Temperature monitoring method for photovoltaic grid-connected cabinet

Abstract

The invention discloses a temperature monitoring method for a photovoltaic grid-connected cabinet, and particularly relates to the technical field of photovoltaic grid-connected cabinets. The method comprises the following steps: arranging multiple types of temperature sensors in a grid-connected cabinet, collecting real-time temperature data, constructing a heat distribution model, comparing a current heat map with a historical steady-state heat field template, and identifying heat field offset and an atypical temperature rise trend; when the system detects that early warning is not triggered by an existing measuring point in a heat abnormal area, redundancy alarm logic is automatically started, temperature abnormal early warning is generated, and the speculated abnormal heat source position is recorded; according to the method, the sensing capability of the thermal anomaly in the blind area of the sensor is effectively improved, and the safety and the intelligent level of the system are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic grid-connected cabinets, and in particular to a temperature monitoring method for photovoltaic grid-connected cabinets. Background Art

[0002] Temperature monitoring for photovoltaic grid-connected cabinets involves real-time monitoring of temperature changes within key components (such as circuit breakers, wiring terminals, and cables) during operation. The goal is to promptly detect abnormal temperature rises caused by overload, poor contact, or equipment aging, thereby preventing fires and equipment failures and ensuring the safe and stable operation of the photovoltaic system.

[0003] The existing technology has the following shortcomings: In multi-point temperature monitoring of photovoltaic grid-connected cabinets, due to airflow disturbances or structural design defects inside the cabinet, the heat from local hot spots may not be effectively transferred to the sensor, but instead be guided to other areas by the air duct, causing the temperature rise at the actual fault point to be ignored. Since the system only relies on fixed measurement points and threshold comparison, it is unable to identify this thermal distribution anomaly, which can lead to serious consequences such as equipment burning or even fire. Summary of the Invention

[0004] The purpose of the present invention is to provide a temperature monitoring method for a photovoltaic grid-connected cabinet to solve the shortcomings in the background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a temperature monitoring method for a photovoltaic grid-connected cabinet, comprising: First-class temperature sensors are installed at multiple electrical nodes inside the photovoltaic grid-connected cabinet and their surrounding areas to collect real-time temperature data; A second type of ambient temperature sensor is installed in the open area of the cabinet to collect data on thermal field changes caused by local air flow; A dynamic thermal distribution model is constructed using the collected temperature data. The model identifies potential thermal field excursions by analyzing the temperature differences, change rates, and spatial position relationships of multiple measurement points. Based on the thermal distribution model, the current temperature data is compared with the thermal field pattern under historical normal conditions to determine whether there is any abnormal thermal field; When the model determines that there is a thermal field anomaly and there is no corresponding temperature rise warning at the deployed measuring points, the redundant alarm logic is triggered, a temperature anomaly alarm is issued, and the estimated location of the abnormal hot area is recorded.

[0006] Preferably, constructing a dynamic heat distribution model includes: Normalize the data collected by each temperature sensor at different time points, unify the unit and sampling period, and establish a time series matrix; Based on the physical layout coordinates of the sensors, a spatial distribution model is established to map each temperature point to the corresponding position of the cabinet's three-dimensional structural model; By using bilinear interpolation or weighted average method, a continuous temperature field distribution map is constructed to form the heat distribution profile inside the cabinet; Generate a thermal distribution frame sequence at set time intervals to form a thermal field dynamic change model.

[0007] Preferably, analyzing the temperature difference, change rate and spatial position relationship of multiple measuring points includes: Calculate the instantaneous temperature difference between any two adjacent measuring points and compare it with the historical steady-state temperature difference benchmark. If it exceeds the preset fluctuation threshold, it will be marked as an abnormal temperature difference area; After analyzing the temperature change rate of each measuring point in a continuous time period, the abnormal temperature rise index is calculated. If the abnormal temperature rise index is greater than the critical rate threshold, it is identified as an abnormal temperature rise trend point; A position correlation analysis is conducted on the abnormal temperature difference areas and abnormal warming trend points in the spatial model. If the high temperature point and the temperature difference boundary exceed 1.5 times the average spacing between the measuring points, it is inferred that there is a thermal field offset. After detecting the thermal field deviation phenomenon, by comparing the current thermal distribution map with the thermal field pattern under historical stable working conditions, it is confirmed whether it is a structural heat conduction anomaly and trigger an abnormality warning.

[0008] Preferably, the method for calculating the instantaneous temperature difference is as follows: the temperature values collected from any two physically adjacent measuring points within a preset sampling period are organized into a time series, and two continuous time-temperature trajectories are constructed respectively; multi-scale wavelet transform processing is performed on the temperature time series of the two measuring points respectively; by selecting the wavelet coefficient level, the temperature time series of each measuring point is inversely transformed and reconstructed to obtain two sets of denoised and enhanced temperature change curves respectively; the difference between the reconstructed temperature curves of the two measuring points is calculated at each time point, thereby obtaining the instantaneous temperature difference.

