A traffic equipment solar power supply fault diagnosis system

Abstract

This invention belongs to the field of intelligent transportation and new energy power supply technology, specifically relating to a fault diagnosis system for solar power supply in transportation equipment. It aims to solve the problems of untimely and inaccurate fault diagnosis in solar power supply systems, as well as the difficulty in distinguishing between transient interference and persistent faults. The system collects multi-source data through environmental perception and electrical parameter monitoring modules, combines dynamic expected power modeling with actual output comparison to generate a power deviation sequence and extract time-series features. The fault identification module employs multi-level logical discrimination and a 12-hour continuous verification mechanism to accurately identify photovoltaic panel contamination, battery aging, poor line contact, and controller faults, and distinguishes transient interference. The decision-making alarm module generates graded alarms based on fault type and level, and links with geographic information to achieve precise location and operation and maintenance scheduling. The system also features internal resistance pulse detection, dual-channel redundant sampling, adaptive threshold adjustment, and model self-learning functions, significantly improving diagnostic accuracy and operation and maintenance efficiency.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation and new energy power supply technology, and specifically relates to a fault diagnosis system for solar power supply of transportation equipment. Background Technology

[0002] In the field of intelligent transportation systems, the energy supply and operational reliability of transportation equipment are core elements for ensuring the efficient and safe operation of urban transportation. With the development of green energy technologies, solar energy, as a clean and renewable energy source, is widely used in the power supply systems of various transportation equipment, including traffic lights, surveillance cameras, and electronic signs. Solar power systems convert solar energy into electrical energy through photovoltaic arrays and combine them with energy storage devices to achieve continuous power supply day and night. Their deployment significantly reduces dependence on the traditional power grid and improves the environmental adaptability and sustainability of transportation infrastructure.

[0003] The stability of the solar power system directly determines the continuous availability of traffic equipment. Since traffic equipment is mostly deployed in complex outdoor environments, it is exposed to adverse conditions such as high temperature, high humidity, dust, and shading for extended periods. This can cause photovoltaic modules to experience performance degradation, connection failures, or localized hot spots, while energy storage batteries may age prematurely due to improper charging and discharging management. If these problems are not detected and addressed in a timely manner, they can lead to power outages, resulting in serious consequences such as traffic signal malfunctions and monitoring failures, impacting traffic order and public safety.

[0004] While some existing systems are equipped with basic voltage and current monitoring functions, significant shortcomings remain: fault detection relies on manual inspections or simple threshold alarms, lacking the ability to deeply integrate and analyze multi-source operational data; the ability to identify early latent faults (such as minor component cracks or junction box aging) is insufficient, making it difficult to accurately locate and classify faults; simultaneously, the system lacks the ability to model the coupling relationship between environmental factors (such as light intensity and temperature) and electrical parameters, resulting in a high false alarm rate and delayed diagnosis. Therefore, there is an urgent need for a fault diagnosis solution that can achieve full-link, intelligent, and highly robust fault diagnosis for solar power supply systems in transportation equipment. Summary of the Invention

[0005] The purpose of this invention is to provide a fault diagnosis system for solar power supply in transportation equipment, addressing the technical problems of untimely fault diagnosis, low diagnostic accuracy, and difficulty in distinguishing between transient interference and persistent faults in existing solar power supply systems for transportation equipment. Currently, with the rapid development of intelligent transportation systems, various traffic lights, monitoring equipment, electronic signs, and other transportation equipment widely rely on solar power supply systems for energy self-sufficiency. However, because these devices are mostly deployed in complex outdoor environments, exposed to sunlight, rain, dust, and drastic temperature changes for extended periods, their solar power supply systems are prone to faults such as photovoltaic panel surface contamination, aging connection lines, battery capacity degradation, and abnormal charge / discharge controllers. In existing technologies, most systems only possess simple voltage or current threshold alarm functions, failing to accurately identify fault types, let alone determine the persistence and severity of faults. This leads to frequent, ineffective inspections by maintenance personnel, or failure to respond promptly when serious faults occur, severely impacting the stable operation of transportation equipment and public safety.

[0006] The technical solution of this invention is a fault diagnosis system for solar power supply in transportation equipment, comprising an environmental perception module, an electrical parameter monitoring module, a status assessment module, a fault identification module, and a decision alarm module. The environmental perception module is used to collect real-time data on light intensity, ambient temperature, precipitation, and wind speed in the deployment area. The electrical parameter monitoring module is used to continuously monitor the output voltage and current of the solar photovoltaic panel, the terminal voltage, charging and discharging current, internal resistance value of the battery, and the operating status signal of the charging and discharging controller. The status assessment module receives raw data from the environmental perception module and the electrical parameter monitoring module, dynamically models the theoretical power generation capacity of the photovoltaic panel, and calculates the expected output power of the photovoltaic panel based on historical data and current environmental parameters. Furthermore, the status assessment module compares the actual output power collected by the electrical parameter monitoring module with the expected output power over time periods to generate a power deviation sequence, and performs time-series sliding window statistical analysis on the power deviation sequence to extract the mean, standard deviation, and slope of the deviation as primary state features.

[0007] The fault identification module performs multi-level fault logic judgment based on primary state characteristics. When the average power deviation continuously exceeds the first preset threshold and the standard deviation is low, it is determined that there is continuous shading or contamination on the photovoltaic panel surface; when the average power deviation does not exceed the limit but the standard deviation increases significantly, it is determined that there is poor line contact or intermittent open circuit; when the battery terminal voltage cannot reach the preset float charge voltage during the charging stage and the internal resistance continuously increases, combined with the distortion characteristics of the charging and discharging current waveform, it is determined that the battery is aging or failed; when the output signal of the charging and discharging controller is abnormal and independent of the photovoltaic input and battery status, it is determined that the controller itself is faulty. The fault identification module further introduces a continuous verification mechanism to perform continuous verification of the initially determined fault types in the time dimension. Specifically, the fault confirmation window duration is set to 12 hours. If the judgment result of the same fault type occurs more than 8 times within this duration, the fault is confirmed as a continuous fault; if the occurrence is less than 3 times and the interval is random, it is determined as transient interference and no alarm is triggered.

