The portable solar generator's fault self-diagnosis system achieves precise fault location and rapid diagnosis through multi-dimensional information collection and intelligent analysis technology. Its core logic lies in integrating hardware sensing, algorithm processing, and data interaction capabilities to form a closed-loop diagnostic system covering the entire process of power generation, energy storage, and power supply. The following analysis focuses on the technical implementation path:
The system first relies on a heterogeneous multimodal sensing module to build its data foundation. This module uses various sensors deployed on key components such as solar panels, charge controllers, energy storage batteries, and inverters to collect parameters such as voltage, current, temperature, and light intensity in real time. For example, the photosensor on the surface of the solar panel can monitor actual lighting conditions. Combined with the electrical performance data output by the current and voltage sensors, it can determine whether there is a decrease in power generation efficiency due to dirt obstruction or component aging. The temperature sensor of the battery pack works in conjunction with the voltage sampling circuit to identify common faults such as overcharging and over-discharging, and can also detect potential internal short-circuit risks through abnormal temperature rise rates. This multi-parameter fusion sensing method effectively avoids misjudgment based on a single signal, providing comprehensive and reliable data support for subsequent diagnosis.
The data preprocessing stage uses a memristor-based in-memory computing unit to achieve edge computing. Traditional diagnostic systems require uploading massive amounts of raw data to the cloud for processing, resulting in high latency and high bandwidth consumption. Memristor technology, however, integrates storage and computation functions near the sensing module, enabling real-time filtering, feature extraction, and preliminary classification of collected data. For example, a sliding window algorithm can filter out transient light fluctuations caused by cloud cover, marking and uploading only persistently abnormal data; fast Fourier transform can extract harmonic components from current signals to preliminarily determine whether there are inverter switching transistor faults or grid pollution issues. This architecture significantly reduces invalid data transmission and improves system response speed.
The multi-scale causal diagnostic engine is the core algorithm module for fault location. This engine combines deep learning and causal graph reasoning techniques to construct a three-layer diagnostic model covering the component level, module level, and system level. At the component level, convolutional neural networks analyze the distortion characteristics of the solar panel's IV curve to accurately identify physical defects such as microcracks and broken grids. At the module level, a fault probability map of each sub-circuit of the charging controller is established based on a Bayesian network. When an abnormal output current is detected despite a normal input voltage, the fault in the power transistor drive circuit can be quickly identified. At the system level, a dynamic causal model is used to analyze the matching relationship between power generation and load demand. When a continuous power shortage occurs, it can distinguish whether it is caused by insufficient sunlight, battery aging, or inverter efficiency degradation. This hierarchical and progressive analysis strategy enables the system to not only locate specific fault points but also reveal the fault propagation path.
An adaptive topology communication network ensures reliable transmission of diagnostic data. Portable solar generators are often used in complex environments such as fieldwork and disaster sites. Traditional wired communication is easily damaged, while wireless communication faces signal obstruction problems. To address this, the system adopts a hybrid networking approach: high-speed interconnection between sensors and the main control board is achieved through flexible printed circuit boards within the device; when communicating between devices or with a remote monitoring platform, it automatically switches between LoRa, 4G/5G, and other multi-mode wireless modules and integrates Mesh self-organizing network functionality. When a communication node fails, the system can dynamically adjust the routing path to ensure real-time interaction between diagnostic commands and status data. For example, in mountainous application scenarios, a portable solar generator cluster can forward data tier by tier to a mountaintop base station via relay nodes, avoiding communication interruptions caused by terrain obstruction.
The digital twin monitoring platform provides visualized operation and maintenance support. Based on a 3D model of the physical generator, the platform maps the equipment's operating status in real time and intuitively displays the fault location and impact range through color coding and animation demonstrations. Maintenance personnel can view key indicators such as solar panel hotspot distribution, battery pack SOC curves, and inverter efficiency heatmaps through the platform without going to the site, and take targeted measures based on the system's automatically generated maintenance suggestions. For example, when the diagnostic engine locates an abnormal power generation efficiency of a solar panel, the platform highlights the component's location and retrieves its historical power generation data, installation angle information, and surrounding environmental parameters to help determine whether the problem is caused by component aging, installation tilt, or tree obstruction, significantly improving maintenance efficiency.
The intelligent evolutionary decision module enables the system to have self-learning capabilities. By continuously accumulating fault cases and maintenance records, the system can automatically optimize diagnostic model parameters and improve its ability to identify new fault types. For example, when connector contact problems caused by salt spray corrosion frequently occur in a certain area, the system will adjust the inspection cycle of equipment in that area and increase the frequency of monitoring connector resistance values; when it is found that a batch of battery packs is prone to capacity fluctuations within a specific temperature range, it will proactively correct its SOC estimation algorithm. This self-improvement mechanism based on experience feedback enables the system to continuously adapt to the differentiated needs of different application scenarios.
A self-reconfigurable sensor network further enhances the system's robustness. Addressing the issue of sensor position changes caused by frequent movement of portable devices, the system employs wirelessly reconfigurable sensor nodes. These nodes have self-localization capabilities and can automatically determine their relative position within the device using ultrasonic or UWB technology, dynamically adjusting data acquisition strategies. For example, when a battery pack is relocated, the temperature sensor will automatically re-register to the new monitoring zone, avoiding fault location errors due to misjudgment of location; when a sensor fails, surrounding nodes will proactively expand the sampling range to compensate for data loss, ensuring diagnostic continuity. This flexible network architecture enables the system to efficiently respond to dynamic changes during the use of portable devices.
