A method and device for self-detecting faults in lighting system

Abstract

The present invention relates to the field of fault self-detection, and discloses a fault self-detection method and device for a lighting system, which are used to improve the accuracy and efficiency of fault detection and enhance the early warning and prediction capabilities of the system. The present invention realizes comprehensive and accurate fault detection of the lighting system by combining multiple sensors, integrated learning models, time series models and Bayesian network models. At the same time, the method also has fault early warning and prediction capabilities, and can take timely measures before the fault occurs to avoid losses. In addition, detailed fault reports and repair suggestions are provided to facilitate users to troubleshoot and repair faults. At the same time, by introducing user feedback and interactive mechanisms, the performance of the system is continuously optimized and improved.

Description

Technical Field

[0001] The present invention relates to the field of fault self-detection, and in particular to a fault self-detection method and device for a lighting system. Background Art

[0002] In modern society, lighting systems in harsh environments play a vital role. They not only ensure the normal operation of various work activities, but also directly affect personnel safety and work efficiency. However, due to the complexity and unpredictability of harsh environments, lighting systems often face various challenges, such as extreme climate, unstable power supply, aging equipment, etc. These factors may lead to system failures.

[0003] Deficiencies in the prior art:

[0004] Lack of intelligent fault detection methods: Traditional lighting system fault detection mainly relies on manual inspection and regular maintenance, which is not only inefficient but also difficult to detect and handle faults in real time. Especially in harsh environments, the difficulty and danger of manual inspections are greatly increased.

[0005] Low fault detection accuracy: Existing fault detection methods are often based on a single parameter or simple model, which makes it difficult to fully and accurately reflect the actual operating status of the lighting system. Especially in complex and changing harsh environments, the fluctuation of a single parameter may be affected by multiple factors, resulting in false alarms or missed alarms.

[0006] Lack of fault warning and prediction capabilities: Traditional fault detection methods can usually only detect and repair faults after they occur, and cannot warn and predict the occurrence of faults in advance. This results in the system being unable to work properly when a fault occurs, causing unnecessary losses.

[0007] Unclear fault information dissemination: After a fault occurs, the existing system can only provide simple fault information, such as fault location, fault type, etc., but cannot provide detailed fault causes, propagation paths and repair suggestions, which brings great difficulties to fault troubleshooting and repair.

[0008] Insufficient user feedback and interaction: Existing lighting system fault detection methods often lack user feedback and interaction mechanisms. Users are unable to verify and provide feedback on the detection results, nor can they make improvement suggestions, which limits the further optimization and upgrading of the system.

[0009] In view of the above shortcomings, this study proposes a fault self-detection method for lighting systems. Summary of the invention

[0010] The present invention provides a fault self-detection method and device for a lighting system, which are used to improve the accuracy and efficiency of fault detection and enhance the early warning and prediction capabilities of the system.

[0011] A first aspect of the present invention provides a fault self-detection method for a lighting system, the fault self-detection method for a lighting system comprising: receiving a fault self-detection signal, the fault self-detection signal including a voltage fluctuation exceeding a set threshold, switching to a secondary mode, and in response to the switching, obtaining existing resource data parameters; obtaining prediction results of multiple base learners based on the existing resource data parameters in combination with a preset integrated learning model, integrating the prediction results of the multiple base learners to obtain a first-level detection result, preliminarily judging whether there is a fault based on the first-level detection result, if the first-level detection result judges that there is no fault, switching to a recovery mode and recording the first-level detection result, if the first-level detection result judges that there is a fault, switching to an off mode; in response to the conversion, obtaining secondary operating status data, and According to the secondary operation status data combined with a preset time series model, a model prediction value is obtained, and the difference between the actual parameter value of the secondary operation status data and the model prediction value is calculated to obtain a warning numerical result; according to the secondary operation status data combined with a preset Bayesian network model of the lighting system, possible results of the fault cause and possible results of the propagation path are obtained, and a secondary detection result is obtained according to the warning numerical result, the possible results of the fault cause and the possible results of the propagation path; a fault self-check result is obtained according to the primary detection result and the secondary detection result, including: using preset historical fault data and normal operation data, assigning weights to the primary detection result and the secondary detection result, assuming that the primary detection result is R1 and the secondary detection result is R2, then the weight assignment formula is: Wherein, NF1 is the number of fault data in the first-level detection result, and NF2 is the number of fault data in the second-level detection result. The first-level detection result and the second-level detection result are fused by using the decision fusion and evidence theory method to obtain the fusion output result. For each detection result, a basic probability distribution function is defined to indicate the support degree of the result for a certain fault type. Let Θ be the set of all possible fault types. For each : Among them, R1(θ) and R2(θ) are the scores or probabilities of the first-level detection results and the second-level detection results for the fault type θ respectively; Dempster's synthesis rule is used to fuse the basic probability distribution of the two detection results. Let m 12 is the basic probability distribution after fusion, then: Among them, K is the conflict coefficient, which indicates the inconsistency between the two pieces of evidence. The fusion output result is mapped to the specific fault type and level through a mapping table or classifier to obtain the fault self-check result.

[0012] Optionally, in a first implementation of the first aspect of the present invention, the fault self-check signal is received, switched to the secondary mode, and in response to the switch, existing resource data parameters are obtained, including: deploying multiple sensors to form a sensor network, continuously collecting a single energy consumption resource parameter of the lighting system, establishing a dynamic sliding window for storing and analyzing the energy consumption data of the past N times, setting a window size N, representing the number of energy consumption data points stored in the sliding window, and calculating the average value of the energy consumption data in the sliding window in real time whenever new energy consumption data arrives, as the energy consumption benchmark value; according to the impact of environmental factors on lighting energy consumption, setting weighting coefficients for different environmental factors; obtaining the difference between the average values ​​based on the weighting coefficients and the energy consumption benchmark value, and the difference exceeding a preset threshold value M, a fault self-check signal is triggered; upon receiving the fault self-check signal, switching to the secondary mode, the fault self-check signal includes energy consumption abnormality, the secondary mode includes an energy-saving mode, in which the lighting system will reduce the brightness, adjust the color temperature or turn off some lamps, the switching of the secondary mode is a smooth switching mechanism, and the brightness and color temperature parameters are adjusted in an exponential gradient manner, and no sudden visual impact is caused to the user when switching to the secondary mode; and / or the lamps in some non-critical areas are dynamically turned off when switching to the secondary mode; in response to the switching, existing resource data parameters are obtained according to the sensor network, the existing resource data parameters include current, voltage, and temperature, and the existing resource data parameters are cleaned, denoised, and standardized.

[0013] Optionally, in a second implementation of the first aspect of the present invention, the prediction results of multiple base learners are obtained based on the existing resource data parameters in combination with a preset integrated learning model, the prediction results of the multiple base learners are integrated to obtain a first-level detection result, and a preliminarily judgment is made based on the first-level detection result whether there is a fault; if the first-level detection result judges that there is no fault, the recovery mode is switched to and the first-level detection result is recorded; if the first-level detection result judges that there is a fault, the shutdown mode is switched to, including: selecting multiple base learners, the base learners include decision trees, support vector machines, and neural networks, using preset historical fault data and normal operating data to train the base learners to obtain a trained integrated learning model type; according to the existing resource data parameters, they are input into the trained ensemble learning model, each base learner gives its own prediction result, and a weighted voting, stacking or bagging ensemble strategy is used to combine the prediction results of multiple base learners to obtain a first-level detection result; a threshold L is preset, if the first-level detection result exceeds the threshold L, it is judged that there is a fault, otherwise, it is judged that there is no fault, if it is judged that there is no fault, it is switched to the recovery mode, the recovery mode includes restarting the system, adjusting parameters, and recording the first-level detection result, if it is judged that there is a fault, it is switched to the shutdown mode, the shutdown mode includes cutting off the power supply, issuing an alarm, and recording the fault information; if within a unit time r, the number of times in response to the switching is greater than n, it is directly switched to the shutdown mode.