[0008] The decision-making alarm module receives the fault confirmation results from the fault identification module and generates differentiated alarm commands based on the fault type and severity level. For photovoltaic panel pollution faults, a low-priority maintenance reminder is generated, suggesting cleaning to be scheduled within the next sunny weather cycle; for medium-severity faults such as battery aging or poor line contact, a medium-priority alarm is generated, requiring on-site verification and component replacement within 48 hours; for high-severity faults such as charge / discharge controller failure or complete battery failure, a high-priority emergency alarm is generated, immediately triggering remote control commands to cut off the load to protect the system, and simultaneously notifying the operation and maintenance platform to dispatch personnel for emergency repairs. The decision-making alarm module also integrates geographic information association functionality, binding alarm information with the device's unique code and geographical location, enabling precise location of faulty equipment and intelligent scheduling of operation and maintenance resources.

[0009] Preferably, the expected output power calculation process in the status assessment module is as follows: First, the solar altitude angle and azimuth angle are calculated based on the latitude and longitude of the equipment installation location and the current date and time; second, the rated power of the photovoltaic panel under standard test conditions is corrected by combining the real-time light intensity and atmospheric transparency coefficient provided by the environmental perception module; third, a temperature compensation factor is introduced, and the output power is corrected by a negative temperature coefficient based on the mapping relationship between ambient temperature and photovoltaic panel surface temperature; finally, a dust accumulation attenuation model is superimposed, which dynamically estimates the light transmittance loss based on historical precipitation, wind speed and the last cleaning time, thereby outputting a high-precision expected power value.

[0010] Preferably, the internal resistance value in the fault identification module is obtained as follows: within the first 5 minutes of each battery charging start-up, a pulse current with a duration of 30 seconds and an amplitude of 1.5 times the rated charging current is injected. Simultaneously, the change in battery terminal voltage before and after the pulse is collected, and the ratio of the voltage change to the pulse current amplitude is taken as the AC internal resistance value at that moment. This internal resistance value is sampled every 24 hours to form a time series for trend analysis.

[0011] Preferably, the sliding window length of the power deviation sequence is set to 6 hours, with a step size of 1 hour, and the statistical characteristics within the window are recalculated after each sliding. The standard deviation threshold is adaptively adjusted according to the season and geographical region, with the winter threshold set to 1.3 times that of summer to accommodate natural differences in light fluctuations.

[0012] Preferably, the differentiated alarm commands in the decision alarm module are sent to the operation and maintenance management terminal through the wireless communication network, and the audible and visual prompt device is triggered locally on the traffic equipment. The prompt type adopts different frequency flashing patterns and tone combinations according to the fault level, which facilitates quick identification by on-site personnel.

[0013] Preferably, the system has a data backtracking and model optimization sub-module, which periodically packages historical fault cases, environmental data, electrical parameters and finally manually confirmed fault types into a training sample set, which is used to update the discrimination logic parameters in the fault identification module, so as to realize the closed-loop self-learning and continuous performance improvement of the diagnostic model.

[0014] Preferably, the electrical parameter monitoring module adopts a dual-channel redundant acquisition architecture, which independently samples and cross-verifies key voltage and current signals. When the deviation between the two channel readings exceeds 2%, the self-diagnosis program is started and the data is marked as abnormal to prevent misjudgment due to sensor failure.

[0015] Preferably, the precipitation sensor in the environmental sensing module has a self-cleaning function, removing surface deposits through periodic ultrasonic vibration to ensure long-term measurement accuracy. The wind speed sensor uses ultrasonic wind measurement technology without mechanical bearings to avoid mechanical wear caused by sand and dust.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: This invention, by constructing a multi-source data fusion system integrating environmental perception and electrical monitoring, achieves comprehensive perception and dynamic modeling of the operating status of solar power supply systems, significantly improving the proactiveness and accuracy of fault diagnosis. The invention introduces a fault confirmation mechanism based on time-series statistics and continuous verification, effectively distinguishing between instantaneous power drops caused by environmental fluctuations and actual hardware faults, greatly reducing false alarm and false negative rates and avoiding waste of maintenance resources. This invention employs hierarchical fault identification logic and internal resistance pulse detection technology, accurately identifying various typical fault types such as photovoltaic panel contamination, poor line contact, battery aging, and controller failure. It also achieves intelligent hierarchical management of maintenance responses through differentiated alarm strategies, improving the timeliness and targeted nature of fault handling. This invention integrates geographic information binding and audio-visual prompts, enhancing the convenience and efficiency of on-site handling. The model self-learning mechanism of this invention enables the system to continuously optimize its diagnostic capabilities over time, extending the system's effective lifespan. Overall, this invention addresses the technical shortcomings of traditional solar power supply systems that prioritize power supply over diagnosis, providing a reliable guarantee for the long-term stable operation of transportation equipment, and has significant engineering application value and promising prospects for widespread adoption. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall technical architecture of the solar power supply fault diagnosis system for transportation equipment proposed in this invention. Detailed Implementation

[0018] Please refer to Figure 1 This invention relates to a fault diagnosis system for solar-powered transportation equipment. Its core objective is to address the technical bottlenecks in existing technologies, such as delayed fault diagnosis, high misjudgment rates, and difficulty in distinguishing between transient disturbances and persistent faults due to the interplay of environmental interference and hardware degradation. This system constructs a multi-source data fusion perception system integrating environmental and electrical data, combined with dynamic power modeling, time-series feature extraction, multi-level logical discrimination, and continuous verification mechanisms, to achieve accurate assessment of the operating status of solar-powered transportation equipment systems and intelligent identification of fault types. In this embodiment, the system is deployed within solar-powered traffic light control boxes along urban main roads, integrated into an embedded industrial-grade computing platform, possessing all-weather autonomous operation capabilities, and maintaining data synchronization with the municipal-level traffic operation and maintenance management platform via a 4G wireless communication network.

