Intelligent vacuum monitoring method and system for vacuum heat collector
By adopting a two-level monitoring system of space-performance joint modeling and dynamic adaptive threshold, the problems of difficult fault location, rigid threshold and lack of prediction in vacuum collectors are solved, achieving high-precision fault location and prediction and reducing operation and maintenance costs.
Patent Information
- Application Number
- CN202511750289.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-26
- Publication Date
- 2026-02-27
AI Technical Summary
Existing technologies cannot accurately distinguish between real faults and accidental fluctuations, making it difficult to quickly locate specific faulty components. Traditional static thresholds cannot adapt to environmental fluctuations, leading to increased downtime losses and an inability to predict latent degradation in post-incident maintenance.
Employing a spatial-performance joint modeling, dynamic adaptive threshold, and a two-level monitoring system, the system performs real-time anomaly monitoring through multi-level grid division and the DBSCAN algorithm. Combined with a predictive model, it predicts health status, achieving meter-level positioning accuracy and dynamic adaptive threshold. It also integrates temporal patterns with environmental coupling for prediction.
It achieves meter-level positioning accuracy for vacuum collectors, reduces false alarms and missed alarms, lowers operation and maintenance costs through a graded handling mechanism, enables the prediction of equipment degradation trends, and improves the accuracy and predictive ability of fault location.
Smart Images
Figure CN121577366A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of solar collector monitoring, and in particular to an intelligent vacuum monitoring method and system for vacuum solar collectors. Background Technology
[0002] Conventional monitoring relies solely on data from a single sensor to determine equipment status, failing to distinguish between genuine faults and occasional fluctuations, and making it difficult to quickly pinpoint specific faulty components (e.g., which component of which collector). Traditional static thresholds (e.g., alarm when vacuum level < XPA) cannot adapt to environmental fluctuations (e.g., legitimate inefficiency during sudden drops in irradiance on cloudy or rainy days). Analyzing each device independently ignores group statistical patterns and is susceptible to interference from individual outliers. Reactive maintenance methods result in downtime losses and cannot predict latent degradation (e.g., slow aging of seals).
[0003] Solar irradiance, ambient temperature, and other factors are strongly correlated with collector performance. An excessive value for a single parameter may be due to external environmental factors rather than equipment malfunction. For example, high temperatures under high irradiance are normal, but a significantly lower temperature difference compared to neighboring equipment under the same irradiance is considered a malfunction. Summary of the Invention
[0004] The purpose of this invention is to provide an intelligent vacuum monitoring method and system for vacuum collectors. Through innovative designs such as space-performance joint modeling, dynamic adaptive threshold, and two-level monitoring system, it systematically solves the three core pain points that have long existed in the new energy field: "difficulty in fault location, rigid threshold, and lack of prediction". It has significant industry demonstration significance.
[0005] To achieve the above objectives, the present invention provides an intelligent vacuum monitoring method for vacuum collectors, comprising the following steps: Obtain the actual distribution data of vacuum collectors in the area, and model the area based on the actual distribution data of vacuum collectors in the area to obtain the area distribution map; The system collects performance parameters and environmental parameters of the vacuum collector. The performance parameters include vacuum level, flow rate, collector tube temperature, collector temperature, and insulation box temperature. The environmental parameters include solar irradiance and irradiance time. The area distribution map is divided using a multi-level grid division method. Distributed monitoring is then performed on the divided multi-level grids to obtain the corresponding environmental and performance parameters. A three-level coding and positioning method is used to code and locate each level of the grid. Two levels of monitoring are set up based on environmental and performance parameters: Level 1 monitoring is real-time anomaly monitoring, and Level 2 monitoring uses a predictive model to predict the health of the vacuum collector. The monitoring status of each vacuum collector is located using the coded location as an index.
[0006] Preferably, a multi-level grid division method is used to divide the area distribution map, including area level, collector level and component level, and the grid is divided level by level.