[0014] Optionally, in a third implementation of the first aspect of the present invention, in response to the conversion, secondary operating status data is obtained, a model prediction value is obtained based on the secondary operating status data in combination with a preset time series model, and the difference between the actual parameter value of the secondary operating status data and the model prediction value is calculated to obtain an early warning numerical result, including: once the system determines that a fault exists and enters a shutdown mode, a data collection mechanism is immediately triggered, and the data collection mechanism is designed to quickly capture and store the secondary operating status data of the lighting system within a short period of time after the fault occurs. The secondary operating status data includes user feedback fault reports, historical fault logs, remote access logs, system resource usage, and environmental and situational data; using preset historical fault data and normal operation The data is trained to obtain a trained time series model, which includes a long short-term memory network, a gated recurrent unit and a Transformer model, to obtain an integrated time series model; the preprocessed secondary operation status data is input into the integrated time series model, the time series model outputs the predicted state data value in a future period of time, and the difference between the actually measured secondary operation status data and the model predicted value is calculated; a variety of indicators are used to measure the difference to obtain a warning numerical result, including mean square error (MSE), mean absolute error (MAE), and dynamic time warping (DTW) distance, and a difference threshold U is set. When the difference between the actual data and the predicted data exceeds this threshold, a warning signal is generated; the warning numerical result includes the difference size and the warning level.

[0015] Optionally, in a fourth implementation manner of the first aspect of the present invention, the possible results of the fault cause and the possible results of the propagation path are obtained based on the secondary operating status data in combination with a preset lighting system Bayesian network model, and the secondary detection result is obtained based on the warning numerical results, the possible results of the fault cause and the possible results of the propagation path, including: inputting the secondary operating status data into a trained lighting system Bayesian network model, using the inference algorithm in the lighting system Bayesian network model to calculate the posterior probability of each fault cause node, and determining the most likely fault cause and its combination according to the size of the posterior probability to obtain the possible results of the fault cause, the inference algorithm including a clique tree algorithm and a variable elimination method; in the Bayesian network, analyzing the dependency relationship between the fault cause node and the fault phenomenon node, and using a path search algorithm to find the possible results of the propagation path from the fault cause to the fault phenomenon, the path search algorithm including a depth-first search and a breadth-first search; combining the warning numerical results, the possible results of the fault cause and the possible results of the propagation path, grading and classifying the warning numerical results, and obtaining the secondary detection result.

[0016] Optionally, in the fifth implementation method of the first aspect of the present invention, it also includes: using visualization technology to display the fault self-inspection results, and generating a detailed fault report, including the fault type, level, possible cause, propagation path and recommended repair measures; allowing users to provide feedback on the fault self-inspection results, including the accuracy and completeness of the results and recommended improvement directions; real-time monitoring of the operating status of the lighting system through the management platform, and promptly triggering the early warning mechanism when a potential fault is detected, and notifying relevant personnel to handle it.

[0017] The second aspect of the present invention provides a fault self-detection device for a lighting system, the fault self-detection device for the lighting system comprising: an acquisition module, for receiving a fault self-detection signal, switching to a secondary mode, and acquiring existing resource data parameters in response to the switching; a detection module, for obtaining prediction results of multiple base learners based on the existing resource data parameters in combination with a preset integrated learning model, integrating the prediction results of the multiple base learners to obtain a first-level detection result, and preliminarily judging whether there is a fault based on the first-level detection result. If the first-level detection result determines that there is no fault, switching to a recovery mode and recording the first-level detection result. If the first-level detection result determines that there is a fault, switching to a shutdown mode. closed mode; a prediction module, used to respond to the conversion, obtain the secondary operation status data, obtain the model prediction value according to the secondary operation status data in combination with a preset time series model, calculate the difference between the actual parameter value of the secondary operation status data and the model prediction value, and obtain the early warning numerical result; a setting module, used to obtain the possible result of the fault cause and the possible result of the propagation path according to the secondary operation status data in combination with a preset Bayesian network model of the lighting system, and obtain the secondary detection result according to the early warning numerical result, the possible result of the fault cause and the possible result of the propagation path; an allocation module, used to obtain the fault self-detection result according to the primary detection result and the secondary detection result.

[0018] The third aspect of the present invention provides a fault self-detection device for a lighting system, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the fault self-detection device for the lighting system executes the above-mentioned fault self-detection method for the lighting system.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned fault self-detection method for a lighting system.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] Improve fault detection accuracy: Through the comprehensive application of ensemble learning models and time series models, it is possible to more accurately detect and analyze lighting system faults and reduce false positives and missed negatives.

[0022] Enhance system reliability: Real-time monitoring and fault self-detection mechanisms can detect potential faults in a timely manner and take measures to prevent the faults from expanding, thereby improving the reliability and stability of the lighting system.

[0023] Optimize resource utilization: In secondary mode, the system is able to reduce energy consumption and adjust parameters such as brightness and color temperature to save resources and extend the service life of the lighting system.

[0024] Improve user experience: Smooth mode switching mechanism and detailed fault reporting can reduce users' troubleshooting concerns and improve user satisfaction and experience.

[0025] Easy maintenance and management: Through the real-time monitoring and fault warning mechanism of the management platform, maintenance personnel can promptly understand the system status and take measures to reduce maintenance costs and workload.

[0026] In summary, the present invention provides an efficient and accurate fault self-detection method for harsh environment lighting systems, which can significantly improve system reliability, optimize resource utilization and enhance user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A schematic diagram of an embodiment of a fault self-detection method for a lighting system according to an embodiment of the present invention;

[0028] Figure 2 A schematic diagram of another embodiment of a fault self-detection method for a lighting system according to an embodiment of the present invention;

[0029] Figure 3 A schematic diagram of an embodiment of a fault self-detection device for a lighting system according to an embodiment of the present invention;

[0030] Figure 4 Schematic diagram of an embodiment of a fault self-detection device for a lighting system in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The embodiments of the present invention provide a method and device for self-checking faults of a lighting system, which are used to improve the accuracy and efficiency of fault detection and enhance the early warning and prediction capabilities of the system. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein.

[0032] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of a fault self-detection method of a lighting system in an embodiment of the present invention includes:

[0033] 101. Receiving a fault self-checking signal, switching to a secondary mode, and acquiring existing resource data parameters in response to the switching;

[0034] It is understandable that the execution subject of the present invention may be a fault self-detection device of the lighting system, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0035] It should be noted that, assuming that an industrial facility is located in a harsh environment with strong winds and dust, its lighting system consists of a main lighting system powered by a main power source and a secondary lighting system powered by a backup battery. The main lighting system includes LED lamps and an intelligent control system, while the secondary lighting system includes fluorescent lamps and manual switch control.

[0036] System monitoring: The central control unit (CCU) of the lighting system periodically performs self-checks on the main power supply, lamp status, communication lines, etc. Fault detection: During a self-check, the CCU detects that the voltage fluctuation of the main power supply line is abnormal, exceeds the set threshold (such as ±10%), and lasts for more than 3 seconds, and determines that there may be a fault in the main power supply. Signal transmission: The CCU immediately sends a fault self-check signal to the secondary lighting system and prepares to switch to the secondary mode.

[0037] Switch to secondary mode: Mode switch command: The CCU generates a mode switch command through internal logic processing and sends it to the controller of the secondary lighting system through a communication line (such as an RS-485 bus). Secondary system response: After receiving the switch command, the controller of the secondary system automatically starts the backup power supply (assuming it is a lead-acid battery pack with a rated voltage of 12V and a capacity of 100Ah) and turns on all fluorescent lamps. Status confirmation: The secondary system controller feeds back a switch completion signal to the CCU and reports the current working status of the secondary system, including battery voltage, number of lamps turned on, etc.

[0038] In response to the switching, obtain the existing resource data parameters: Data parameter acquisition: After receiving the switching completion signal fed back by the secondary system, the CCU immediately obtains the existing resource data parameters through the interface. Secondary power supply parameters: battery voltage (11.8V, indicating that the battery is discharging but in good condition), remaining capacity estimation (assuming 80%, i.e. 80Ah available). Lamp status: All fluorescent lamps are turned on normally, and no fault report is reported. Ambient light intensity: The current ambient light intensity data is obtained through the ambient light sensor (such as 50lux, indicating that the lighting needs to be turned on).

[0039] 102. According to the existing resource data parameters combined with a preset integrated learning model, prediction results of multiple base learners are obtained, and the prediction results of the multiple base learners are integrated to obtain a first-level detection result. According to the first-level detection result, it is preliminarily determined whether there is a fault. If the first-level detection result determines that there is no fault, the recovery mode is switched and the first-level detection result is recorded. If the first-level detection result determines that there is a fault, the shutdown mode is switched.

[0040] It should be noted that the ensemble learning model is constructed as follows: Data collection and preprocessing: Collect historical fault data, including voltage fluctuations, current anomalies, lamp brightness changes, ambient temperature, humidity and other parameters. Clean the data, remove outliers and missing values, and then normalize them. Base learner selection: Select three different base learners: support vector machine (SVM), decision tree (DT) and random forest (RF). Each base learner is trained using the same training data set, but with different algorithms and parameter settings.