[0019] The system's overall technical process begins with the synchronous acquisition of multi-dimensional environmental parameters and key electrical quantities. The environmental sensing module is activated first, continuously acquiring information on light intensity, ambient temperature, precipitation, and wind speed in the equipment deployment area. Simultaneously, the electrical parameter monitoring module operates, performing high-precision sampling of the voltage and current at the photovoltaic panel output, the battery bank's terminal voltage, charging and discharging current, internal resistance, and the operating status codes output by the charging and discharging controller. These two types of data are transmitted in real-time to the status assessment module, where dynamic modeling of theoretical power generation capacity and quantitative comparison of actual output performance are completed, generating a primary state characteristic sequence reflecting system deviations from normal operating conditions. This sequence is then passed as input to the fault identification module, which performs a preliminary fault type determination based on preset multi-level criteria and introduces a continuity verification mechanism in the time dimension to confirm the persistence of the fault. Finally, the decision-making alarm module generates differentiated alarm commands and executes corresponding actions based on the confirmed fault type and its severity level, while binding the alarm information to the equipment's geographical coordinates to achieve precise scheduling of maintenance responses. The system also includes a data backtracking and model optimization submodule, which regularly collects historical operating data for updating diagnostic logic parameters, forming a closed-loop self-learning mechanism.

[0020] The environmental sensing module consists of four types of dedicated sensors, responsible for collecting four key environmental parameters: light intensity, ambient temperature, precipitation, and wind speed. The light intensity sensor uses a high-sensitivity silicon photodiode with a spectral response range covering 300 nm to 1100 nm, a sampling frequency set to once per minute, a measurement range of 0 lux to 200,000 lux, and a resolution of no less than 10 lux. The sensor surface is covered with an anti-reflective coating and a hydrophobic film to reduce the impact of dust adhesion and rainwater residue on measurement accuracy. The ambient temperature sensor uses a digital thermistor, encapsulated in a ventilated louvered box to avoid measurement deviations caused by direct sunlight. Its temperature measurement range is -40°C to 85°C, with an accuracy of ±0.5°C. The precipitation sensor uses a tipping bucket rain gauge combined with an ultrasonic self-cleaning device. The tipping bucket has a capacity of 0.2 mm of precipitation equivalent. Each tipping triggers a pulse signal, and the system calculates the cumulative precipitation by counting the number of pulses per unit time. The ultrasonic self-cleaning device is set to activate every 6 hours, vibrating continuously for 30 seconds to effectively remove leaves, insects, or dust clogging the sensor's water inlet surface. The wind speed sensor employs a mechanically-free anemometer based on the ultrasonic time-difference propagation principle, consisting of three orthogonally arranged ultrasonic transceiver units. It calculates the wind speed vector by measuring the difference in sound wave propagation time in the downwind and upwind directions, avoiding the jamming and wear caused by sand and dust intrusion into the bearings of traditional cup-type sensors. The wind speed measurement range is 0 m / s to 60 m / s, with a resolution of 0.1 m / s. All environmental sensors are equipped with lightning protection circuits and electromagnetic shielding housings to ensure data integrity and equipment safety during thunderstorms. The collected environmental data is timestamped, packaged into structured data frames, and transmitted to the status assessment module via an internal serial peripheral interface.

[0021] The electrical parameter monitoring module adopts a dual-channel redundant architecture design, implementing independent dual-channel acquisition and cross-verification for four key electrical quantities: photovoltaic panel output voltage, photovoltaic panel output current, battery terminal voltage, and charging / discharging current. Each signal's two acquisition channels are powered by independent signal conditioning circuits, analog-to-digital converters, and isolated power supplies, achieving physical isolation at the hardware level. Voltage signals are stepped down through a high-precision resistor divider network before being input to a differential amplifier, while current signals are sampled through a low-temperature-drift manganese-copper shunt and converted into standard voltage signals by an isolated operational amplifier. The analog-to-digital converter uses a 24-bit Σ-Δ converter chip with a sampling rate of 10 times per second and an effective bit depth of at least 20 bits to ensure the detectability of minute fluctuations. The system is configured to compare the dual-channel readings every 5 minutes. If the relative deviation of the same parameter between the two channels consistently exceeds 2%, it is determined that the sensor or acquisition link is abnormal, and a self-diagnostic program is initiated. The self-diagnostic procedure first switches to the channel with stable readings as the primary channel, simultaneously sending a "sensor health status abnormal" warning to the maintenance platform and freezing the parameter's eligibility for subsequent fault diagnosis to prevent cascading misjudgments caused by single-point sensor failure. For acquiring the battery's internal resistance, the module performs a specific pulse excitation operation: at 2:00 AM daily, when the system detects that the battery is in a constant current charging phase and the charging current is stable, it controls the charge / discharge controller to superimpose a pulse current with an amplitude 1.5 times the rated charging current and a duration of 30 seconds onto the original charging current. One second before and one second after the pulse injection, the high-precision voltage acquisition unit synchronously records the voltage values ​​across the battery terminals, calculates the voltage difference ΔU between the two measurements, and defines the ratio of ΔU to the pulse current amplitude I_pulse as the AC internal resistance R_ac at that moment, i.e., R_ac = ΔU / I_pulse. This internal resistance value is sampled once daily, forming a continuous time series for subsequent trend analysis. The charging and discharging controller's operating status signal is reported every 30 seconds via the controller's built-in communication interface in the form of a heartbeat packet. The signal includes fields such as operating mode, fault code, and temperature protection status, serving as a direct basis for determining whether the controller itself is malfunctioning.