[0009] Preferably, the calculation method of the abnormal index of heating rate is as follows: for each temperature measuring point, the temperature change data of the set time window is collected to construct a continuous time series; the empirical mode decomposition operation is performed on the temperature time series of each measuring point to decompose the original temperature curve into several intrinsic mode functions (IMFs); a mode sensitive to heating is selected from multiple IMFs, and the local slope change trend, slope fluctuation amplitude and heating duration period length are extracted according to their fluctuation trend to form a characteristic vector of the heating behavior of each measuring point in the current window; the heating characteristic vectors of all measuring points are input into the fuzzy C-means clustering algorithm for classification, and a membership value belonging to the abnormal heating class is assigned to each measuring point, that is, the abnormal index of heating rate.

[0010] Preferably, the current temperature data is compared with the thermal field pattern under the historical normal state to determine whether there is an abnormal thermal field, including: Collect stable operating temperature data of photovoltaic grid-connected cabinets under multiple states, generate multiple sets of thermal distribution maps, and establish a standardized historical thermal field benchmark library through averaging and outlier elimination technology. Each type of thermal field mode corresponds to a typical distribution template; Normalize the thermal distribution map at the current moment to make it consistent with the historical thermal field template in size, scale and sampling frequency; The structural similarity measurement algorithm is used to compare the current thermal map with each historical thermal field template one by one, and the similarity scores in terms of overall structure, thermal center position and temperature difference distribution are calculated; If the similarity between the current heat map and all historical templates is lower than the set threshold, it is judged as a thermal field anomaly; the anomaly level is further divided according to the degree of similarity deviation, including slight offset and major heat source reconstruction, and the corresponding alarm strategy is triggered.

[0011] Preferably, the thermal field image similarity measurement includes: Extract the characteristic vector of the thermal distribution map, including the coordinates of the thermal center, the maximum temperature difference, the directionality of the thermal gradient between each measuring point, and the temperature mean and variance; The principal component analysis method is used to compress the dimensions of historical and current feature vectors and extract the most discriminative thermal distribution feature axis; The multidimensional distance between the current thermal map and the main axis of each historical thermal field is calculated using the Mahalanobis distance algorithm. If the distance exceeds the specified threshold, it is determined that the current thermal field pattern has deviated from the normal operating state and is marked as an abnormal thermal distribution state.

[0012] Preferably, triggering the redundant alarm logic includes: After determining that the thermal field is abnormal, the characteristic area of the abnormal heat zone is marked in the current thermal distribution map, and a thermal field response map covering the abnormal area is generated; By comparing the location of the abnormal area in the thermal field response diagram with the sensing range of the currently deployed temperature sensors, a spatial overlap detection algorithm is used to determine whether there is a thermal anomaly area that is not monitored by the sensor. If the thermal anomaly area is outside the radial sensing range of all measuring points, or the temperature change does not form an effective gradient at the sensor, it is identified as a warning blind spot. When it is confirmed that a thermal field anomaly exists and there is no warning at the measured point, the redundant alarm logic is activated to trigger a temperature anomaly warning.

[0013] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. This invention achieves precise monitoring of thermal anomalies within photovoltaic grid-connected cabinets by constructing a dynamic heat distribution model and introducing an intelligent spatial temperature field recognition mechanism. Compared to traditional passive early warning methods that rely on measurement point thresholds, this invention proactively identifies potential hotspots within measurement point blind spots and implements model-driven redundant alarms, significantly improving the system's ability to detect thermal field excursions, asymmetric temperature rise, and structural heat dissipation anomalies.

[0014] 2. This invention provides a comprehensive anomaly identification and inferred location mechanism. It can spatially invert and provide early warning of abnormal hotspots without triggering alarms at traditional measurement points, effectively addressing monitoring blind spots in scenarios such as insufficient monitoring points, environmental disturbances, and sensor failure. This technical solution offers high precision, adaptability, and scalability, making it suitable for a variety of practical applications, including smart grids, distributed photovoltaics, and unmanned stations, with significant security implications. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0016] Figure 1 This is a mind map of the method of the present invention. DETAILED DESCRIPTION

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0018] For examples, see Figure 1 As shown, the temperature monitoring method of a photovoltaic grid-connected cabinet described in this embodiment includes: First-class temperature sensors are installed at multiple electrical nodes inside the photovoltaic grid-connected cabinet and their surrounding areas to collect real-time temperature data; A second type of ambient temperature sensor is installed in the open area of the cabinet to collect data on thermal field changes caused by local air flow; A dynamic thermal distribution model is constructed using the collected temperature data. The model identifies potential thermal field excursions by analyzing the temperature differences, change rates, and spatial position relationships of multiple measurement points. Based on the thermal distribution model, the current temperature data is compared with the thermal field pattern under historical normal conditions to determine whether there is any abnormal thermal field; When the model determines that there is a thermal field anomaly and there is no corresponding temperature rise warning at the deployed measuring points, the redundant alarm logic is triggered, a temperature anomaly alarm is issued, and the estimated location of the abnormal hot area is recorded.