[0041] Ensemble learning model training: Use bagging (self-service sampling method) to generate training subsets for each base learner. Train each base learner and obtain its prediction results. Use majority voting as the ensemble strategy, that is, determine the final fault judgment based on the prediction results of the majority of base learners.

[0042] Obtain existing resource data parameters: obtain real-time data parameters such as lighting system voltage (220V±5%), current (0.5A±0.1A), lamp brightness (90%±5%), ambient temperature (25℃±5℃), humidity (60%±10%), etc.

[0043] Base learner prediction: Input the acquired data parameters into the trained SVM, DT and RF base learners. SVM prediction result: No fault (probability is 0.85). DT prediction result: Fault exists (probability is 0.6). RF prediction result: No fault (probability is 0.75).

[0044] Ensemble prediction results: According to the voting method, two base learners (SVM and RF) predict no fault, and one base learner (DT) predicts a fault. Therefore, the first-level detection result is no fault.

[0045] Initial judgment and mode conversion: Since the first-level test results show no fault, the system switches to recovery mode. In recovery mode, the system will try to restart or reset the unstable components and record the first-level test results for subsequent analysis.

[0046] Recording and feedback: The first-level detection results (including the prediction results of each base learner, the integrated prediction results, and the mode conversion information) are recorded in the fault log. If a similar situation occurs again in the future, the model can be optimized or the strategy can be adjusted by consulting the fault log.

[0047] If there is a fault: If the first-level detection result determines that there is a fault, the system will immediately switch to the shutdown mode to prevent the fault from expanding or causing safety hazards. In the shutdown mode, the system will cut off the power supply, stop all lamps from working, and issue an alarm to notify maintenance personnel to handle it.

[0048] 103. In response to the conversion, secondary operation status data is acquired, a model prediction value is obtained based on the secondary operation status data in combination with a preset time series model, and a difference between an actual parameter value of the secondary operation status data and the model prediction value is calculated to obtain a warning numerical result;

[0049] It should be noted that the time series model is constructed as follows: Data collection and preprocessing: Collect secondary operating status data of the lighting system in the past year, including voltage, current, power factor, lamp brightness, ambient temperature, etc. The data sampling interval is 1 hour. Clean the data, remove outliers and missing values, and then perform stabilization processing and feature extraction.

[0050] Model selection and training: Select the ARIMA (AutoRegressive Integrated Moving Average) model as the time series forecasting model. Use the data from the past year to train the ARIMA model and determine the model parameters (such as order p, d, q) and optimization algorithm. Verify and test the model to ensure its forecasting accuracy and stability.

[0051] Respond to the transition: When the lighting system switches from normal operation mode to shutdown mode (for example, because the primary detection results indicate that there may be a potential fault), the acquisition of the secondary operation status data is triggered. Acquire secondary operation status data: Real-time acquisition of the secondary operation status data of the lighting system, such as voltage (230V), current (0.45A), power factor (0.98), lamp brightness (85%), and ambient temperature (30°C). Combined with the time series model for prediction: The acquired secondary operation status data is input into the trained ARIMA model. The model outputs the predicted values ​​for the next 24 hours, including voltage prediction values, current prediction values, etc.

[0052] Calculate the difference and warning value results: Calculate the difference between the actual parameter value of the secondary operation status data and the model prediction value, for example:

[0053] Voltage difference = actual voltage - predicted voltage = 230V - 228V = 2V

[0054] Current difference = Actual current - Predicted current = 0.45A - 0.46A = -0.01A

[0055] According to the preset threshold value (for example, the voltage difference threshold value is ±5V, and the current difference threshold value is ±0.1A), it is determined whether the difference exceeds the normal range.

[0056] In this example, both the voltage difference and the current difference do not exceed the threshold, so the warning value result is normal.

[0057] Early warning and response: If the early warning value exceeds the threshold, the early warning signal is triggered to notify maintenance personnel to conduct inspection and repair. If the early warning value is within the normal range, the system continues to maintain the early warning mode and continuously monitors the secondary operation status data. Recording and feedback: Record the secondary operation status data, model prediction value, early warning value results and other information in the fault early warning log. Regularly analyze the early warning log, optimize the time series model parameters and threshold settings, and improve the accuracy of early warning.

[0058] 104. According to the secondary operation status data and the preset Bayesian network model of the lighting system, possible results of the fault cause and possible results of the propagation path are obtained, and the secondary detection result is obtained according to the warning value result, the possible results of the fault cause and the possible results of the propagation path;

[0059] It should be noted that the Bayesian network model is constructed as follows: Data collection and preprocessing: Collect fault records of lighting systems in the past few years, including fault causes, fault phenomena, propagation paths, etc. Clean and organize the data to extract characteristic variables related to fault diagnosis. Model structure and parameter learning: Based on the fault records, construct the structure of the Bayesian network and determine the dependencies between nodes (such as abnormal voltage, abnormal current, lamp failure, etc.). Use learning algorithms (such as maximum likelihood estimation or Bayesian estimation) to estimate the parameters of the model, that is, the conditional probability table between nodes.

[0060] Combined with the Bayesian network model for diagnosis: The acquired secondary operation status data is input into the trained Bayesian network model. Based on the input data, the model calculates the possible results of each fault cause and the possible results of the propagation path. The model concludes that the probability of "abnormal voltage causing the lamp to not light up" is 0.8, and the probability of "abnormal current causing controller failure" is 0.6.

[0061] Calculate the warning value result: According to the preset threshold, determine whether the actual parameter value in the secondary operation status data exceeds the normal range. Set the threshold of the battery panel voltage to 18-22V and the threshold of the battery voltage to 11-13V.

[0062] In this example, the voltage of the solar panel is slightly high (20V), but within the normal range; the voltage of the battery is normal (12V); the output current of the controller and the brightness of the LED lamp are also within the normal range; but the ambient temperature is high (45℃), so attention should be paid to heat dissipation.

[0063] Secondary test results: The secondary test results are obtained by combining the warning numerical results, possible fault cause results and possible propagation path results. The warning numerical results show that the current system status is basically normal, but the ambient temperature is high, which requires attention. The possible fault cause results show that the possibility of abnormal voltage causing the lamp to not light up is high (0.8), and it is necessary to focus on checking the working status of the solar panel and battery.

[0064] The possible results of the propagation path show that if the voltage anomaly persists, it will cause the controller to fail (with a probability of 0.6), thus affecting the normal operation of the entire lighting system.

[0065] Response and processing: Take corresponding response measures according to the secondary detection results. Strengthen the monitoring of solar panels and batteries to ensure voltage stability; increase heat dissipation measures to reduce ambient temperature; prepare backup controllers just in case.

[0066] Recording and feedback: Record the secondary detection results, response measures taken, and other information in the fault self-check log. Regularly analyze the log data, optimize the structure and parameters of the Bayesian network model, and improve the accuracy and efficiency of fault diagnosis.

[0067] 105. Obtain a fault self-detection result according to the first-level detection result and the second-level detection result.

[0068] It should be noted that the first-level test results and data examples are as follows:

[0069] Time: 16:00, September 19, 2024

[0070] Test results: Voltage test: The system voltage fluctuates abnormally and is lower than the normal threshold (220V±10%). The measured voltage is 190V.

[0071] Current detection: The main circuit current is basically normal, but there is a brief overcurrent phenomenon in a specific period of time (such as when the wind speed changes suddenly).

[0072] Conclusion: The preliminary judgment is that the power supply is unstable or there is a potential fault in the line.

[0073] Secondary test results, data example:

[0074] Time: Immediately after the first level test

[0075] Detailed detection: Voltage anomaly analysis: Through Bayesian network model analysis, combined with ambient temperature (-20℃) and wind speed (15m / s) data, it is inferred that voltage fluctuations may be caused by increased resistance of the transmission line.

[0076] Current overcurrent analysis: Using the time series model to predict and compare the current data, it was found that the overcurrent phenomenon is highly correlated with sudden changes in wind speed, which may be caused by fluctuations in generator output.

[0077] Lamp status detection: The brightness of some LED lamps is reduced, and the infrared sensor detects that the surface temperature of the lamp is high, which is initially judged to be poor heat dissipation of the lamp.