[0022] The state assessment module is the core component for achieving accurate state perception in the system. Its function is to integrate environmental parameters and electrical monitoring data to establish a dynamic prediction model of the theoretical output power of the photovoltaic panel, and generate quantifiable state deviation characteristics by comparing it with the actual output power. The module's internal workflow consists of four stages: solar geometric parameter calculation, light intensity correction, temperature compensation, and dust attenuation modeling. In the first stage, based on the geographical latitude and longitude coordinates of the equipment installation location and the current UTC time, the astronomical algorithm library is used to calculate the sun's altitude and azimuth angles at that moment. The calculation process considers the Earth's orbital eccentricity, the obliquity of the ecliptic, and the precession effect, ensuring that the calculation error of the sun's position at any time of year is less than 0.1 degrees. In the second stage, combined with the real-time light intensity measurement and atmospheric transparency coefficient provided by the environmental perception module, the rated power P_STC of the photovoltaic panel under standard test conditions is linearly corrected. The atmospheric transparency coefficient is obtained from tables based on the daily precipitation, air humidity, and visibility data, and is used to characterize the degree of atmospheric scattering and absorption of solar radiation. The corrected theoretical power P_irrad is calculated using the formula... Given, where I_actual is the measured light intensity, and I_STC is the light intensity under standard test conditions of 1000 lux. The atmospheric transparency coefficient ranges from 0.7 to 1.0. In the third stage, a temperature compensation factor is introduced. The output power of the photovoltaic panel has a negative temperature coefficient, typically -0.4% per degree Celsius. The module estimates the actual operating temperature T_panel of the photovoltaic panel based on ambient temperature sensor readings and the impact of wind speed on heat dissipation from the panel surface. The temperature compensation factor k_temp is calculated by k_temp = 1 + α × (T_panel - T_STC), where α is the power temperature coefficient and T_STC is the standard test temperature of 25 degrees Celsius. The final temperature-corrected theoretical power P_temp = P_irrad × k_temp. In the fourth stage, a dust accumulation attenuation model is superimposed. The model uses the last cleaning time t_clean as a baseline, combined with historical daily precipitation and average wind speed data, to dynamically estimate the light transmittance loss on the photovoltaic panel surface. Specifically, after each effective precipitation (daily rainfall greater than 2 mm), the system assumes that surface dust has been partially washed away, and the light transmittance recovers to a certain proportion. During periods without precipitation, the light transmittance decays exponentially over time, with the decay rate adjusted by the average daily wind speed; the higher the wind speed, the slower the decay. The current light transmittance η_dust is calculated using the empirical formula η_dust=exp(-k_decay×(t_now-t_clean) / v_wind), where k_decay is the baseline decay coefficient, t_now is the current time, and v_wind is the recent average wind speed. The final expected output power P_expected=P_temp×η_dust. This expected power is updated every 15 minutes and compared point-by-point with the actual output power P_actual reported by the electrical parameter monitoring module to generate a 24-hour power deviation sequence δP(t)=P_actual(t)-P_expected(t).

[0023] After obtaining the power deviation sequence, the state assessment module further performs time-series sliding window statistical analysis. The sliding window length is set to 6 hours, with a step size of 1 hour, meaning that the window advances by 1 hour every hour, recalculating the statistical characteristics of all deviation data within the window. For each sliding window, the module calculates three primary state characteristics: the deviation mean μ_δP, the deviation standard deviation σ_δP, and the change slope k_δP. The deviation mean reflects the degree of degradation of the overall output capability of the system during that period; the deviation standard deviation characterizes the severity of output fluctuations, with high variance potentially indicating poor contact or intermittent open circuits; the change slope is obtained by fitting the time-series trend line of the deviation data within the window using a linear regression method, used to capture signs of acceleration or mitigation of performance degradation. The calculation process of statistical features is as follows: Assume the sliding window contains N power deviation sampling points δP_1 to δP_N, with timestamps t_1 to t_N. Then the mean μ_δP = (1 / N) × Σ(δP_i), the standard deviation σ_δP = √[(1 / (N-1)) × Σ(δP_i - μ_δP) squared], and the slope k_δP is obtained by the least squares method, satisfying k_δP = [N × Σ(t_i × δP_i) - Σt_i × ΣδP_i] / [N × Σ(t_i squared) - (Σt_i) squared]. These primary state features, along with environmental parameters, battery internal resistance, and controller status, together constitute the input feature vector of the fault identification module.

[0024] The fault identification module performs multi-level fault logic discrimination based on the input feature vector and introduces a continuous verification mechanism in the time dimension to distinguish between transient interference and real faults. The module's internal logic is divided into a primary discrimination layer and a continuous verification layer. The primary discrimination layer matches the feature combinations at the current moment according to preset rules and outputs a preliminary fault type. The specific criteria are as follows: When the average power deviation μ_δP is consistently below -15% and the standard deviation σ_δP is less than 5%, it is determined that there is continuous shading or contamination on the surface of the photovoltaic panel. This type of fault causes the output power to be consistently low but with small fluctuations. When the average power deviation μ_δP is within the normal range (greater than -10%) but the standard deviation σ_δP exceeds 8%, it is determined that there is poor line contact or intermittent circuit breakage, which is manifested as frequent jumps in output power. When the battery terminal voltage fails to reach the preset float charge voltage (e.g., 13.8 volts) during the charging phase and the AC internal resistance measured on the day increases by more than 15% compared to the average of the previous 5 days, and the charging current waveform shows non-periodic distortion or interruption, it is determined that the battery is aging or has failed. When the operating status code reported by the charge and discharge controller contains preset fault codes (such as overvoltage, overcurrent, and temperature protection), and this abnormal state is independent of the changes in photovoltaic input power and battery voltage, that is, an error is still reported when both input and output are normal, it is determined that the controller itself is faulty.