[0007] Preferably, real-time anomaly monitoring includes the following steps: Clustering features within the same time window were selected, including core performance features and spatial features; the core performance features included efficiency ratio, normalized heat and appearance temperature difference ratio, and the spatial features included three-level coded coordinates. Standardize the clustering characteristics of all solar collectors within the current time period; The DBSCAN algorithm is used for cluster analysis to obtain core clusters, edge points, and outliers. Based on the cluster analysis results, the individual devices and devices in adjacent grids are compared to obtain real-time monitoring results.
[0008] Preferably, based on the cluster analysis results, a comparison is made between a single device and devices in adjacent grids to obtain real-time monitoring results, including... From the core cluster obtained by spatial clustering, a batch of solar collectors are randomly selected, and the average value of their performance parameters is calculated to construct a baseline time series. The dynamic judgment matrix is set as: baseline time series ± α × standard deviation matrix; α represents the dynamic coefficient. Based on the dynamic judgment matrix, the time series data of the collectors in the edge points and outliers are compared with the dynamic judgment matrix. When the comparison result is abnormal, the problematic collector is located through three-level coding. Core clusters of the same type and from the same region with similar installation angles were selected as a reference group; Calculate the parameter differences ΔP and ΔT between the target point and the reference group, and use t-test to verify the differences.
[0009] Preferably, secondary monitoring involves using a predictive model to predict the health of the vacuum collector, including the following steps: Feature extraction is performed on performance parameters and environmental sampling parameters to obtain time-series features and environmental coupling features; the time-series features include statistical features, trend features, frequency domain features, and slope features; the environmental coupling features include irradiance correction factor, temperature gradient ratio, cross-combination features, and hysteresis effect features; By inputting time-series features and environmental coupling features into the time-series prediction model, and analyzing the correlation between time-series features and environmental coupling features, the predicted health status of the solar collector can be obtained.
[0010] Preferably, the loss of the time series prediction model is: ; In the formula, Indicates the total loss. Indicates MSE loss. This indicates a loss of trend consistency.
[0011] A smart vacuum monitoring system for vacuum collectors includes: The modeling module is used to obtain the actual distribution data of vacuum collectors in the area, and to model the area based on the actual distribution data of vacuum collectors in the area to obtain the area distribution map. The data acquisition module is used to collect performance parameters and environmental parameters of the vacuum collector. The performance parameters include vacuum level, flow rate, collector tube temperature, collector temperature, and insulation box temperature; the environmental parameters include solar irradiance and irradiance time. The grid division module is used to divide the area distribution map using a multi-level grid division method, perform distributed monitoring on the divided multi-level grids, obtain the corresponding environmental and performance parameters, and use a three-level coding and positioning method to code and locate each level of grid. The monitoring module is used to set two levels of monitoring based on environmental and performance parameters: the first level is real-time anomaly monitoring, and the second level is to use a predictive model to predict the health of the vacuum collector; the monitoring status of each vacuum collector is located by using the code location as an index.
[0012] Therefore, the present invention employs the above-mentioned intelligent vacuum monitoring method and system for vacuum collectors, and the technical effects are as follows: Spatial-performance joint modeling: meter-level positioning accuracy is achieved through three-level coding, and DBSCAN is used to discover group anomaly patterns.
[0013] Dynamic adaptive threshold: Set differentiated α coefficients for different parameter characteristics to avoid false alarms and missed alarms.
[0014] Multimodal prediction: Integrating temporal patterns with environmental coupling to predict equipment degradation trends in advance.
[0015] Tiered response mechanism: Level 1 provides rapid response to emergency faults, while Level 2 provides guidance on preventative maintenance to reduce operation and maintenance costs. Attached Figure Description
[0016] Figure 1 This is a flowchart of an intelligent vacuum monitoring method and system for vacuum collectors according to the present invention. Detailed Implementation
[0017] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.