[0078] Conclusion: The voltage fluctuation may be caused by aging of the transmission line or insufficient design margin. Overcurrent is a normal phenomenon, but attention should be paid to the impact of wind turbine output stability on the system. If the brightness of the lamp is reduced, the cooling system should be checked to prevent overheating and damage.

[0079] Comprehensive fault self-test results, data integration: Comprehensive primary and secondary test results Fault self-test results: Major fault: unstable power supply, specifically manifested as abnormal voltage fluctuation. Minor fault: poor heat dissipation of the lamp, resulting in reduced brightness. Potential risks: Generator output fluctuations may affect the system current and require long-term monitoring.

[0080] Recommended measures: Immediate action: Check and replace aging or improperly designed transmission lines to ensure voltage stability. Short-term plan: Clean and optimize the lamp cooling system to restore the normal brightness of the lamp. Long-term monitoring: Establish a real-time monitoring mechanism for generator output and system current to prevent potential failures.

[0081] In the embodiment of the present invention, the central control unit (CCU) performs self-inspection on the main power supply, lamp status, communication line, etc. at regular intervals. Once an abnormality is found (such as voltage fluctuation exceeding the set threshold), a fault self-inspection signal is immediately triggered. After receiving the fault self-inspection signal, the system switches to the secondary mode and obtains existing resource data parameters, including secondary power supply parameters, lamp status, ambient light intensity, etc. According to the acquired data parameters, combined with the preset integrated learning model (including base learners such as SVM, DT and RF), the prediction results of multiple base learners are obtained, and the first-level detection results are obtained by integration through the voting method to preliminarily judge whether there is a fault. If the first-level detection result determines that there is a fault or further diagnosis is required, the system obtains the second-level operating status data, and combines the preset time series model (such as ARIMA model) and Bayesian network model to obtain the model prediction value, possible results of the fault cause and possible results of the propagation path, and integrates this information to obtain the second-level detection result. Finally, based on the first-level detection result and the second-level detection result, the final fault self-inspection result is obtained, and corresponding recommended measures are proposed. Through the comprehensive use of the integrated learning model and the time series model, it is possible to more accurately detect and diagnose the fault of the lighting system and reduce false alarms and missed alarms. When a fault occurs, the system can quickly switch to the secondary mode to ensure that the lighting is not affected. At the same time, it provides fault warning and diagnostic information, which helps to quickly restore the normal operation of the system. Through the fault self-check results, maintenance personnel can conduct inspections and repairs more targetedly, reduce unnecessary inspections and repair costs, and improve maintenance efficiency. Considering the impact of harsh environments on lighting systems, such as windy, dusty, and low temperatures, it ensures that the fault self-check method has sufficient adaptability and robustness to cope with the fault detection needs under various harsh environmental conditions. The implementation of this technology helps to promote the intelligent upgrade of lighting systems and provide strong technical support and guarantee for future intelligent lighting systems. In summary, this technology provides an efficient, accurate, and stable solution for fault self-checking of lighting systems, which is of great significance for improving system reliability, optimizing maintenance efficiency, and promoting intelligent upgrades.

[0082] See also Figure 2 Another embodiment of the fault self-detection method of the lighting system in the embodiment of the present invention includes:

[0083] 201. Receiving a fault self-checking signal, switching to a secondary mode, and acquiring existing resource data parameters in response to the switching;

[0084] Specifically, multiple sensors are deployed to form a sensor network to continuously collect the single energy consumption resource parameters of the lighting system. A dynamic sliding window is established to store and analyze the energy consumption data of the past N times. A window size N is set to represent the number of energy consumption data points stored in the sliding window. Whenever new energy consumption data arrives, the sliding window is updated, the oldest data is removed, and the latest data is added. Let E_i be the energy consumption data collected for the i-th time. The energy consumption data set in the sliding window is , the average energy consumption is μ, then: Among them, Ei is the energy consumption data collected for the i-th time, which is directly measured by the sensor; N is the size of the sliding window; whenever new energy consumption data arrives, the average value of the energy consumption data in the sliding window is calculated in real time as the energy consumption benchmark value; according to the impact of environmental factors on lighting energy consumption, weighted coefficients are set for different environmental factors, and the energy consumption adjustment factor affected by comprehensive environmental factors is set as F, then: Among them, wj is the weighted coefficient of the jth environmental factor, xj is the real-time value of the jth environmental factor, and k is the total number of environmental factors; the energy consumption benchmark value μadj after updating the environmental factors is: The difference in the mean values, ΔE, is: When the difference exceeds the preset threshold value M, a fault self-check signal is triggered; upon receiving the fault self-check signal, the system switches to the secondary mode, wherein the fault self-check signal includes energy consumption abnormality, and the secondary mode includes an energy-saving mode. In the energy-saving mode, the lighting system will reduce the brightness, adjust the color temperature, or turn off some lamps. The switching of the secondary mode is a smooth switching mechanism, which uses an exponential gradient method to adjust the brightness and color temperature parameters. When switching to the secondary mode, the system will not cause a sudden visual impact on the user. Assume that the initial brightness is L0 and the target brightness is L target , the switching time is T, the current time is t, then the change of brightness L(t) with time t is: Wherein, λ is a parameter for controlling the gradient speed; and / or dynamically turning off some lamps in non-critical areas when switching to the secondary mode; in response to the switching, obtaining existing resource data parameters according to the sensor network, the existing resource data parameters including current, voltage, and temperature, and performing cleaning, denoising and standardization processing on the existing resource data parameters;

[0085] It should be noted that in a large industrial plant, a sensor network consisting of 10 sensors is deployed to monitor the energy consumption of the lighting system and its related environmental factors. Each sensor is responsible for collecting a single energy resource parameter of the lighting system (such as power consumption) and recording environmental factors (such as temperature, humidity, and light intensity). The sensor network is wirelessly connected to the central control system to achieve real-time transmission and processing of data.

[0086] Sliding window and energy consumption data analysis, set the sliding window size N=20, which means that the sliding window stores the most recent 20 energy consumption data points. Whenever new energy consumption data arrives, update the sliding window, remove the oldest data, and add the latest data. For example, the energy consumption data set in the current sliding window is , the newly collected energy consumption data is E21, then the updated sliding window is According to the energy consumption data in the sliding window, the average energy consumption value μ is calculated: The average energy consumption value obtained in a certain calculation is μ=100 W.

[0087] Environmental factors and energy consumption adjustment factors, consider three environmental factors: temperature (x1), humidity (x2), light intensity (x3), and set their weighting coefficients w1=0.5, w2=0.3, w3=0.2 respectively. The real-time measured environmental factor values ​​are: temperature x1=25℃, humidity x2=60%, light intensity x3=300 lux.

[0088] Here, their impact on energy consumption is set to be linear, and based on standard values ​​(such as temperature 25°C, humidity 50%, and light intensity 200 lux). The greater the deviation from the standard value, the greater the energy consumption adjustment factor F. To simplify the calculation, when each environmental factor deviates from the standard value, the energy consumption adjustment factor increases by 0.01 for each unit deviation (such as temperature deviates from 1°C, humidity deviates from 1%, and light intensity deviates from 10lux).

[0089] Then the energy consumption adjustment factor F is calculated as: =0.0032

[0090] The energy consumption benchmark μadj after updating the environmental factors is: Fault self-check and secondary mode switching, the newly collected energy consumption data Enew=110 W, calculate the difference between the average values ​​DeltaE: Set the preset threshold M=5 W. Since ΔE>M, the fault self-check signal is triggered. After receiving the fault self-check signal, the system switches to energy-saving mode. Set the initial brightness L0=100%, the target brightness Ltarget=70%, the switching time T=5 seconds, and the parameter λ for controlling the gradient speed=0.2. Then the change of brightness L(t) with time t is: At t=5 seconds, the brightness smoothly transitions to 70%. At the same time, the system dynamically turns off some lights in non-critical areas to further reduce energy consumption.

[0091] Data cleaning and processing, in response to the switch, the system obtains the existing resource data parameters, including current, voltage, temperature, etc., based on the sensor network. These data are cleaned, denoised and standardized for subsequent analysis. For example, outliers are removed, data fluctuations are smoothed, and data is normalized to a specific range.

[0092] 202. According to the existing resource data parameters and a preset integrated learning model, prediction results of multiple base learners are obtained, and the prediction results of the multiple base learners are integrated to obtain a first-level detection result. According to the first-level detection result, it is preliminarily determined whether there is a fault. If the first-level detection result determines that there is no fault, the recovery mode is switched and the first-level detection result is recorded. If the first-level detection result determines that there is a fault, the shutdown mode is switched.