[0025] Based on the initial discrimination, a continuous verification layer is initiated to perform continuous verification on the initially determined fault types over time to confirm them as persistent faults. The system sets the fault confirmation window to 12 hours, during which initial discrimination is performed once per hour. If the discrimination result of the same fault type occurs more than 8 times consecutively within 12 hours, the fault is confirmed as a persistent fault, and a fault confirmation signal is output. If the occurrences are less than 3 times and the time intervals between occurrences are random and irregular, it is determined to be transient interference, such as temporary cloud cover, transient lightning interference, or sensor noise. The system does not generate an alarm, but only records the event log for subsequent analysis. The system adaptively adjusts the standard deviation threshold according to the season and geographical region. In the temperate regions of the Northern Hemisphere, the natural fluctuation of sunlight is large in winter. The standard deviation threshold is set to 1.3 times that of summer. For example, if the summer threshold is 8%, it is adjusted to 10.4% in winter to avoid false alarms caused by normal seasonal fluctuations. The regional adjustment factor is calculated based on historical meteorological data of the equipment deployment location and is pre-written into the system configuration table.

[0026] The decision-making alarm module receives the fault confirmation results output by the fault identification module, generates differentiated alarm commands based on the fault type and severity level, and executes corresponding control actions. For confirmed photovoltaic panel pollution faults, the system classifies them as low priority, generates a maintenance reminder command, and pushes it to the operation and maintenance management terminal via the wireless communication network, suggesting that cleaning operations be scheduled during the next consecutive sunny weather period to avoid ineffective operations on rainy or windy days. For battery aging or poor line contact faults, the system classifies them as medium severity level, generates a medium priority alarm, and requires operation and maintenance personnel to arrive on-site within 48 hours to retest the battery internal resistance, tighten terminals, or replace lines. For faults such as charge / discharge controller failure or complete battery failure (internal resistance rises by more than 50% or terminal voltage cannot be maintained), the system classifies them as high priority emergency faults and immediately generates an emergency alarm command. This command triggers a remote relay via the local control bus to cut off power to the load, preventing damage from deep battery discharge or fire caused by controller short circuits. Simultaneously, it sends a high-priority alarm packet to the maintenance platform via the wireless communication network. This packet includes the device's unique code, geographic coordinates, fault type, time of occurrence, and suggested handling, and automatically initiates the emergency repair work order process. The decision-making alarm module also integrates geographic information association; all alarm information is bound to latitude and longitude coordinates obtained from the device's built-in global navigation satellite system module, with an accuracy of no less than 5 meters, enabling precise location of faulty equipment. Locally, the system triggers an audio-visual alert device, which consists of a high-brightness LED array and a piezoelectric buzzer. The alert mode is set differently according to the fault level: low priority uses a green light that flashes once every 10 seconds and a low-frequency buzzer sound (500 Hz, lasting 0.2 seconds); medium priority uses a yellow light that flashes once every 2 seconds and a medium-frequency buzzer sound (1000 Hz, lasting 0.3 seconds); high priority uses a red light that flashes twice per second and a high-frequency rapid buzzer sound (2000 Hz, lasting 0.5 seconds, with 0.5-second intervals), which allows on-site inspectors to quickly identify the severity of the fault from a distance.

[0027] The system includes a data backtracking and model optimization submodule to achieve closed-loop self-learning and continuous improvement of diagnostic capabilities. The submodule periodically (e.g., weekly) packages complete operational data from the past 7 days into a training sample set. Each sample contains: environmental parameter sequences within the time window, electrical parameter sequences, primary features output by the status assessment module, preliminary and final judgment results from the fault identification module, alarm commands generated by the decision alarm module, and most importantly, a manual confirmation label. The manual confirmation label is entered by on-site maintenance personnel via mobile terminal after maintenance is completed, indicating the actual fault cause (e.g., "bird droppings covering photovoltaic panels," "sulfation of battery plates," "MOSFET breakdown in controller"), serving as the ground truth for supervised learning. The sample set is uploaded to a cloud-based analysis platform, where machine learning algorithms are used to fine-tune the judgment thresholds, weight coefficients, and logical rules in the fault identification module. For example, grid search and cross-validation methods are used to optimize the combination of judgment thresholds for the mean and standard deviation of power deviation, minimizing the sum of false alarm and false negative rates in a specific climate zone. The updated diagnostic model parameters are then transmitted back to each on-site device via a secure channel, completing the remote upgrade of the diagnostic logic. This mechanism enables the system to adapt to equipment aging, environmental changes, and the emergence of new failure modes, maintaining the stability and advancement of long-term diagnostic performance.