[0018] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0019] Example 1 like Figure 1As shown, an intelligent vacuum monitoring method for a vacuum collector includes the following steps: Obtain the actual distribution data of vacuum collectors in the area, and model the area based on the actual distribution data of vacuum collectors in the area to obtain the area distribution map; The system collects performance parameters and environmental parameters of the vacuum collector. The performance parameters include vacuum level, flow rate, collector tube temperature, collector temperature, and insulation box temperature. The environmental parameters include solar irradiance and irradiance time. The area distribution map is divided using a multi-level grid division method. Distributed monitoring is then performed on the divided multi-level grids to obtain the corresponding environmental and performance parameters. A three-level coding and positioning method is used to code and locate each level of the grid. The distribution map of the area is divided using a multi-level grid division method, including the area level, collector level and module level, which are divided into grids in sequence. Two levels of monitoring are set up based on environmental and performance parameters: Level 1 monitoring is real-time anomaly monitoring, and Level 2 monitoring uses a predictive model to predict the health of the vacuum collector. Real-time anomaly monitoring includes the following steps: Clustering features within the same time window were selected, including core performance features and spatial features; the core performance features included efficiency ratio, normalized heat and appearance temperature difference ratio, and the spatial features included three-level coded coordinates. Standardize the clustering characteristics of all solar collectors within the current time period; Cluster analysis was performed using the DBSCAN algorithm to obtain core clusters, edge points, and outliers. Based on the cluster analysis results, individual devices were compared with devices in adjacent grids to obtain real-time monitoring results, including... From the core cluster obtained by spatial clustering, a batch of solar collectors are randomly selected, and the average value of their performance parameters is calculated to construct a baseline time series. The dynamic judgment matrix is set as: baseline time series ± α × standard deviation matrix; α represents the dynamic coefficient; vacuum degree: α = 2.0 (strict monitoring), temperature: α = 2.5 (allowing large fluctuations), flow rate: α = 1.5 (relatively stable).
[0020] Specifically, the parameters of the DBSCAN (density-based spatial clustering application) algorithm are as follows: eps (neighborhood radius): set to 0.3, indicating that in the feature space, the maximum Euclidean distance between two sample points considered to be in the same neighborhood is 0.3. This value determines the cluster density; smaller values result in denser clusters. min_samples (minimum number of samples): set to 5, indicating the minimum number of neighborhood samples (including itself) required for a point to be defined as a core point. That is, if a point's neighborhood contains at least 5 points, then that point is considered a core point. metric (distance metric): set to 'euclidean', indicating that Euclidean distance is used to calculate the similarity between sample points, which is the most commonly used distance metric. algorithm (algorithm selection): set to 'auto', indicating that DBSCAN will automatically select the most suitable algorithm (such as ball tree or kd-tree) to efficiently calculate the neighborhood, based on the characteristics of the input data.
[0021] Based on the dynamic judgment matrix, the time series data of the collectors in the edge points and outliers are compared with the dynamic judgment matrix. When the comparison result is abnormal, the problematic collector is located through three-level coding. Core clusters of the same type and from the same region with similar installation angles were selected as a reference group; Calculate the parameter differences ΔP and ΔT between the target point and the reference group, and use t-test to verify the differences.
[0022] Secondary monitoring involves using predictive models to forecast the health of vacuum collectors, and includes the following steps: Feature extraction is performed on performance parameters and environmental sampling parameters to obtain time-series features and environmental coupling features; the time-series features include statistical features, trend features, frequency domain features, and slope features; the environmental coupling features include irradiance correction factor, temperature gradient ratio, cross-combination features, and hysteresis effect features; By inputting time-series features and environmental coupling features into the time-series prediction model, and analyzing the correlation between time-series features and environmental coupling features, the predicted health status of the solar collector can be obtained.
[0023] The loss of the time series prediction model is: ; In the formula, Indicates the total loss. Indicates MSE loss. This indicates a loss of trend consistency.
[0024] The monitoring status of each vacuum collector is located using the coded location as an index.