[0093] Specifically, multiple base learners are selected, and the base learners include decision trees, support vector machines, and neural networks. The preset historical fault data and normal operation data are used to train the base learners to obtain a trained integrated learning model; according to the existing resource data parameters, they are input into the trained integrated learning model, and each base learner gives its own prediction result. The prediction results of multiple base learners are combined by using weighted voting, stacking or bagging integration strategies to obtain a first-level detection result; a threshold L is preset, and if the first-level detection result exceeds the threshold L, it is judged that there is a fault, otherwise, it is judged that there is no fault. If it is judged that there is no fault, it is switched to a recovery mode, and the recovery mode includes restarting the system, adjusting parameters, and recording the first-level detection result. If it is judged that there is a fault, it is switched to a shutdown mode, and the shutdown mode includes cutting off the power supply, sounding an alarm, and recording the fault information; if the number of times in response to the switching within a unit time r is greater than n, it is directly switched to the shutdown mode;

[0094] It should be noted that the data preparation is as follows: Historical fault data: contains 1000 records, each record includes multiple features (such as current, voltage, temperature, etc.) and the corresponding fault type label. Normal operation data: contains 2000 records, each record also includes multiple features, but no fault type label (or the label is "normal").

[0095] Base learner training: Decision tree: Use information gain as the splitting criterion, and the maximum depth of the tree is set to 5 to avoid overfitting. Support vector machine: Use radial basis function (RBF) as the kernel function, and select the optimal penalty parameter C and kernel parameter γ through cross-validation. Neural network: Construct a three-layer neural network consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the hidden layer is set to half the number of features, ReLU is used as the activation function, Adam is selected as the optimizer, and the loss function is cross entropy. Ensemble learning model: Use a weighted voting strategy to combine the prediction results of the three base learners. The weight is determined based on the performance of each base learner on the validation set (such as accuracy).

[0096] First-level detection and mode switching, data input and prediction: The existing resource data parameters (such as current, voltage, temperature, etc.) are input into the trained integrated learning model in real time. Each base learner gives its own prediction result (fault or normal), and the first-level detection result is obtained according to the weighted voting strategy.

[0097] Threshold setting and judgment: The preset threshold L is 0.6 (that is, if the proportion of faults in the weighted voting results exceeds 60%, it is considered that there is a fault). If the first-level detection result exceeds the threshold L, it is judged that there is a fault; otherwise, it is judged that there is no fault.

[0098] Mode switching: If it is determined that there is no fault, switch to recovery mode. Recovery mode includes restarting the system or adjusting parameters (such as adjusting the brightness and color temperature of the lamp). And record the first-level test results for subsequent analysis. If it is determined that there is a fault, switch to shutdown mode. Shutdown mode includes cutting off the power supply, issuing an alarm (such as notifying maintenance personnel through an audible and visual alarm or sending a text message / email), and recording fault information (such as fault time, fault type, etc.).

[0099] Frequent switching processing: Set the unit time r to 10 minutes and the number of responses to switching n to 3. If the number of responses to switching is greater than 3 times within 10 minutes (i.e., frequent switching from fault-free to recovery mode), directly switch to shutdown mode to avoid possible system instability or potential failure.

[0100] In the lighting system of an industrial plant, the existing resource data parameters input in real time are: current = 1.2A, voltage = 220V, temperature = 45°C. After these parameters are input into the trained ensemble learning model, the following prediction results are obtained: decision tree: fault; support vector machine: normal; neural network: fault; according to the weighted voting strategy (assuming the weights are 0.4, 0.3, and 0.3 respectively), the calculated fault ratio is 0.7 (i.e. (0.41 + 0.30 + 0.3*1) = 0.7), which exceeds the preset threshold L = 0.6, so it is judged that there is a fault.

[0101] The system then switches to shutdown mode, cuts off power and issues an alarm. After receiving the alarm, maintenance personnel go to the site to troubleshoot and repair the fault. At the same time, the system records the fault information for subsequent analysis and optimization.

[0102] 203. In response to the conversion, obtain secondary operation status data, obtain a model prediction value based on the secondary operation status data in combination with a preset time series model, calculate the difference between the actual parameter value of the secondary operation status data and the model prediction value, and obtain an early warning numerical result;

[0103] Specifically, once the system determines that there is a fault and enters the shutdown mode, a data collection mechanism is immediately triggered. The data collection mechanism is designed to quickly capture and store the secondary operating status data of the lighting system within a short period of time after the fault occurs. The secondary operating status data includes user feedback fault reports, historical fault logs, remote access logs, system resource usage, and environmental and contextual data; the collected secondary operating status data is cleaned to remove outliers and noise, and the data is standardized or normalized to extract features related to fault prediction, including trends, seasonality, and periodicity in time series data. The sliding window technology is used to capture the time dependency of the data; the preset historical fault data and normal operating data are used for training to obtain a trained time series model. The time series model includes a long short-term memory network, a gated recurrent unit, and a Transformer model to obtain an integrated time series model; set P LSTM (x) , P GRU (x) and P Transformer (x) are the prediction values ​​of the long short-term memory network, gated recurrent unit and Transformer model respectively, then the integrated prediction value P ensemble (x) for: The preprocessed secondary operation status data is input into the integrated time series model, and the time series model outputs the predicted status data value in the future period, and the difference between the actually measured secondary operation status data and the model predicted value is calculated; a variety of indicators are used to measure the difference to obtain the warning numerical result, including mean square error (MSE), mean absolute error (MAE), dynamic time warping (DTW) distance, and a difference threshold U is set. When the difference between the actual data and the predicted data exceeds this threshold, a warning signal is generated; the warning numerical result includes the difference size and the warning level;

[0104] It should be noted that the fault self-detection and data collection mechanism, fault judgment and shutdown mode: the system determines that there is a fault through previous methods (such as the base learner integration model) and has entered the shutdown mode.

[0105] Data collection mechanism triggering: Once entering the shutdown mode, the data collection mechanism is immediately triggered to quickly capture and store the secondary operating status data of the lighting system within a short time (e.g., within 5 seconds) after the fault occurs.

[0106] Secondary operation status data: User feedback fault reports: 2, describing the fault phenomenon. Historical fault logs: 10, recording past faults and repairs. Remote access logs: 5, recording remote access system operations. System resource usage: CPU usage 75%, memory usage 60%, disk usage 50%. Environmental and situational data: temperature 25°C, humidity 60%, light intensity 300 lux.

[0107] Data preprocessing and data cleaning: remove outliers and noise. For example, the instantaneous data of CPU usage suddenly jumping to 100% is regarded as an outlier. Data standardization processing: normalize each data item to the interval [0,1]. Feature extraction: extract trend, seasonality, and periodicity features from time series data. Use sliding window technology and set the window size to 10 to capture the time dependency of data.

[0108] Time series model training and prediction,Model training: Use the preset historical fault data and normal operation data (assuming each contains 1,000 records) for training. Train the long short-term memory network (LSTM), gated recurrent unit (GRU) and Transformer models to obtain an integrated time series model.

[0109] Integrated prediction: Input the preprocessed secondary operation status data into the integrated model. Set the prediction values ​​of LSTM, GRU, and Transformer models to 0.7, 0.6, and 0.8 respectively (after normalization).

[0110] According to the integrated prediction formula, the integrated prediction value is calculated: Difference calculation: The actual measured secondary operating status data (assuming the same preprocessing) is compared with the model prediction value. The difference is calculated, assuming that the actual data is 0.65.

[0111] Warning numerical results and threshold settings, difference measurement indicators:

[0112] Mean Square Error (MSE): (0.7−0.65) 2 =0.0025

[0113] Mean absolute error (MAE): |0.7−0.65|=0.05

[0114] Dynamic Time Warping (DTW) distance: Since this example is a single time point data, the DTW distance is not applicable and can be ignored or replaced with other indicators suitable for single point comparison.

[0115] Threshold setting and warning signal: Set the difference threshold U to 0.1. Since MAE = 0.05 is less than the threshold U, no warning signal is generated.

[0116] Warning numerical results: Difference size: 0.05; Warning level: None (because the difference does not exceed the threshold).

[0117] In this embodiment, after the system determines that there is a fault and enters the shutdown mode, it successfully triggers the data collection mechanism and quickly captures the secondary operating status data of the lighting system. After data preprocessing and feature extraction, the data is input into the integrated time series model for prediction. By calculating the difference between the actual data and the predicted data and measuring it with multiple indicators, it is finally determined that the difference does not exceed the set threshold, so no warning signal is generated.