[0028] The dual-channel redundant acquisition architecture of the electrical parameter monitoring module is a key technology for ensuring system reliability. The two independent acquisition links are completely symmetrical in hardware design, including signal input terminals, protection circuits, amplifiers, analog-to-digital converters, and microcontroller interfaces. During the startup phase, the system performs channel consistency calibration, using a precision DC source to output standard voltage and current signals, recording the reading deviation between the two channels and generating a correction coefficient matrix. During normal operation, the main control unit reads the dual-channel data every 5 minutes and calculates the relative deviation rate. If the deviation rate of a certain parameter exceeds a preset threshold of 2% for three consecutive times, the system determines that the parameter channel is abnormal. At this time, the main control unit automatically switches to the channel with the smaller deviation as the current valid data source, while marking the other channel as "failed" and adding a data quality flag to the locally stored and reported data. If both channels fail, the system enters a safety mode, suspends fault diagnosis based on that parameter, relies solely on other normal parameters for status assessment, and immediately reports a critical alarm for "critical sensor dual-channel failure." This redundancy mechanism effectively prevents erroneous data acquisition caused by aging of individual sensors, loose wiring, or electromagnetic interference, significantly improving the overall fault tolerance and diagnostic reliability of the system.

[0029] The precipitation sensor in the environmental sensing module achieves its self-cleaning function through ultrasonic vibration. The sensor body is made of stainless steel, with a piezoelectric ceramic transducer integrated at the bottom. The control circuit applies a 40 kHz AC voltage to the transducer at a preset cycle (every 6 hours), exciting the sensor surface to generate high-frequency micro-vibrations. The amplitude is controlled within 10 micrometers, sufficient to remove attached water droplets, dust, or insect remains from the surface without affecting the mechanical balance of the tipping bucket. The vibration process lasts for 30 seconds, during which precipitation measurement is paused to avoid false triggering. The wind speed sensor uses ultrasonic phase difference wind measurement technology, with three sets of ultrasonic transceivers arranged in a ring at a 120-degree angle. The system measures the time difference Δt between the propagation of ultrasonic waves in the downwind and upwind directions. The wind speed v is calculated using the formula v=(L×Δt) / (2×d×cosθ), where L is the transceiver spacing, d is the projected length of the sound wave path in the wind direction, and θ is the angle between the sound wave path and the wind direction. Since there are no rotating parts, the sensor is completely maintenance-free and can operate stably for extended periods in harsh environments such as sandstorms and salt spray.

[0030] The dust accumulation and decay model in the condition assessment module is a key component in improving the accuracy of expected power calculations. The model uses the equipment's last cleaning time, t_clean, as the initial reference point, and then updates the transmittance daily based on meteorological conditions. If the daily precipitation is greater than 2 mm, the system considers it an effective cleaning, and the transmittance recovers to the level of the day after the last cleaning, i.e., η_dust recovers to exp(-k_decay×1 / v_wind_ref). If there is no effective precipitation, the transmittance decays exponentially daily, with the decay coefficient k_decay adjusted by the daily average wind speed v_wind, expressed as k_decay_adj=k_decay×(1-0.3×(v_wind / v_max)), where v_max is the local historical maximum wind speed, ensuring that higher wind speeds result in slower dust settling and a smaller decay coefficient. This model uses historical cleaning records and concurrent power data for parameter fitting to ensure good predictive accuracy across different climatic regions.

[0031] The continuous verification mechanism of the fault identification module effectively solves the problem of confusion between environmental disturbances and real faults. Taking cloud cover as an example, the power drop caused by it is usually short-lived (a few minutes to tens of minutes) and has a high standard deviation (due to rapid changes in sunlight), which may be falsely reported as "low output" in the initial identification. However, since such events occur infrequently and discontinuously within a 12-hour window, the continuous verification layer filters them out and does not confirm them as persistent faults. Conversely, if the photovoltaic panel is covered by leaves for a long period of time, the average power deviation will remain below the threshold, and if multiple daily identifications match this pattern, it will be confirmed as a real fault after more than 8 occurrences. This mechanism enables the system to maintain high sensitivity while possessing strong anti-interference capabilities.

[0032] The differentiated alarm strategy of the decision-making alarm module enables optimal allocation of operation and maintenance resources. Low-priority alerts avoid unnecessary emergency deployments, medium-priority alarms provide reasonable response time, and high-priority emergency alarms ensure that critical faults are handled immediately. The coded design of the audible and visual prompts allows on-site personnel to make a preliminary judgment on the nature of the fault without approaching the equipment, improving operational safety and efficiency.

[0033] The closed-loop learning mechanism of the data backtracking and model optimization submodules endows the system with evolutionary capabilities. By continuously accumulating real-world fault cases, the system can identify new fault modes not covered by existing rules, such as the soft fault characteristics of a specific controller model under high temperatures, thereby updating the judgment logic and improving the coverage and accuracy of future diagnoses.

[0034] This embodiment constructs a complete fault diagnosis technology system for solar power supply in transportation equipment through the collaborative work of the aforementioned modules. The system not only achieves accurate identification of various typical faults, but also solves the fundamental flaw of high false alarm rates in traditional threshold alarm methods through continuous verification and differentiated alarm mechanisms. Deep fusion and dynamic modeling of environmental and electrical multi-source data ensure the scientific rigor and timeliness of the diagnostic benchmark. Internal resistance pulse detection technology provides direct quantitative evidence for battery health status assessment. Dual-channel redundant acquisition and self-cleaning sensor design improve the long-term reliability of the system in harsh outdoor environments. A closed-loop self-learning mechanism ensures continuous optimization of diagnostic performance. The overall solution significantly enhances the autonomous health management capability of transportation equipment energy systems, reduces operation and maintenance costs, and ensures the stable operation of intelligent transportation infrastructure.

[0035] In existing technologies, fault diagnosis of solar power systems often relies on simple low voltage alarms or current interruption detection, lacking a detailed breakdown of fault types and an assessment of severity. Such systems cannot distinguish between temporary shading and permanent damage to photovoltaic panels, nor can they quantify the aging of batteries, leading to a lack of basis for operation and maintenance decisions. This invention introduces dynamic expected power modeling, incorporating environmental factors into the diagnostic benchmark, ensuring that power deviation calculations are based on scientific predictions. Multi-level discrimination logic combined with internal resistance detection enables refined classification of fault types. A continuous verification mechanism filters noise from the time dimension, ensuring alarm reliability. Differentiated alarms and geographic binding functions connect the entire chain from diagnosis to handling. The synergistic effect of these technical features fundamentally changes the traditional "passive response" maintenance model, constructing a new operation and maintenance system of "proactive early warning, precise location, and tiered response."