[0019] To achieve real-time monitoring of key heat-generating areas within the grid cabinet, the first-class temperature sensors must be placed in close proximity to the heat source, ensuring comprehensive coverage and avoiding blind spots. Their placement should be based on thermal simulation analysis or empirical thermal field distribution results to ensure effective monitoring of localized temperature rises caused by poor contact, overload, and electrical aging.

[0020] Electrical nodes refer to locations in the grid cabinet that are prone to heat, aging, or high current density, including but not limited to: Circuit breaker terminals: These terminals have high contact resistance and carry current for long periods of time, making them common hot spots. The sensor should be installed outside the terminal housing, near the contact points.

[0021] Knife switch connection part: Frequent operation and wear of the connector may cause local heat, so the sensor should be close to the conductive connector.

[0022] Busbar junctions and joints: They carry the total current and should be placed at key points, especially where multiple busbars are connected in parallel.

[0023] The connection between the cable lug and the copper busbar: This is where local hot spots are most likely to occur due to poor crimping. The sensor can adopt an enveloping mounting structure.

[0024] DC bus input and AC output terminals: As the system input / output interface, loose cables or oxidized contacts are common.

[0025] Ground bus: It may heat up instantaneously in the event of a ground fault and should also be monitored.

[0026] To meet the requirements of high voltage, high magnetic field, high temperature and vibration in electrical environment, the following types of sensors are preferred: NTC thermistor / RTD (such as PT100): suitable for surface mount installation and fast response speed.

[0027] Thermocouple (such as K type): suitable for high temperature areas and flexible installation.

[0028] Infrared non-contact sensors (such as IR temperature sensing modules): Suitable for inconvenient or electrical areas to avoid safety hazards.

[0029] Flexible sensors: For installation in curved or space-constrained locations, such as cable elbows.

[0030] The installation method should ensure reliable contact and avoid vibration and falling off. Double-sided thermal conductive adhesive, clip fixation or special clamping structure should be used. High-temperature and pressure-resistant cables should be used for wiring and uniformly lead to the temperature acquisition module.

[0031] All first-category sensors must be connected to a unified data acquisition module, which should support the following functions: Sampling frequency: generally set between 1 second and 10 seconds to ensure that the temperature rise dynamics can be captured; Multi-channel synchronous sampling: avoid timing deviation caused by sampling delay; Analog / digital mixed acquisition: supports the coexistence of analog sensors and intelligent digital temperature sensors; The module should have anti-interference capability and be able to access RS485, CAN or industrial Ethernet communications.

[0032] After the installation is complete, the following verifications should be performed: Thermal response test: Verify temperature perception sensitivity by artificially heating local nodes (such as blowing hot air); Installation strength check: ensure that the sensor is well fixed under vibration or cabinet door opening and closing; Regular inspection: The system should support regular self-inspection function to determine whether the sensor is out of position or has failed.

[0033] By accurately positioning, rationally selecting, and standardizedly installing the first-class temperature sensor layout strategy at multiple electrical nodes inside the photovoltaic grid-connected cabinet, not only is comprehensive coverage of key heating points achieved, but high-quality basic data support is also provided for subsequent thermal field modeling and abnormality analysis, significantly improving the system's monitoring accuracy and safety assurance capabilities.

[0034] The second type of ambient temperature sensor not only monitors the overall thermal environment within the grid cabinet but, more importantly, detects localized air heat conduction anomalies, providing data for identifying "thermal field excursions." By monitoring temperature trends within the non-contact area, it helps identify potential heat sources not detected by the first type of sensor.

[0035] The installation location should avoid high-heat generating components and be placed in a relatively open area with free air flow, so as to sense the overall characteristics of the temperature field inside the cabinet that changes with time and the air duct structure. Typical installation areas include: Upper part of the cabinet near the ventilation openings: Hot air rises and concentrates, making it suitable for monitoring heat accumulation or abnormal heat dissipation.

[0036] The central axis area at the bottom of the cabinet: senses temperature gradients and the phenomenon of cold at the bottom and hot at the top.

[0037] The side of the cabinet near the door panel or air duct outlet: used to determine the horizontal convection condition of the cabinet.

[0038] Transition space between the busbar layer and the distribution layer: monitor the dynamics of heat conduction in the vertical direction.

[0039] Depending on the cabinet volume and ventilation structure, 2-4 environmental sensing points are generally set to ensure that the overall heat flow trend can be reflected.

[0040] The second type of sensor focuses on the ability to respond to slow changes in ambient temperature, and the following sensors are usually used: High-precision digital temperature sensor (such as DS18B20, TMP117): convenient for multi-point networking and strong stability.

[0041] Air thermal sensor (NTC / PTC): small size, low cost, suitable for dense deployment.

[0042] Integrated MEMS environmental sensing module: supports temperature, humidity, and even airflow detection for more complex thermal field analysis.