[0118] 204. According to the secondary operation status data and the preset Bayesian network model of the lighting system, possible results of the fault cause and possible results of the propagation path are obtained, and the secondary detection result is obtained according to the warning value result, the possible results of the fault cause and the possible results of the propagation path;

[0119] Specifically, an initial Bayesian network structure is constructed, initial probability values ​​are assigned to nodes in the network, and the Bayesian network model is trained based on historical fault data of the lighting system. A method combining structure learning and parameter learning is used to optimize the network structure and node probability values ​​to obtain a trained Bayesian network model of the lighting system; the secondary operating status data is input into the trained Bayesian network model of the lighting system, and the posterior probability of each fault cause node is calculated using the inference algorithm in the Bayesian network model of the lighting system. According to the size of the posterior probability, the most likely fault cause and its combination are determined to obtain the possible result of the fault cause, and the inference algorithm includes a clique tree algorithm and a variable elimination method; in the Bayesian network, the dependency relationship between the fault cause node and the fault phenomenon node is analyzed, and the path search algorithm is used to find the possible result of the propagation path from the fault cause to the fault phenomenon, and the path search algorithm includes a depth-first search and a breadth-first search; the warning numerical result, the possible result of the fault cause and the possible result of the propagation path are combined, and the warning numerical result is graded and classified to obtain the secondary detection result;

[0120] It should be noted that the Bayesian network is constructed and trained to construct the initial Bayesian network structure: the lighting system Bayesian network is set to contain the following nodes: power consumption (P), temperature (T), humidity (H), light intensity (L), bulb life (B), circuit failure (C), power failure (S) and fault phenomenon (F).

[0121] Based on expert knowledge and experience, preliminary dependencies between nodes are established, such as power consumption is affected by bulb life and circuit failure, and failure symptoms are affected by all other nodes.

[0122] Assign initial probability values: Assign an initial probability value to each node, such as the initial failure probability of light bulb life B is 0.01, the initial failure probability of circuit fault C is 0.02, etc.

[0123] Model training: Use the historical fault data of the lighting system (including fault type, fault cause and corresponding environmental factor data) to train the Bayesian network model. Use a combination of structural learning and parameter learning to optimize the network structure and node probability values. For example, adjust the dependencies between nodes through structural learning, and update the probability values ​​of nodes through parameter learning. After training, an optimized Bayesian network model is obtained, in which the probability values ​​of each node have been adjusted according to historical data.

[0124] Fault cause reasoning and propagation path analysis, input secondary operation status data: It is assumed that at a certain moment, the secondary operation status data collected by the sensor network is: power consumption is abnormally increased, the temperature is 25°C, the humidity is 60%, the light intensity is normal, the bulb life has been used for 80%, there is no obvious abnormality in the circuit, and the power supply is stable.

[0125] Calculate the posterior probability: Input the above data into the trained Bayesian network model, use the clique tree algorithm or variable elimination method for reasoning, and calculate the posterior probability of each fault cause node. The calculated results are: the posterior probability of the bulb life is insufficient is 0.7, the posterior probability of the circuit failure is 0.2, and the posterior probability of the power failure is 0.1.

[0126] Determine the cause of the fault: According to the size of the posterior probability, the most likely cause of the fault is the insufficient life of the bulb. Analyze the propagation path: Use a path search algorithm (such as depth-first search or breadth-first search) to analyze the dependency between the fault cause node and the fault phenomenon node. The propagation path found is: insufficient life of the bulb -> abnormal increase in power consumption -> fault phenomenon occurs.

[0127] The secondary detection results are generated and combined with the warning numerical results: the warning numerical results at this time are: mean square error MSE=0.01, mean absolute error MAE=0.05, and do not exceed the set difference threshold.

[0128] Grading and classification processing: Combine the warning value results, possible results of the fault cause (lamp life is insufficient) and possible results of the propagation path (lamp life is insufficient -> power consumption increases abnormally -> fault phenomenon occurs).

[0129] According to the severity of the fault cause and propagation path, the warning numerical results are graded and classified. In this example, the fault cause "lamp life is insufficient" is classified as "maintainability fault" and the warning level is set to "low" (because the MAE does not exceed the threshold and the fault cause is clear and can be repaired).

[0130] Generate secondary detection results: The final secondary detection results are: the cause of the failure is the insufficient life of the bulb, the warning level is low, and the recommended maintenance measure is to replace the bulb.

[0131] 205. Obtain a fault self-detection result according to the first-level detection result and the second-level detection result.

[0132] Specifically, the preset historical fault data and normal operation data are used to assign weights to the primary detection result and the secondary detection result. Assuming that the primary detection result is R1 and the secondary detection result is R2, the weight assignment formula is: Wherein, NF1 is the number of fault data in the first-level detection result, and NF2 is the number of fault data in the second-level detection result. The first-level detection result and the second-level detection result are fused by using the decision fusion and evidence theory method to obtain the fusion output result. For each detection result, a basic probability distribution function is defined to indicate the support degree of the result for a certain fault type. Let Θ be the set of all possible fault types. For each : Among them, R1(θ) and R2(θ) are the scores or probabilities of the first-level detection results and the second-level detection results for the fault type θ respectively; the Dempster synthesis rule is used to fuse the basic probability distribution of the two detection results. Let m12 be the basic probability distribution after fusion, then: Among them, K is the conflict coefficient, which indicates the inconsistency between the two pieces of evidence. The fusion output result is mapped to the specific fault type and level through a mapping table or a classifier to obtain the fault self-check result.

[0133] It should be noted that data preparation, historical fault data and normal operation data: historical fault data includes various fault types and their corresponding quantities, such as 100 times of bulb failure, 50 times of circuit failure, etc. Normal operation data includes various parameter values ​​and their corresponding quantities when there is no fault.

[0134] Level 1 test results and level 2 test results: Level 1 test results (R1): Fault judgment obtained through preliminary analysis of sensor data, such as abnormal power consumption, with a score of 0.8 (assuming the full score is 1); Level 2 test results (R2): Fault judgment obtained through in-depth analysis of sensor data and historical data, such as insufficient bulb life, with a score of 0.9.

[0135] Weight distribution, calculation of the number of fault data: Number of fault data in the first-level detection results (NF1): Assumed to be 1 (i.e., 1 power consumption abnormality was detected). Number of fault data in the second-level detection results (NF2): Assumed to be 1 (i.e., 1 bulb life shortage was detected).

[0136] Calculate the weight according to the weight distribution formula: Basic probability distribution, defining a set of fault types:

[0137] Θ={bulb failure, circuit failure, power failure, other failure}

[0138] Calculate the basic probability distribution function: For the first-level detection results: = =0

[0139] Since the first-level detection result only points out abnormal power consumption, it can be regarded as indirect support for other fault types. To simplify the calculation, it is set to be evenly distributed to other fault types, that is:

[0140] m1(circuit fault)=m1(power supply fault)=m1(other faults)=3 / 0.8≈0.27;

[0141] For secondary test results: (Note: The secondary results clearly indicate that the lamp life is insufficient)

[0142] m2(circuit fault)=m2(power supply fault)=m2(other fault)=0

[0143] Dempster synthesis rule fusion, calculate the conflict coefficient: (Note: Since the secondary results are clear and the primary results do not conflict with the secondary results, the conflict coefficient is 0)

[0144] Fusion basic probability distribution: (Note: It is understood here that the first-level result's support for other fault types is partially covered by the second-level result that is clearly a bulb fault);

[0145] However, since the secondary result is very clear and the primary result directly supports the bulb failure, to simplify the understanding, it can be considered that the fusion mainly depends on the secondary result, that is:

[0146] m 12 (Bulb failure)≈1

[0147] m 12 (Circuit Fault) = m 12 (Power failure) = m 12 (Other faults)≈0

[0148] Map to specific fault types and levels through a mapping table or classifier: According to the basic probability distribution after fusion, the fault type is determined to be "bulb failure". The fault levels are set to be minor, medium, and severe. Since the insufficient life of a bulb is usually a maintainability fault, it is classified as "minor".

[0149] Fault self-test results, fault type: lamp failure; fault level: minor.

[0150] 206. Use visualization technology to display fault self-inspection results and generate detailed fault reports, including fault type, level, possible cause, propagation path and recommended repair measures; allow users to provide feedback on fault self-inspection results, including the accuracy and completeness of the results and recommended improvement directions; monitor the operating status of the lighting system in real time through the management platform, and trigger the early warning mechanism in time when potential faults are detected, and notify relevant personnel to handle them.