[0036] This embodiment details the calculation process for the expected output power of the condition assessment module. The core objective is to provide a high-precision power prediction method that integrates astronomical, meteorological, and material characteristics, offering a reliable benchmark for fault diagnosis. The calculation process strictly follows a four-level model: solar geometry, illumination correction, temperature compensation, and dust attenuation. The solar altitude and azimuth angles are calculated using the internationally recognized Spencer formula, with Julian day, geographical latitude and longitude, and local time as inputs. The output angle accuracy meets engineering requirements. Illumination intensity correction considers atmospheric transparency, with coefficients obtained from tables based on air quality index and visibility, ensuring a reasonable reduction in theoretical power under hazy conditions. Temperature compensation uses the power temperature coefficient provided by the photovoltaic module manufacturer, combined with a surface temperature estimation model corrected for wind speed, improving compensation accuracy. The dust attenuation model incorporates the dual effects of precipitation erosion and wind speed-inhibited sedimentation, making the transmittance prediction closer to reality. The four-level models are executed in series, and the final expected power output serves as a reference benchmark for fault diagnosis.

[0037] This embodiment specifies the method for obtaining the internal resistance value of the fault identification module. The core objective is to provide a non-invasive, periodically executable method for quantifying battery health. Pulse current injection is performed at the initial stage of charging, when the battery is in a constant current phase, the system is stable, and the influence of external loads is minimal. The pulse amplitude is set to 1.5 times the rated charging current, lasting 30 seconds, which generates measurable voltage changes without damaging the battery. Voltage sampling is performed one second before and after the pulse, and the average value of multiple points is taken to eliminate noise. The AC internal resistance calculation formula R_ac=ΔU / I_pulse is a fundamental relationship in electrochemical impedance spectroscopy, exhibiting good linearity and repeatability in engineering applications. Daily sampling generates a trend curve; by calculating the slope or setting a threshold, the degradation rate can be determined, providing data support for predictive maintenance.

[0038] In this embodiment, the sliding window parameters for the power deviation sequence are specified. The setting of a 6-hour sliding window length and a 1-hour step size balances the stability and timeliness of feature extraction. The 6-hour window is sufficient to cover typical weather change cycles, such as a morning transition from cloudy to sunny, ensuring the representativeness of statistical features; the 1-hour step size ensures timely updates to the state assessment. The adaptive standard deviation adjustment mechanism reflects the system's adaptability to regional and seasonal differences. Due to large fluctuations in winter sunlight, the threshold is relaxed to 1.3 times that of summer to avoid false alarms; different regions are assigned regional factors based on historical data, such as a higher threshold for foggy coastal areas than for arid inland areas, achieving localized optimization of the diagnostic strategy.

[0039] In this embodiment, the alarm issuance and notification methods of the decision-making alarm module are specified. The wireless communication network adopts the 4G LTE standard and supports the TCP / IP protocol stack. The alarm command is encapsulated as a JSON format data packet, containing the device ID, fault type code, severity level, timestamp, and location information. The control signal for the local audio-visual notification is directly driven by the microcontroller's general-purpose input / output port. The LED flashing frequency and buzzer tone are precisely controlled by a pulse width modulation signal to ensure the standardization and recognizability of the notification mode.

[0040] In this embodiment, the operational mechanism of the data backtracking and model optimization submodule is specified. The sample packaging cycle is set to weekly to avoid excessive data volume affecting transmission efficiency. Manual confirmation of tag entry is completed through a dedicated mobile application, including the fault location, phenomenon description, handling measures, and replacement part model. The cloud-based analysis platform uses random forest or gradient boosting tree algorithms to train the classification model, using traditional rule-based criteria as initial weights, and learns and optimizes the contribution of each feature through historical samples. Model updates adopt a canary release strategy, first testing on 10% of devices to verify the effect before full rollout, ensuring system stability.

[0041] This embodiment, through a comprehensive explanation of the aforementioned technical details, fully reveals the implementation path and engineering solution of a solar power supply fault diagnosis system for transportation equipment. The functions of each module are clearly defined, data flow is orderly, logical judgment is rigorous, and control actions are explicit, forming a complete, reliable, and intelligent fault diagnosis technology system.