[0043] The performance requirements are as follows: Temperature measurement range: -40℃~+125℃, meeting the requirements of high-temperature cabinet scenarios; Response time: ≤2 seconds, ensuring sensitivity to local temperature fluctuations; Accuracy error: ±0.2℃ or better, easy for comparative analysis; It has strong anti-electromagnetic interference capability and is suitable for high-voltage power distribution environments.

[0044] The sensor should be installed using a hanging or bracket method, suspended in the air rather than attached to any surface to avoid the influence of heat conduction; All sensor cables use shielded cables or twisted pairs to prevent signal distortion; If wireless sensors are used, it is necessary to verify signal stability and plan a reasonable mesh communication structure.

[0045] The sampling period is set slightly lower than that of the first type of sensor (e.g. once every 5 seconds), which is sufficient to capture the overall temperature field change trend; The data is uploaded to the central monitoring system synchronously with the first type of sensors; It is used to construct a heat flow distribution map, combining time series and spatial coordinates to assist in determining whether there is a situation where the heat source is "not directly detected".

[0046] When the system detects: a certain environmental sensor point continues to heat up but the surrounding first-class sensors do not alarm; the temperature difference between the high and low positions of the cabinet exceeds the set threshold, showing an abnormal gradient; abnormal heat conduction occurs in the horizontal convection path; it is preliminarily judged that there may be a heat source position offset or sensor blind spot abnormality, and further analysis or operation and maintenance warning is triggered.

[0047] The second type of ambient temperature sensor breaks the traditional monitoring limitation of "only measuring key nodes." By dynamically sensing the heat flow within the cabinet, it enables intelligent identification of heat field migration, heat accumulation, and hidden heat sources. It complements the first type of sensor, forming a highly sensitive, distributed, and three-dimensional monitoring system for thermal anomalies, significantly enhancing the intrinsic safety of photovoltaic grid-connected cabinets.

[0048] In order to fully understand the distribution and dynamic changes of heat inside photovoltaic grid-connected cabinets, the present invention first constructs a continuous and updateable heat distribution model, which specifically includes the following steps: The system collects real-time temperature data from multiple first-class temperature sensors and second-class ambient temperature sensors. Because sensors may come from different models and have varying sampling accuracy and response times, data normalization is required during the initial processing phase. This process includes unit conversion, sampling period alignment, and missing value interpolation. The temperature variation over time at each measurement point is constructed as a one-dimensional time series. Multiple measurement points form a two-dimensional time series matrix, which serves as the foundation for subsequent modeling.

[0049] Each sensor's physical coordinates (X, Y, and Z axis positions) are recorded upon installation. The system uses this information to construct a three-dimensional spatial mapping model of the cabinet interior. The real-time temperature values at each measurement point are projected onto the cabinet's structural model to form a point cloud of the temperature spatial distribution.

[0050] Considering that sensors are not fully distributed, the model uses bilinear interpolation or weighted average interpolation to estimate the temperature of unmeasured areas. The weighting method can be adjusted using an inverse function of the distance to the measurement point, gradually reducing the weight of the temperature estimate in areas farther away from the measurement point to avoid edge errors.

[0051] The thermal distribution field is recorded at fixed intervals (e.g., every 10 seconds) and the consecutive frames are combined into a frame sequence to form a dynamic thermal distribution graph. This model not only reflects the current temperature distribution state but also shows the heat propagation path and changing trends, forming the core foundation for thermal anomaly analysis.

[0052] Based on the established dynamic thermal model, to accurately identify abnormal temperature rise and thermal field excursions, it is necessary to further analyze the temperature difference and temperature change rate between each measuring point. The following are the detailed technical steps of this process.

[0053] In traditional methods, the instantaneous temperature difference is usually obtained by simply subtracting the temperatures of two adjacent points. This method is susceptible to noise interference and is insensitive to sudden abnormalities. Therefore, the present invention introduces a multi-scale wavelet transform method to perform difference enhancement calculation. The main steps are as follows: Randomly select two physically adjacent measuring points, collect and organize their temperature values within a time window (for example, 10 seconds) to form two continuous time series curves.

[0054] A multiscale wavelet transform was performed on each of the two time series, decomposing them into components at several scales, each representing temperature fluctuations at a different frequency. High-frequency components reflect sudden temperature rises, while low-frequency components represent steady trends, effectively removing noise.

[0055] The mid- and high-frequency components most relevant to the fault characteristics are selected from the multi-scale components of each measuring point, and the curve is reconstructed using inverse transformation to form a denoised and enhanced temperature change curve.

[0056] At each time point, the temperature difference between the two reconstructed curves is calculated to obtain a more physically meaningful series of instantaneous temperature differences. This difference series truly reflects the changes in heat transfer efficiency and heat source distribution between measurement points.

[0057] The current difference sequence is compared with the historical steady-state temperature difference model built based on long-term operation. If the difference exceeds the preset fluctuation threshold (this threshold can be dynamically adjusted based on the measurement point location, cabinet structure, etc.), the area between the measurement point pair is marked as an "abnormal temperature difference area."