[0151] It should be noted that visualization technology displays the fault self-detection results in the form of charts and maps through the graphical user interface (GUI) of the central control system.

[0152] Specific data: Chart display: such as a line graph of power consumption over time, with abnormal peaks and corresponding times marked. Map display: The location of each sensor is marked on the plant layout map, and the sensor status (normal, abnormal) is represented by different colors or icons. Fault point location: The location of the faulty sensor is accurately marked on the map, and the fault type icon (such as bulb failure, circuit failure, etc.) is displayed.

[0153] Generate a detailed fault report, implementation method: When the fault self-checking system detects a fault, it automatically generates a fault report containing the following information: Fault type: such as bulb failure. Fault level: minor (such as bulb brightness drops), moderate (such as bulb flickers frequently), severe (such as bulb does not light up at all). Possible causes: bulb aging, poor circuit contact, etc. Propagation path: Analyze the areas or systems that the fault may affect, such as the impact of a single bulb failure on the surrounding light intensity. Recommended repair measures: replace the bulb, check the circuit connection, etc.

[0154] Specific data: Fault report example: "At 10:00 on March 15, 2023, sensor 3 detected a bulb failure. The fault level is medium. The possible cause is bulb aging. It is recommended to replace the bulb. The fault may affect the plant area A. Please ask relevant personnel to handle it in time."

[0155] User feedback mechanism, implementation method: set feedback button on GUI to allow users to provide feedback on fault self-test results. Specific data: User feedback form example: including fault report number, user name, feedback time, feedback content (such as "the report is accurate, but it is recommended to add circuit inspection items"), satisfaction score (such as 1-5 stars).

[0156] Real-time monitoring and early warning on the management platform. Implementation method: Real-time monitoring of the operating status of the lighting system through the management platform, setting thresholds to trigger early warning mechanisms. Specific data: Real-time monitoring data: For example, the current power consumption of sensor 3 is 120W, the normal range is 100-110W, and exceeding the threshold triggers an early warning. Early warning notification: When a potential fault is detected, the system automatically sends an early warning notification to the mobile phone or email of the relevant personnel. The notification content includes the fault type, location, recommended measures and urgency. Early warning processing record: Record the sending time, receiving personnel, processing time and processing results of the early warning notification.

[0157] In the embodiment of the present invention, after a fault is detected, the system can quickly switch to the secondary mode or shutdown mode to reduce the impact of the fault on production, and quickly locate and isolate the fault point through detailed fault cause and propagation path analysis. Through real-time monitoring and early warning mechanisms, potential faults can be discovered in advance to avoid greater losses caused by fault expansion, reduce unplanned downtime, and thus reduce maintenance costs. In the secondary mode, the energy consumption of the lighting system is optimized and resource utilization efficiency is improved by adjusting the brightness, color temperature or turning off the lamps in non-critical areas. The fault self-check results are displayed through visualization technology, and a detailed fault report is generated, so that users can intuitively understand the system status and improve user experience and satisfaction. Allow users to provide feedback on the fault self-check results, and continuously optimize model parameters and algorithms based on user suggestions to improve the system's adaptability and intelligence level. In summary, this technology realizes efficient fault self-check and diagnosis of the lighting system through the comprehensive use of multiple technical means, which is of great significance in improving system reliability, reducing maintenance costs, and optimizing resource utilization.

[0158] The above describes the fault self-detection method of the lighting system in the embodiment of the present invention. The following describes the fault self-detection device of the lighting system in the embodiment of the present invention. Figure 3In one embodiment of the present invention, a fault self-detection device for a lighting system includes: an acquisition module 301, which is used to receive a fault self-detection signal, switch to a secondary mode, and obtain existing resource data parameters in response to the switch; a detection module 302, which is used to obtain prediction results of multiple base learners based on the existing resource data parameters in combination with a preset integrated learning model, integrate the prediction results of multiple base learners to obtain a primary detection result, and preliminarily determine whether there is a fault based on the primary detection result. If the primary detection result determines that there is no fault, switch to the recovery mode and record the primary detection result. If the primary detection result determines that there is a fault, switch to the shutdown mode; prediction module 303, used to obtain secondary operating status data in response to the conversion, obtain model prediction values ​​based on the secondary operating status data in combination with a preset time series model, calculate the difference between the actual parameter value of the secondary operating status data and the model prediction value, and obtain an early warning numerical result; a setting module 304, used to obtain possible results of the cause of the fault and possible results of the propagation path based on the secondary operating status data in combination with a preset Bayesian network model of the lighting system, and obtain a secondary detection result based on the early warning numerical result, the possible results of the cause of the fault and the possible results of the propagation path; an allocation module 305, used to obtain a fault self-detection result based on the primary detection result and the secondary detection result.

[0159] In the embodiment of the present invention, by combining the integrated learning model and the time series model, the fault of the lighting system can be detected more accurately and the false alarm and missed alarm phenomena can be reduced; by calculating the difference between the actual parameter value and the predicted value, an early warning can be issued before the fault occurs, providing a time window for timely taking measures to prevent the fault from expanding; by analyzing the cause of the fault and the propagation path through the Bayesian network model, it can help maintenance personnel quickly locate the fault point and improve maintenance efficiency and accuracy; the fault self-detection device can monitor the operating status of the lighting system in real time, and promptly discover and handle potential faults, thereby enhancing the reliability and stability of the entire lighting system; through intelligent and automated fault detection and early warning, the number of manual inspections and repairs can be reduced, and the maintenance cost of the lighting system can be reduced.

[0160] above Figure 3 and Figure 4 The fault self-detection device of the lighting system in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The fault self-detection device of the lighting system in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0161] Figure 44 is a schematic diagram of a structure of a fault self-checking device for a lighting system provided by an embodiment of the present invention. The fault self-checking device 400 for the lighting system may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 410 (for example, one or more processors) and a memory 420, and one or more storage media 430 (for example, one or more mass storage devices) storing application programs 433 or data 432. Among them, the memory 420 and the storage medium 430 may be temporary storage or permanent storage. The program stored in the storage medium 430 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the fault self-checking device 400 for the lighting system. Furthermore, the processor 410 may be configured to communicate with the storage medium 430 to execute a series of instruction operations in the storage medium 430 on the fault self-checking device 400 for the lighting system.

[0162] The fault self-diagnosis device 400 of the lighting system may also include one or more power supplies 440, one or more wired or wireless network interfaces 450, one or more input and output interfaces 460, and / or one or more operating systems 431, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be understood by those skilled in the art that Figure 4 The structure of the fault self-checking device of the lighting system shown does not constitute a limitation on the fault self-checking device of the lighting system, and may include more or less components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0163] The present invention also provides a fault self-detection device for a lighting system, the fault self-detection device for the lighting system comprising a memory and a processor, the memory storing computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the fault self-detection method for the lighting system in the above-mentioned embodiments.

[0164] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the fault self-detection method of the lighting system.

[0165] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0166] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0167] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault self-detection method for a lighting system, characterized in that: The fault self-detection method of the lighting system comprises: Acquire actual energy consumption, and if the difference between the actual energy consumption and the reference value exceeds a preset threshold, trigger a fault self-check signal, receive the fault self-check signal, switch to a secondary mode, and in response to the switch, acquire existing resource data parameters; According to the existing resource data parameters combined with a preset integrated learning model, prediction results of multiple base learners are obtained, and the prediction results of the multiple base learners are integrated to obtain a first-level detection result, and a preliminary judgment is made on whether there is a fault according to the first-level detection result. If the first-level detection result judges that there is no fault, the system is switched to a recovery mode and the first-level detection result is recorded. If the first-level detection result judges that there is a fault, the system is switched to a shutdown mode. In response to the conversion, secondary operation status data is acquired, a model prediction value is obtained based on the secondary operation status data in combination with a preset time series model, and a difference between an actual parameter value of the secondary operation status data and the model prediction value is calculated to obtain a warning numerical result; According to the secondary operation status data combined with a preset Bayesian network model of the lighting system, possible results of the fault cause and possible results of the propagation path are obtained, and the secondary detection result is obtained according to the warning value result, the possible results of the fault cause and the possible results of the propagation path; Obtaining a fault self-test result based on the first-level test result and the second-level test result, including: Using the preset historical fault data and normal operation data, weights are assigned to the primary detection results and the secondary detection results. Assuming the primary detection result is R1 and the secondary detection result is R2, the weight assignment formula is: Among them, N F1 is the number of fault data in the first-level detection results, N F2 is the number of fault data in the secondary detection results; The first-level detection results and the second-level detection results are fused by using the method of decision fusion and evidence theory to obtain a fusion output result. For each detection result, a basic probability distribution function is defined to indicate the support degree of the result for a certain fault type. Let Θ be the set of all possible fault types. For each θ∈Θ: Among them, R1(θ) and R2(θ) are the scores or probabilities of the first-level detection results and the second-level detection results for the fault type θ, respectively. θ′ represents another specific fault type in the fault type set θ. R1(θ′) represents the support degree of the first-level detection result for the fault type θ′, and R2(θ′) represents the support degree of the second-level detection result for the fault type θ′. Use Dempster's synthesis rule to fuse the basic probability distribution of two detection results. Let m 12 is the basic probability distribution after fusion, then: Among them, K is the conflict coefficient, which indicates the degree of inconsistency between two pieces of evidence, and A i represents a subset or category in the fault type set Θ, which is used to classify or map the fault types; The fusion output results are mapped to specific fault types and levels through a mapping table or classifier to obtain the fault self-detection results.

