] The arc rate and arc spacing can be customized according to the experimental requirements. The arc rate setting range is 0-10mm / s, and the arc spacing setting range is 0-100mm. During the experimental operation, the experimental parameter settings are completed using a small-volume matrix keyboard.

[0037] (2) Photovoltaic energy storage system fault arc fire multimodal data acquisition system based on Raspberry Pi

[0038] The photovoltaic energy storage system is a complex process from the generation of fault arcs to the initiation of electrical fires. The evolution of the fault arc can be divided into the arc starting stage, arc burning stage, and stable combustion stage. Since the DC fault arc does not cross the zero point and is difficult to extinguish naturally, the mutual influence of the fault arc and the external environment should be considered, such as arc extinguishing in a windy environment, the expansion of the fire area due to the accumulation of flammable materials around, and the thermal decomposition of cable insulation caused by the fault arc. It is difficult to obtain the above information from the current and voltage signals. The smoke and temperature signals have a lag, that is, the effect is not shown until the middle and late stages of the fire, making it difficult to achieve early warning. In addition, the detection effect is poor in the outdoor environment of the photovoltaic system due to poor airtightness.

[0039] Image data acquisition is performed based on Raspberry Pi to monitor the evolution of the fault arc and the interaction between the fault arc and the environment. Current and voltage data are collected to obtain early information of the fault arc, and environmental signals such as smoke, temperature, and flame are collected to obtain late information of the fault arc evolution.

[0040] like Figure 3 As shown in the figure, the main tasks of realizing multi-modal information collection of fault arc on Raspberry Pi are as follows: 1. Hardware preparation: high-performance Raspberry Pi, camera module for image acquisition, current and voltage sensor ADS1115, smoke sensor MQ-2, temperature sensor DHT22, flame sensor; 2. Software installation: Raspberry Pi OS, Python programming environment, sensor driver and library; 3. Sensor connection with Raspberry Pi: camera module connected to CSI interface, current and voltage sensor connected to I2C interface, smoke sensor connected to ADC interface, temperature and flame sensor connected to GPIO interface; 4. Image acquisition: configure the camera module to acquire images through OpenCV; 5. Multi-modal data integration and synchronous acquisition: use multi-threaded synchronous acquisition technology to ensure the real-time and synchronization of data; 6. Data storage: data is stored in Raspberry Pi and cloud.

[0041] (3) Fault arc fire warning method based on multimodal information fusion technology

[0042] Since photovoltaic systems and energy storage equipment are exposed to various environments and working conditions for a long time, insulation aging, poor contact, line damage, animal gnawing and other problems lead to frequent fault arcs, thus causing fires. The existing early warning schemes mainly include fire warning methods that combine text modal data (electrical parameters, environmental parameters) with machine learning, and fire warning methods that combine image data with deep learning. The input data sources are all single-modal data. The fault arc fire warning method that uses multimodal information fusion technology fuses text modal data with image modal data to improve the reliability, accuracy and anti-interference ability of fault arc detection and early warning.

[0043] The implementation process of the multi-modal early warning method for arc fire in photovoltaic energy storage system is as follows: Figure 4 As shown in the figure, various sensors and camera modules are used to collect multimodal data of fault arcs, and the electrical parameters and environmental parameters of photovoltaic systems and energy storage equipment are monitored in real time. Then, the collected data is preprocessed by filtering and denoising. Then, the text type data is feature extracted, such as current, voltage peak, mean, frequency centroid, waveform distortion rate temperature anomaly, etc., and the image data is extracted with features such as arc area, flame color, background change, cremation and arc recognition. After the image and text modal features are Transformer encoded, the attention mechanism is applied to different modal data based on the Co-attention model: the mutual attention of the two modal features is calculated, that is, the attention matrix of the correlation between the modalities is generated, and the features of each modality are weighted and synthesized into a new feature set according to the matrix. The weighted current, voltage, smoke, temperature, flame, and image features are fused. Then, the multimodal fusion features are input into the deep neural network for training and data enhancement and regularization techniques are used to improve the generalization ability of the model. Finally, the fault arc detection model is deployed to the Raspberry Pi for real-time monitoring and fire warning of the photovoltaic energy storage system.

Claims

1. A portable photovoltaic energy storage system fault arc fire multi-modal early warning device, characterized in that: Including a portable automatic precision control fault arc simulation device for photovoltaic energy storage systems and a multi-modal data acquisition system for fault arc fires in photovoltaic energy storage systems based on Raspberry Pi; The portable automatic precise control fault arc simulation device for photovoltaic energy storage system comprises an embedded device Raspberry Pi, a stepper motor, a driver, and a mobile power supply; the mobile power supply is connected to the embedded device Raspberry Pi and the driver; The embedded device Raspberry Pi and the driver are connected to the stepper motor; The multimodal data acquisition system for arc fault fires in photovoltaic energy storage systems based on Raspberry Pi collects image type data of arc faults based on the Raspberry Pi camera module, and collects text type data of arc faults based on current, voltage, smoke, flame, and temperature sensors, and establishes a multimodal arc fault database.

2. A portable photovoltaic energy storage system fault arc fire multi-modal early warning device according to claim 1, characterized in that: The mobile power supply provides working power for the Raspberry Pi and the driver through the power conversion module; the driver connects the stepper motor A and B phases to enable the motor, and controls the electrode movement speed v by sending a pulse frequency f through the Raspberry Pi. The relationship is shown in formula (1): The step angle of the stepper motor is α, the number of driver subdivisions is N, the lead of the screw is d, and the sending frequency of the Raspberry Pi is f.

3. A portable photovoltaic energy storage system fault arc fire multi-modal early warning device according to claim 2, characterized in that: The realization of multi-modal information acquisition of fault arc on Raspberry Pi includes: 1) Hardware preparation: high-performance Raspberry Pi, camera module for image acquisition, current and voltage sensor ADS1115, smoke sensor MQ-2, temperature sensor DHT22, flame sensor; 2) Software installation: Raspberry Pi OS, Python programming environment, sensor driver and library; 3) Sensor connection with Raspberry Pi: camera module connected to CSI interface, current and voltage sensors connected to I2C interface, smoke sensor connected to ADC interface, temperature and flame sensor connected to GPIO interface; 4) Image acquisition: configure the camera module to acquire images through OpenCV; 5) Multi-modal data integration and synchronous acquisition: use multi-threaded synchronous acquisition technology to ensure the real-time and synchronization of data; 6) Data storage: data is stored in Raspberry Pi and cloud.

4. A portable photovoltaic energy storage system fault arc fire multi-modal early warning device according to claim 3, characterized in that: The multimodal data of fault arcs are collected through various sensors and camera modules to monitor the electrical parameters and environmental parameters of photovoltaic systems and energy storage equipment in real time. Then, the collected data is preprocessed by filtering and denoising. Then, the text type data is feature extracted, and the image data is extracted with features such as arc area, flame color, background change, cremation and arc recognition. After the image and text modal features are Transformer encoded, the attention mechanism is applied to the different modal data based on the Co-attention model: the mutual attention of the two modal features is calculated, that is, the attention matrix of the correlation between the modalities is generated, and the features of each modality are weighted according to the matrix to synthesize a new feature set, and the weighted current, voltage, smoke, temperature, flame, and image features are fused. Then, the multimodal fusion features are input into the deep neural network for training and data enhancement and regularization techniques are used to improve the generalization ability of the model. Finally, the fault arc detection model is deployed to the Raspberry Pi for real-time monitoring and fire warning of the photovoltaic energy storage system.

Images