[0058] This paper introduces a combination of empirical mode decomposition (EMD) and fuzzy C-means clustering (FCM) to obtain the abnormal temperature rise index of each measuring point within the current time window. The calculation steps are as follows: A time window (e.g., 30 seconds) is set for each measuring point, and the temperature data of this section is collected to form a continuous time series, and an analysis window sequence is formed with a sliding step of 5 seconds.

[0059] By performing EMD decomposition on each temperature series, we obtain several intrinsic mode functions (IMFs) with local time scale significance. These IMFs represent short-term fluctuations, medium-term trends, and long-term background changes.

[0060] Three key features related to warming are extracted from the medium and high frequency IMFs: Local slope change trend: indicates the speed of temperature increase per unit time; Slope fluctuation amplitude: judge whether the heating process is stable or jumpy; Length of the warming period: reflects whether the warming is a trend or sporadic.

[0061] The combination of the three features constitutes the “temperature rise behavior feature vector” of each measuring point.

[0062] The feature vectors of all measurement points are input into the FCM algorithm for cluster analysis. This algorithm allows measurement points to belong to multiple cluster centers simultaneously. By calculating their membership in the "abnormal temperature rise category," each measurement point's temperature rise anomaly index is obtained. This index reflects the degree to which the current temperature rise behavior at the measurement point deviates from normal, and the value range is generally between 0 and 1.

[0063] If the abnormal temperature rise index of a certain measuring point exceeds the critical threshold set by the system (for example, 0.7), it will be marked as an "abnormal temperature rise trend point" and enter the next stage of spatial anomaly analysis.

[0064] Combined with the identified "abnormal temperature difference areas" and "abnormal temperature rise trend points", further spatial logical judgment is performed to confirm whether there is a thermal field offset phenomenon: The physical coordinates of the measuring points in the three-dimensional model are used to determine the spatial distance between the abnormal temperature rise point and its adjacent abnormal temperature difference area. If the offset exceeds 1.5 times the average spacing between the measuring points, it is determined that "thermal field offset" exists.

[0065] Further comparison of the current heat distribution map with the historical normal heat map is conducted. If features such as displacement of the center of gravity of the hotspot distribution, change in path, and asymmetric expansion are found, it will be identified as "structural heat conduction anomaly."

[0066] Once thermal field deviation or conduction anomaly is confirmed, the system will automatically trigger the alarm mechanism and highlight the deviation area through the visual interface to assist operation and maintenance personnel in quickly locating potential hidden dangers.

[0067] Compared to traditional temperature monitoring systems that rely solely on absolute temperature or fixed threshold alarms, this invention achieves higher monitoring intelligence and fault prediction capabilities through the following technological innovations: Introducing a dynamic heat distribution model to achieve joint modeling of temperature in spatial and temporal dimensions; Use wavelet transform to accurately extract instantaneous temperature difference features and enhance sensitivity to abnormal heat sources; Combining EMD and fuzzy clustering techniques, a heating rate anomaly index is constructed to quantify potential abnormal behaviors; Spatial topological relationships are used to identify thermal field offsets and solve the problem of identifying sensor blind spots or atypical thermal anomalies.

[0068] During initial deployment or scheduled maintenance, the system collects long-term samples of the thermal distribution of the PV grid cabinet under various typical operating conditions, such as varying power generation loads, ambient temperatures, or seasonal conditions. During each sampling session, the temperature values of all temperature sensors are recorded and a thermal profile is generated based on these values at that moment.

[0069] In order to establish a representative and comparable thermal distribution reference library, the system will classify these historical thermal images according to the corresponding working conditions and perform standardization on each type of image, including but not limited to: Unified alignment of spatial dimensions: enables the coordinates of measurement points of different thermal maps to be directly mapped in the same three-dimensional structure; Temperature normalization: compress the temperature value to between 0 and 1 to facilitate subsequent image similarity comparison; Abnormal elimination and averaging: eliminate occasional abnormal temperature rise points in historical data and use sliding average or median filtering to construct a "steady-state" template diagram.

[0070] Finally, a thermal field benchmark library is formed, in which each type of working condition corresponds to one or more standard thermal distribution templates, which serve as a reference for the current thermal field comparison.

[0071] During operation, the system collects temperature data from all current measurement points in real time and constructs a thermal distribution map at the current time point. In order to enable direct comparison of this map with the historical benchmark template, the current thermal map must be standardized. This process includes: Sizing: Project the thermal data of the current measurement point into the same three-dimensional spatial model as the historical thermal map through the coordinate mapping algorithm; Temperature range remapping: linearly map the temperature values of the current heat map according to the temperature range of the historical baseline map to ensure the consistency of the temperature difference ratio during the comparison process; Symmetry check: Determine whether the current heat distribution map has a geometric structure similar to the historical template, such as whether the high-temperature area is symmetrically distributed along the center of the busbar and whether it is concentrated near the terminal blocks.