Claims

1. A fault diagnosis system for solar power supply of transportation equipment, characterized in that, include: The environmental sensing module is used to collect real-time data on light intensity, ambient temperature, precipitation, and wind speed in the deployment area. The electrical parameter monitoring module is used to continuously monitor the output voltage and current of the solar photovoltaic panel, the terminal voltage of the battery, the charging and discharging current, the internal resistance value, and the working status signal of the charging and discharging controller. The state assessment module receives raw data from the environmental sensing module and the electrical parameter monitoring module, dynamically models the theoretical power generation capacity of the photovoltaic panel, and calculates the expected output power of the photovoltaic panel based on historical data and current environmental parameters. The state assessment module compares the actual output power collected by the electrical parameter monitoring module with the expected output power in time intervals to generate a power deviation sequence, and performs time-series sliding window statistical analysis on the power deviation sequence to extract the mean, standard deviation and slope of the deviation as primary state features. The fault identification module is used to perform multi-level fault logic judgment based on primary state characteristics. When the average power deviation is continuously lower than the first preset threshold and the standard deviation is less than the first set range, it is determined that there is continuous shading or pollution on the surface of the photovoltaic panel. When the average power deviation is within the normal range but the standard deviation exceeds the second preset threshold, it is determined that there is poor line contact or intermittent open circuit. When the battery terminal voltage cannot reach the preset float charge voltage during the charging stage and the internal resistance continues to rise, and the charging and discharging current waveform shows non-periodic distortion, it is determined that the battery is aging or has failed. When the operating status code reported by the charging and discharging controller contains a preset fault code and the abnormality is independent of the photovoltaic input and the battery status, it is determined that the controller itself is faulty. The decision alarm module is used to receive the fault confirmation results output by the fault identification module and generate differentiated alarm commands based on the fault type and severity level. The internal resistance value in the fault identification module is obtained as follows: within the first 5 minutes of each battery charging start, when the system detects that it is in the constant current charging stage and the current is stable, the charge and discharge controller is controlled to inject a pulse current with a duration of 30 seconds and an amplitude of 1.5 times the rated charging current. The battery terminal voltage is collected synchronously within 1 second before and after the pulse, and the voltage change ΔU is calculated. The ratio of ΔU to the pulse current amplitude is used as the AC internal resistance value at that moment. The AC internal resistance value is sampled once a day to form a time series for trend analysis. Differentiated alarm commands in the decision alarm module are sent to the operation and maintenance management terminal via wireless communication network; the audible and visual prompt device is triggered locally on the traffic equipment, which consists of a high-brightness LED array and a piezoelectric buzzer; the prompt mode is set with different combinations according to the fault level: low priority uses a green light that flashes once every 10 seconds and a 500 Hz low-frequency buzzer sound; Medium priority uses a yellow light that flashes once every 2 seconds and a 1000 Hz medium frequency buzzer; high priority uses a red light that flashes twice per second and a 2000 Hz high frequency rapid buzzer, which makes it easy for on-site personnel to quickly identify the severity of the fault from a distance. The electrical parameter monitoring module adopts a dual-channel redundant acquisition architecture, implementing independent dual-channel acquisition and cross-verification for photovoltaic panel output voltage, photovoltaic panel output current, battery terminal voltage, and charging and discharging current. The two acquisition channels for each type of signal are powered by independent signal conditioning circuits, analog-to-digital converters, and isolated power supplies, achieving physical isolation at the hardware level. The system compares the dual-channel readings every 5 minutes. If the relative deviation continues to exceed 2%, the self-diagnosis program is initiated and the channel with stable readings is switched as the primary channel.

2. The traffic equipment solar power supply fault diagnosis system according to claim 1, characterized in that, The expected output power calculation process in the status assessment module is as follows: Based on the geographical latitude and longitude of the equipment installation location and the current date and time, the solar altitude angle and azimuth angle are calculated; combined with the real-time light intensity and atmospheric transparency coefficient provided by the environmental perception module, the rated power of the photovoltaic panel under standard test conditions is linearly corrected; a temperature compensation factor is introduced to estimate the actual operating temperature based on the impact of ambient temperature and wind speed on the heat dissipation of the photovoltaic panel surface, and the output power is corrected with a negative temperature coefficient; a dust accumulation attenuation model is superimposed, which uses the equipment's last cleaning time as a benchmark and combines historical precipitation and average wind speed data to dynamically estimate the light transmittance loss of the photovoltaic panel surface, and finally outputs the expected output power.

3. The solar power supply fault diagnosis system for a traffic device according to claim 2, wherein The dust accumulation attenuation model is implemented as follows: when the daily precipitation is greater than 2 mm, the system considers it to be an effective cleaning, and the light transmittance is restored to the level of the first day after the last cleaning; during the period without effective precipitation, the light transmittance decays with time according to an exponential law, and the decay rate is adjusted by the daily average wind speed. The higher the wind speed, the slower the decay. The current light transmittance is calculated by empirical formula and used to correct the expected output power.

4. The traffic equipment solar power supply fault diagnosis system according to claim 1, characterized in that, The power deviation sequence in the condition assessment module is statistically analyzed using a time-series sliding window method. The sliding window length is set to 6 hours, and the step size is 1 hour. After each sliding, the mean, standard deviation, and slope of the deviation within the window are recalculated. The slope of the change is obtained by fitting the time series trend line of the power deviation data within the window using the least squares method, which is used to capture signs of acceleration or mitigation of performance degradation.

5. A fault diagnosis system for solar power supply of transportation equipment according to claim 4, characterized in that, The standard deviation threshold is adaptively adjusted according to the season and geographical region. In the temperate region of the Northern Hemisphere, the standard deviation threshold in winter is set to 1.3 times that in summer. The regional adjustment factor is calculated based on historical meteorological data of the equipment deployment location and is pre-written into the system configuration table to adapt to the natural differences in light fluctuations under different climatic conditions.

6. The solar power supply fault diagnosis system for a traffic device according to claim 5, wherein During the startup phase, the electrical parameter monitoring module performs channel consistency calibration by using a precision DC source to output a standard signal, recording the reading deviation between the two channels and generating a correction coefficient matrix.

7. The traffic equipment solar power supply fault diagnosis system according to claim 1, characterized in that, It also includes a data backtracking and model optimization sub-module, which periodically packages the complete operating data from the past 7 days into a training sample set; each sample contains an environmental parameter sequence, an electrical parameter sequence, primary state characteristics, preliminary and final fault judgment results, generated alarm commands, and manually confirmed actual fault cause labels; The sample set is uploaded to the cloud analysis platform, and machine learning algorithms are used to optimize the parameters of the discrimination threshold and logic rules in the fault identification module. The updated diagnostic model parameters are then transmitted back to the field equipment through a secure channel to complete the remote upgrade.

Images