2. The fault self-detection method of the lighting system according to claim 1, characterized in that: The receiving of the fault self-checking signal, switching to the secondary mode, and obtaining the existing resource data parameters in response to the switching, include: Deploy multiple sensors to form a sensor network, continuously collect the single energy consumption resource parameters of the lighting system, establish a dynamic sliding window to store and analyze the energy consumption data of the past N times, set a window size N, which represents the number of energy consumption data points stored in the sliding window, and calculate the average value of the energy consumption data in the sliding window in real time whenever new energy consumption data arrives, as the energy consumption benchmark value; According to the impact of environmental factors on lighting energy consumption, weighted coefficients are set for different environmental factors; The difference between the average values ​​is obtained based on the weighted coefficient and the energy consumption benchmark value. If the difference exceeds the preset threshold value M, a fault self-check signal is triggered; Upon receiving a fault self-check signal, switching to a secondary mode, wherein the fault self-check signal includes abnormal energy consumption, and the secondary mode includes an energy-saving mode. In the energy-saving mode, the lighting system will reduce the brightness, adjust the color temperature, or turn off some lamps. The switching of the secondary mode is a smooth switching mechanism, which uses an exponential gradient method to adjust the brightness and color temperature parameters, and will not cause a sudden visual impact on the user when switching to the secondary mode; and / or dynamically turning off some lights in non-critical areas when switching to secondary mode; In response to the switching, existing resource data parameters are acquired according to the sensor network, the existing resource data parameters including current, voltage, and temperature, and the existing resource data parameters are cleaned, denoised, and standardized.

3. The fault self-detection method of the lighting system according to claim 1, characterized in that: The method of obtaining prediction results of multiple base learners based on the existing resource data parameters in combination with a preset integrated learning model, integrating the prediction results of the multiple base learners to obtain a first-level detection result, and preliminarily judging whether there is a fault based on the first-level detection result. If the first-level detection result judges that there is no fault, switching to a recovery mode and recording the first-level detection result. If the first-level detection result judges that there is a fault, switching to a shutdown mode includes: Select multiple base learners, including decision trees, support vector machines, and neural networks, and use preset historical fault data and normal operation data to train the base learners to obtain a trained integrated learning model; According to the existing resource data parameters, the parameters are input into the trained ensemble learning model, each base learner gives its own prediction result, and a weighted voting, stacking or bagging ensemble strategy is used to combine the prediction results of multiple base learners to obtain a first-level detection result; A threshold L is preset. If the first-level detection result exceeds the threshold L, it is judged that there is a fault. Otherwise, it is judged that there is no fault. If it is judged that there is no fault, it is switched to the recovery mode. The recovery mode includes restarting the system, adjusting parameters, and recording the first-level detection result. If it is judged that there is a fault, it is switched to the shutdown mode. The shutdown mode includes cutting off the power supply, issuing an alarm, and recording the fault information. If the number of times in response to the switching within the unit time r is greater than n, it is directly converted to the off mode.

4. The fault self-detection method of the lighting system according to claim 1, characterized in that: The step of obtaining secondary operation status data in response to the conversion, obtaining a model prediction value based on the secondary operation status data in combination with a preset time series model, calculating a difference between an actual parameter value of the secondary operation status data and the model prediction value, and obtaining a warning numerical result includes: Once the system determines that there is a fault and enters shutdown mode, a data collection mechanism is immediately triggered. The data collection mechanism is designed to quickly capture and store secondary operating status data of the lighting system within a short period of time after the fault occurs. The secondary operating status data includes user feedback fault reports, historical fault logs, remote access logs, system resource usage, and environmental and contextual data; Use the preset historical fault data and normal operation data for training to obtain a trained time series model, which includes a long short-term memory network, a gated recurrent unit, and a Transformer model to obtain an integrated time series model; The preprocessed secondary operation status data is input into the integrated time series model, the time series model outputs the predicted status data value in the future period, and the difference between the actually measured secondary operation status data and the model predicted value is calculated; A variety of indicators are used to measure the difference and obtain the warning numerical results, including mean square error (MSE), mean absolute error (MAE), and dynamic time warping (DTW) distance. A difference threshold U is set. When the difference between the actual data and the predicted data exceeds this threshold, a warning signal is generated. The warning numerical result includes the difference size and the warning level.

5. The fault self-detection method of the lighting system according to claim 1, characterized in that: The possible results of the fault cause and the possible results of the propagation path are obtained based on the secondary operation status data in combination with the preset Bayesian network model of the lighting system, and the secondary detection results are obtained based on the warning value results, the possible results of the fault cause and the possible results of the propagation path, including: According to the secondary operation status data, the trained Bayesian network model of the lighting system is input, and the posterior probability of each fault cause node is calculated using the inference algorithm in the Bayesian network model of the lighting system. According to the size of the posterior probability, the most likely fault cause and its combination are determined to obtain the possible results of the fault cause, and the inference algorithm includes a clique tree algorithm and a variable elimination method; In the Bayesian network, the dependency between the fault cause node and the fault phenomenon node is analyzed, and the possible results of the propagation path from the fault cause to the fault phenomenon are found using the path search algorithm. The path search algorithm includes depth-first search and breadth-first search. The warning numerical result, the possible result of the fault cause and the possible result of the propagation path are combined, and the warning numerical result is graded and classified to obtain a secondary detection result.

6. The fault self-detection method of the lighting system according to claim 1, characterized in that: Also includes: Use visualization technology to display fault self-check results and generate detailed fault reports, including fault type, level, possible cause, propagation path and recommended repair measures; Allow users to provide feedback on fault self-check results, including accuracy, completeness and suggested improvement directions; The operating status of the lighting system is monitored in real time through the management platform, and the early warning mechanism is triggered in time when potential faults are detected to notify relevant personnel for processing.

7. A fault self-detection device for a lighting system, characterized in that: The fault self-detection device of the lighting system comprises: An acquisition module, configured to receive a fault self-detection signal, switch to a secondary mode, and acquire existing resource data parameters in response to the switching; A detection module, used to obtain prediction results of multiple base learners according to the existing resource data parameters combined with a preset integrated learning model, integrate the prediction results of the multiple base learners to obtain a first-level detection result, and preliminarily determine whether there is a fault according to the first-level detection result. If the first-level detection result determines that there is no fault, switch to recovery mode and record the first-level detection result. If the first-level detection result determines that there is a fault, switch to shutdown mode; A prediction module, configured to obtain secondary operation status data in response to the conversion, obtain a model prediction value based on the secondary operation status data in combination with a preset time series model, calculate a difference between an actual parameter value of the secondary operation status data and the model prediction value, and obtain a warning numerical result; A setting module is used to obtain possible results of fault causes and possible results of propagation paths according to the secondary operation status data combined with a preset Bayesian network model of the lighting system, and obtain secondary detection results according to the warning value results, the possible results of fault causes and the possible results of propagation paths; The allocation module is used to obtain the fault self-detection result according to the first-level detection result and the second-level detection result.

8. A fault self-checking device for a lighting system, characterized in that: The fault self-checking device of the lighting system comprises: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory to enable the fault self-detection device of the lighting system to execute the fault self-detection method of the lighting system according to any one of claims 1 to 6.

9. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the fault self-detection method of the lighting system according to any one of claims 1 to 6 is implemented.

Images