[0072] At the same time, several representative features are extracted from the current heat map for comparative analysis, mainly including the following: Thermal center coordinates: the location of the highest temperature point in the current thermal map; Temperature variance and range: used to measure the uniformity or concentration of the overall heat distribution; Gradient directionality between measuring points: indicates the trend of temperature change with spatial distribution; Heat diffusion path morphology: Evaluates the geometric characteristics of heat diffusion from high temperature areas to the outside.

[0073] The comparison stage is the core link of the present invention's intelligent judgment of thermal anomalies. The system introduces a variety of image analysis and clustering recognition technologies to gradually complete the anomaly recognition process: First, the system uses a structural similarity metric to compare the current heatmap with the historical template image. This algorithm comprehensively evaluates the similarity between the two heatmaps based on local brightness, contrast, and texture structure, and outputs a similarity score. This score typically ranges from 0 to 1, with higher values indicating greater similarity between the current heatmap and the historical template.

[0074] When the similarity between the current thermal map and all historical template maps is lower than the preset threshold (for example, less than 0.75), the system preliminarily determines that the current thermal distribution pattern deviates from the normal thermal state and there is a suspected thermal field anomaly.

[0075] To further verify the accuracy of the judgment, the system also compares the characteristic vectors of the current heat map (such as the coordinates of the heat center, variance, and gradient direction) with the historical template features. This comparison uses principal component analysis to reduce the multidimensional heat distribution features to principal feature components. The system then uses the Mahalanobis distance (a distance metric that considers the correlation between feature variables) to determine whether the current heat map deviates significantly from the historical template.

[0076] If the multidimensional distance exceeds the statistical tolerance interval when the historical template is constructed, it can be confirmed that the current heat map shows abnormal thermal distribution behavior.

[0077] In addition to static comparisons of single heat maps, the system also incorporates a time-series-based heat map evolution model, using dynamic time warping technology to compare the degree of deviation between the current heat map's trajectory and its historical trajectory. If the current heat map's evolution path deviates from the historical stable pattern and persists for a period of time without returning to normal, it is considered a trending thermal field excursion.

[0078] To achieve intelligent response and risk classification management, the system divides the thermal anomaly levels according to the above comparison results: Level 1 warning: The current heat map has a slight deviation from the historical template similarity, and the Mahalanobis distance is close to the upper threshold. A warning is issued but no action is triggered. Secondary anomaly: The thermal center coordinates are significantly offset, the thermal gradient direction is reversed, or there is an irregular diffusion pattern. The system records and prompts for inspection. Level 3 emergency: The heat map structure changes significantly and the trend deviation persists. The system will trigger shutdown protection or remote alarm.

[0079] When the system continuously detects that a certain heat distribution pattern is abnormal but does not cause a fault, or gradually becomes the new normal with equipment aging and seasonal changes, the system can incorporate the heat map sample into the model training set and update the thermal field benchmark library, thereby realizing the "adaptive evolution" of the intelligent monitoring system.

[0080] The thermal field anomaly identification method based on thermal distribution map comparison proposed in this invention breaks the limitations of traditional temperature threshold judgment. Through technical means such as image similarity analysis, multi-dimensional feature distance judgment and time trajectory modeling, it realizes the visualization, intelligent and hierarchical responsive identification of thermal anomalies, greatly improving the fault prediction capability and system operation and maintenance efficiency of photovoltaic grid-connected cabinets, and has significant novelty and practicality.

[0081] The temperature monitoring system of this invention features dynamic thermal distribution modeling. After collecting data from all measurement points, it generates a real-time thermal map of the cabinet interior. This thermal map is a spatial distribution diagram. The temperature values of each measurement point are projected onto the cabinet structure model using a three-dimensional coordinate system. An interpolation algorithm is then used to estimate the temperature of unmeasured areas, ultimately forming a continuous, realistic temperature field image that reflects the thermal state of the cabinet.

[0082] After comparing the current heat map with the historical stable heat map pattern (or statistical heat template), the model determines that there is a thermal anomaly. That is, the structural similarity between the current heat map and any historical normal heat map is lower than the set threshold, or the thermal center of gravity position shows significant drift. In this case, a thermal anomaly is determined.

[0083] The system then extracts the spatial features of the hotspot area from the anomaly heat map, including but not limited to: Thermal center coordinates (the spatial location of the hottest point in the thermal map); Thermal anomaly boundaries (envelope of areas with abrupt temperature gradient changes); The distribution pattern of thermal anomalies (e.g., whether they develop along one side of the busbar or occur locally); Heat diffusion direction (heat flow trend is determined based on the gradient path of local high temperature points).

[0084] The above information is integrated to form a thermal field abnormality response diagram, which serves as the graphical basis for subsequent alarms and judgments.

[0085] Thermal map anomalies are not equivalent to triggering system alarms. The system needs to further determine whether such thermal anomalies are covered by existing measurement points.

[0086] The system includes a built-in sensing area model for each temperature sensor. This model, centered around the sensor's mounting point, defines a radius or spatial cube based on its sensitivity and thermal conductivity, representing its effective monitoring range. Typically, this range is the local direct contact area or a nearby heat diffusion zone within 5–10 cm.