[0025] Example 2 Using the method in Example 1, a solar energy park contains 100 vacuum collectors distributed across three zones: A, B, and C. Each device has been coded at three levels: zone level, collector level, and module level (e.g., the 2nd module of collector number 5 in zone A is coded as A-05-02). Monitoring is required for the period from 14:00 to 16:00 on a given day.
[0026] Table 1 Data collected
[0027] Note: B-03-01 triggered an initial alarm due to the vacuum level being lower than the dynamic lower limit.
[0028] The three-level grid system is as follows: Area level: Areas A, B, and C are treated as independent subnets. Collector level: Each zone contains approximately 30 units managed as a single unit. Component level: Each device contains 4 components for fine-grained control. Example code parsing: B-03-01 = Component 1 of collector number 3 in zone B; Distributed monitoring deployment: Each area is equipped with an IoT gateway to collect all parameters at 5-minute intervals.
[0029] Edge computing nodes execute the DBSCAN clustering algorithm in real time.
[0030] Level 1 monitoring: Real-time anomaly detection Construct cluster feature vectors, selecting the following features from all devices within the current window (14:00~16:00): Core performance characteristics: efficiency ratio (η = heat production / irradiance), normalized heat output (Q) norm =Q / maxQ), surface temperature difference ratio (ΔT_surface / ΔT_ambient); Spatial features: three-level coded coordinates (e.g., A-01-01 is decomposed into ['A','01','01']).
[0031] Standardization and cluster analysis Z-score normalization is performed on all features, with the DBSCAN parameters configured as follows: eps=0.3 (strictly control neighborhood density), min_samples=5 (at least 5 neighboring points constitute the core point), metric='euclidean'+algorithm='auto' (automatically optimize the index structure).
[0032] The clustering results are determined as follows: Table 2. Clustering Results Judgment ;
[0033] Dynamic threshold verification Ten devices were randomly selected from the core cluster to calculate the baseline time series and standard deviation matrix.
[0034] Dynamic judgment matrix = baseline value ± α × σ: Vacuum degree: α=2.0, with the allowable deviation range expanded to ±2σ; Temperature: α=2.5, tolerating greater thermal inertia fluctuations; Flow rate: α=1.5, strictly ensuring fluid stability; B-03-01 Verification: Measured vacuum degree 6.5 × 10⁻⁶ -4 Pa < lower limit 6.8 × 10 -4 Pa, an anomaly has been confirmed.
[0035] Spatial correlation analysis Compare the device parameters of B-03-01 with its adjacent grids (B-02, B-04): ΔP (pressure difference) = -1.2 × 10 -4 Pa, ΔT (temperature difference) = -24℃ The t-test showed a pressure P < 0.01, indicating a significant difference, and was therefore determined to be a true positive fault.
[0036] Secondary monitoring: Health prediction Table 3 Feature Engineering
[0037] Model Inference Input feature vectors into the LSTM+Attention hybrid model, and combine the loss functions as follows: MSE Loss: Error in fitting the actual output; Trend Consistency Loss: Constraints the prediction of trends to be consistent with historical patterns; B-03-01 Prediction Result: Health level will drop to 62% in the next 24 hours (normal ≥80%), and it is recommended to shut down the machine for maintenance.
[0038] Table 4 Final Monitoring Report
[0039] Therefore, the present invention adopts the above-mentioned intelligent vacuum monitoring method and system for vacuum collectors. Through innovative designs such as space-performance joint modeling, dynamic adaptive threshold, and two-level monitoring system, it systematically solves the three core pain points that have long existed in the new energy field: "difficulty in fault location, rigid threshold, and lack of prediction". It has significant industry demonstration significance.
[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A smart vacuum monitoring method for vacuum collectors, characterized in that, Includes the following steps: Obtain the actual distribution data of vacuum collectors in the area, and model the area based on the actual distribution data of vacuum collectors in the area to obtain the area distribution map; The system collects performance parameters and environmental parameters of the vacuum collector. The performance parameters include vacuum level, flow rate, collector tube temperature, collector temperature, and insulation box temperature. The environmental parameters include solar irradiance and irradiance time. The area distribution map is divided using a multi-level grid division method. Distributed monitoring is then performed on the divided multi-level grids to obtain the corresponding environmental and performance parameters. A three-level coding and positioning method is used to code and locate each level of the grid. Two levels of monitoring are set up based on environmental and performance parameters: Level 1 monitoring is real-time anomaly monitoring, and Level 2 monitoring uses a predictive model to predict the health of the vacuum collector. The monitoring status of each vacuum collector is located using the coded location as an index.
2. The intelligent vacuum monitoring method for a vacuum collector according to claim 1, characterized in that, A multi-level grid division method is used to divide the area distribution map, including area level, collector level and component level, and the grid is divided level by level.
3. The intelligent vacuum monitoring method for a vacuum collector according to claim 1, characterized in that, Real-time anomaly monitoring includes the following steps: Clustering features within the same time window were selected, including core performance features and spatial features; the core performance features included efficiency ratio, normalized heat and appearance temperature difference ratio, and the spatial features included three-level coded coordinates. Standardize the clustering characteristics of all solar collectors within the current time period; The DBSCAN algorithm is used for cluster analysis to obtain core clusters, edge points, and outliers. Based on the cluster analysis results, the individual devices and devices in adjacent grids are compared to obtain real-time monitoring results.
4. The intelligent vacuum monitoring method for a vacuum collector according to claim 3, characterized in that, Based on the cluster analysis results, a comparison is made between a single device and devices in adjacent grids to obtain real-time monitoring results, including... From the core cluster obtained by spatial clustering, a batch of solar collectors are randomly selected, and the average value of their performance parameters is calculated to construct a baseline time series. The dynamic judgment matrix is set as: baseline time series ± α × standard deviation matrix; α represents the dynamic coefficient. Based on the dynamic judgment matrix, the time series data of the collectors in the edge points and outliers are compared with the dynamic judgment matrix. When the comparison result is abnormal, the problematic collector is located through three-level coding. Core clusters of the same type and from the same region with similar installation angles were selected as a reference group; Calculate the parameter differences ΔP and ΔT between the target point and the reference group, and use t-test to verify the differences.
5. The intelligent vacuum monitoring method for a vacuum collector according to claim 1, characterized in that, Secondary monitoring involves using predictive models to forecast the health of vacuum collectors, and includes the following steps: Feature extraction is performed on performance parameters and environmental sampling parameters to obtain time-series features and environmental coupling features; the time-series features include statistical features, trend features, frequency domain features, and slope features; the environmental coupling features include irradiance correction factor, temperature gradient ratio, cross-combination features, and hysteresis effect features; By inputting time-series features and environmental coupling features into the time-series prediction model, and analyzing the correlation between time-series features and environmental coupling features, the predicted health status of the solar collector can be obtained.
6. The intelligent vacuum monitoring method for a vacuum collector according to claim 5, characterized in that, The loss of the time series prediction model is: ； In the formula, Indicates the total loss. Indicates MSE loss. This indicates a loss of trend consistency.
7. An intelligent vacuum monitoring system for a vacuum collector, characterized in that, include: The modeling module is used to obtain the actual distribution data of vacuum collectors in the area, and to model the area based on the actual distribution data of vacuum collectors in the area to obtain the area distribution map. The data acquisition module is used to collect performance parameters and environmental parameters of the vacuum collector. The performance parameters include vacuum level, flow rate, collector tube temperature, collector temperature, and insulation box temperature; the environmental parameters include solar irradiance and irradiance time. The grid division module is used to divide the area distribution map using a multi-level grid division method, perform distributed monitoring on the divided multi-level grids, obtain the corresponding environmental and performance parameters, and use a three-level coding and positioning method to code and locate each level of grid. The monitoring module is used to set two levels of monitoring based on environmental and performance parameters: the first level is real-time anomaly monitoring, and the second level is to use a predictive model to predict the health of the vacuum collector; the monitoring status of each vacuum collector is located by using the code location as an index.