[0087] This abnormal response map is then overlaid with the sensing area models of all sensors for spatial comparison. If the thermal anomaly area falls completely outside the sensing range, or if it is partially covered but the corresponding sensor does not record sufficient temperature rise (i.e., the current temperature or temperature rise rate does not reach the alarm threshold), the system determines that area as a temperature warning blind spot.

[0088] This blind spot may be caused by the following reasons: Insufficient distribution density, especially in structural corners or multi-layer wiring areas; The cabinet ventilation or heat conduction structure changes, causing heat transfer to deviate from the original design path; A measuring point is damaged or has decreased in accuracy and is unable to respond normally to thermal changes.

[0089] When it is confirmed that there is a thermal field anomaly and the anomaly has not been triggered by the current distribution point alarm, the system no longer relies on the logical judgment of the measurement point, but takes the model as the core basis to actively trigger the redundant alarm mechanism.

[0090] The activation criteria for this mechanism are as follows: The structural similarity between the current heat map and the historical heat map is less than the set value (such as 0.75); The temperature in the center of the abnormal area deviates significantly from the historical average (for example, more than 20% higher); The abnormal area has no effective intersection with the sensing areas of all measuring points, or the sensors in the intersection area do not trigger an alarm.

[0091] When any two of the above conditions are met, the system will bypass the regular measurement point alarm rules and directly issue a temperature anomaly warning.

[0092] These alarms can be divided into two categories: Model-driven one-time alarm: used for sudden anomaly identification; Trend-type enhanced alarm: When the abnormal area persists in the blind spot and shows a continuous temperature increase trend, it will be upgraded to a continuous alarm and can even trigger the system load limit or power off protection logic.

[0093] After the alarm is triggered, the system will immediately record the spatial coordinates, thermal center point, abnormal area estimation, expansion direction trend and other information of the abnormal area into the system log to form an event-level thermal anomaly file.

[0094] This information can be used to: On-site inspection and target positioning by operation and maintenance personnel; Subsequent thermal model training sample accumulation and abnormal pattern learning; Check sensor blind spots to provide data support for the next hardware layout optimization; Determine whether it is an intermittent fault, local aging, or special circumstances such as environmental thermal interference.

[0095] The system can also be configured with a thermal map snapshot storage mechanism to capture a thermal map snapshot when an alarm occurs and perform a visual comparison with the historical map to improve the efficiency and accuracy of subsequent analysis.

[0096] In summary, this implementation successfully constructs a temperature anomaly identification method that surpasses traditional threshold judgment methods through the combined logic of "model-driven + blind-spot identification + redundant judgment mechanism." This system is capable of identifying thermal anomalies and accurately inferring their locations without sensor alarms, significantly enhancing the intelligent capabilities of the grid-connected cabinet system in terms of anomaly identification, equipment protection, and fault traceability. Its independent judgment logic, model-driven spatial deduction mechanism, and redundant alarm design demonstrate the invention's originality and practical value.

[0097] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0098] It should be understood that the term "and / or" herein is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone, where A and B may be singular or plural. In addition, the character " / " herein generally indicates that the objects associated with each other are in an "or" relationship, but it may also indicate an "and / or" relationship, which can be understood by referring to the context. A person of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0099] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for monitoring the temperature of a photovoltaic grid-connected cabinet, characterized by: include: First-class temperature sensors are installed at multiple electrical nodes inside the photovoltaic grid-connected cabinet and their surrounding areas to collect real-time temperature data; A second type of ambient temperature sensor is installed in the open area of the cabinet to collect data on thermal field changes caused by local air flow; A dynamic thermal distribution model is constructed using the collected temperature data. The model identifies potential thermal field excursions by analyzing the temperature differences, change rates, and spatial position relationships of multiple measurement points. Based on the thermal distribution model, the current temperature data is compared with the thermal field pattern under historical normal conditions to determine whether there is any abnormal thermal field; When the model determines that there is a thermal field anomaly and there is no corresponding temperature rise warning at the deployed measuring points, the redundant alarm logic is triggered, a temperature anomaly alarm is issued, and the estimated location of the abnormal hot area is recorded.

2. A temperature monitoring method for a photovoltaic grid-connected cabinet according to claim 1, characterized in that: The construction of dynamic heat distribution model includes: Normalize the data collected by each temperature sensor at different time points, unify the unit and sampling period, and establish a time series matrix; Based on the physical layout coordinates of the sensors, a spatial distribution model is established to map each temperature point to the corresponding position of the cabinet's three-dimensional structural model; By using bilinear interpolation or weighted average method, a continuous temperature field distribution map is constructed to form the heat distribution profile inside the cabinet; Generate a thermal distribution frame sequence at set time intervals to form a thermal field dynamic change model.

3. A temperature monitoring method for a photovoltaic grid-connected cabinet according to claim 2, characterized in that: The analysis of temperature differences, change rates, and spatial position relationships at multiple measuring points includes: Calculate the instantaneous temperature difference between any two adjacent measuring points and compare it with the historical steady-state temperature difference benchmark. If it exceeds the preset fluctuation threshold, it will be marked as an abnormal temperature difference area; After analyzing the temperature change rate of each measuring point in a continuous time period, the abnormal temperature rise index is calculated. If the abnormal temperature rise index is greater than the critical rate threshold, it is identified as an abnormal temperature rise trend point; A position correlation analysis is conducted on the abnormal temperature difference areas and abnormal warming trend points in the spatial model. If the high temperature point and the temperature difference boundary exceed 1.5 times the average spacing between the measuring points, it is inferred that there is a thermal field offset. After detecting the thermal field deviation phenomenon, by comparing the current thermal distribution map with the thermal field pattern under historical stable working conditions, it is confirmed whether it is a structural heat conduction anomaly and trigger an abnormality warning.

4. A temperature monitoring method for a photovoltaic grid-connected cabinet according to claim 3, characterized in that: The calculation method of the instantaneous temperature difference is as follows: the temperature values collected by any two physically adjacent measuring points within a preset sampling period are organized into a time series, and two continuous time-temperature trajectories are constructed respectively. Multi-scale wavelet transform processing is performed on the temperature time series of the two measuring points respectively. By selecting the wavelet coefficient level, the temperature time series of each measuring point is inversely transformed and reconstructed to obtain two sets of denoised and enhanced temperature change curves. The difference between the reconstructed temperature curves of the two measuring points is calculated at each time point to obtain the instantaneous temperature difference.

5. A temperature monitoring method for a photovoltaic grid-connected cabinet according to claim 4, characterized in that: The calculation method of the heating rate anomaly index is as follows: for each temperature measuring point, the temperature change data of the set time window is collected to construct a continuous time series; the empirical mode decomposition operation is performed on the temperature time series of each measuring point to decompose the original temperature curve into several intrinsic mode functions (IMFs); the mode sensitive to heating is selected from multiple IMFs, and the local slope change trend, slope fluctuation amplitude and heating duration period are extracted according to its fluctuation trend to form the heating behavior feature vector of each measuring point in the current window; the heating feature vectors of all measuring points are input into the fuzzy C-means clustering algorithm for classification, and each measuring point is assigned a membership value belonging to the abnormal heating class, namely, the heating rate anomaly index.

6. The temperature monitoring method of a photovoltaic grid-connected cabinet according to claim 1, characterized in that: Compare the current temperature data with the thermal field pattern under historical normal conditions to determine whether there are any abnormal thermal conditions, including: Collect stable operating temperature data of photovoltaic grid-connected cabinets under multiple states, generate multiple sets of thermal distribution maps, and establish a standardized historical thermal field benchmark library through averaging and outlier elimination technology. Each type of thermal field mode corresponds to a typical distribution template; Normalize the thermal distribution map at the current moment to make it consistent with the historical thermal field template in size, scale and sampling frequency; The structural similarity measurement algorithm is used to compare the current thermal map with each historical thermal field template one by one, and the similarity scores in terms of overall structure, thermal center position and temperature difference distribution are calculated; If the similarity between the current heat map and all historical templates is lower than the set threshold, it is judged as a thermal field anomaly; the anomaly level is further divided according to the degree of similarity deviation, including slight offset and major heat source reconstruction, and the corresponding alarm strategy is triggered.

7. A temperature monitoring method for a photovoltaic grid-connected cabinet according to claim 6, characterized in that: Thermal field image similarity metrics include: Extract the characteristic vector of the thermal distribution map, including the coordinates of the thermal center, the maximum temperature difference, the directionality of the thermal gradient between each measuring point, and the temperature mean and variance; The principal component analysis method is used to compress the dimensions of historical and current feature vectors and extract the most discriminative thermal distribution feature axis; The multidimensional distance between the current thermal map and the main axis of each historical thermal field is calculated using the Mahalanobis distance algorithm. If the distance exceeds the specified threshold, it is determined that the current thermal field pattern has deviated from the normal operating state and is marked as an abnormal thermal distribution state.

8. The temperature monitoring method of a photovoltaic grid-connected cabinet according to claim 1, characterized in that: Triggering redundant alarm logic includes: After determining that the thermal field is abnormal, the characteristic area of the abnormal heat zone is marked in the current thermal distribution map, and a thermal field response map covering the abnormal area is generated; By comparing the location of the abnormal area in the thermal field response diagram with the sensing range of the currently deployed temperature sensors, a spatial overlap detection algorithm is used to determine whether there is a thermal anomaly area that is not monitored by the sensor. If the thermal anomaly area is outside the radial sensing range of all measuring points, or the temperature change does not form an effective gradient at the sensor, it is identified as a warning blind spot. When it is confirmed that a thermal field anomaly exists and there is no warning at the measured point, the redundant alarm logic is activated to trigger a temperature anomaly warning.

Images