Intelligent turnover cabinet data analysis method and system integrated with intelligent sensor

Abstract

The invention relates to the technical field of intelligent turnover cabinets, and discloses an intelligent turnover cabinet data analysis method and system integrated with an intelligent sensor. The method comprises the following steps: carrying out distributed set member filtering processing on multi-source data from an intelligent turnover cabinet storage position sensor, an RFID reading module and an environment monitoring sensor to obtain storage position state data and measuring instrument position data; calculating an ellipsoid volume ratio index, and dynamically adjusting the acquisition frequency of the sensor based on the ellipsoid volume ratio index to obtain a sensor data stream; performing sensor behavior feature analysis on the sensor data stream to obtain an abnormal behavior detection result; and according to the abnormal behavior detection result and the measuring instrument information, constructing a storage location adaptation degree matrix and generating an access control instruction. According to the invention, rapid identification of the information of the measuring instrument is ensured, the problems of slow identification speed and low precision of a traditional system are solved, and accurate identification of abnormal behaviors such as random taking and random placing and wrong code reading is realized.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent turnover cabinets, and in particular to a data analysis method and system for intelligent turnover cabinets with integrated intelligent sensors. Background Art

[0002] As the power marketing system's requirements for metering asset management increase, the traditional metering instrument turnover management system can no longer meet the needs of intelligent and refined management. Existing metering instrument turnover cabinets generally have problems such as insufficient perception, serious data islands, and difficulty in multi-system coordination. They cannot achieve real-time and accurate monitoring of storage status and metering instrument locations, resulting in low asset management efficiency and frequent equipment loss or misplacement. Especially under the dual principles of "first in, first out" and "first in, first out", manual judgment is prone to deviations, resulting in waste of metering instrument resources and delays in business processes.

[0003] The sensor collection method of traditional turnover cabinets usually adopts periodic sampling with a fixed frequency, which does not take into account the difference in data change frequency and business importance. It not only leads to a large amount of redundant data transmission and storage, but also increases system power consumption and communication burden. At the same time, due to the lack of efficient fusion processing mechanism for sensor data, the existing system is difficult to deal with the problem of inconsistency of multi-sensor data, especially in complex environments such as RFID reading delay and signal interference. The system recognition accuracy is insufficient and it is unable to effectively monitor and respond to abnormal operation behaviors. Summary of the invention

[0004] The present application provides a data analysis method and system for an intelligent turnover cabinet with integrated intelligent sensors. The present application ensures rapid identification of measuring instrument information, solves the problems of slow identification speed and low accuracy of traditional systems, and realizes accurate identification of abnormal behaviors such as random picking and placing and code reading errors.

[0005] In a first aspect, the present application provides a data analysis method for an intelligent turnover cabinet integrated with an intelligent sensor, the data analysis method for an intelligent turnover cabinet integrated with an intelligent sensor comprising: Perform distributed set member filtering on multi-source data from intelligent turnover cabinet storage sensors, RFID reading modules and environmental monitoring sensors to obtain storage status data and measuring instrument location data; Calculating an ellipsoid volume ratio index using the storage position status data and the measuring instrument position data, and dynamically adjusting the sensor acquisition frequency based on the ellipsoid volume ratio index to obtain a sensor data stream; Performing sensor behavior feature analysis on the sensor data stream to obtain an abnormal behavior detection result; A storage location adaptation matrix is ​​constructed according to the abnormal behavior detection result and the measuring instrument information, and an access control instruction is generated based on the storage location adaptation matrix.

[0006] In a second aspect, the present application provides a smart turnover cabinet data analysis system integrated with a smart sensor, the smart turnover cabinet data analysis system integrated with a smart sensor comprising: A processing module is used to perform distributed set member filtering processing on multi-source data from the intelligent turnover cabinet storage position sensor, RFID reading module and environmental monitoring sensor to obtain storage position status data and measuring instrument position data; A calculation module, used to calculate an ellipsoid volume ratio index using the storage position status data and the measuring instrument position data, and dynamically adjust the sensor acquisition frequency based on the ellipsoid volume ratio index to obtain a sensor data stream; A feature analysis module, used to perform sensor behavior feature analysis on the sensor data stream to obtain an abnormal behavior detection result; A generation module is used to construct a storage location adaptation matrix according to the abnormal behavior detection result and the measuring instrument information, and to generate an access control instruction based on the storage location adaptation matrix.

[0007] In the technical solution provided by the present application, multi-source sensor data is processed by distributed set member filtering technology, which overcomes the data island problem of the traditional system and provides high-precision storage status and metering instrument position monitoring capabilities; the dynamic adjustment mechanism of sensor acquisition frequency based on the ellipsoid volume ratio index reduces unnecessary data acquisition and transmission, and reduces the overall energy consumption of the system; the specified time least squares recognition algorithm ensures the rapid recognition of metering instrument information, and solves the problems of slow recognition speed and low accuracy of the traditional system; the abnormal behavior detection model based on Markov jump characteristics can adapt to the time-varying delay environment of sensor reading, and realizes the accurate recognition of abnormal behaviors such as "random picking and placing" and "code reading errors"; the access control strategy optimized by linear matrix inequality takes into account both the "first-in, first-out" and "first-in, first-out" principles, greatly optimizing the efficiency of storage location selection and path planning; the present application also integrates historical data and real-time data, and establishes an adaptive learning mechanism, so that the intelligent turnover cabinet system can continuously optimize parameter configuration during long-term operation and maintain an efficient and stable operating state, thereby comprehensively improving the metering asset management level of the power marketing system. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0009] Figure 1This is a schematic diagram of an embodiment of a data analysis method for an intelligent turnover cabinet with integrated intelligent sensors in an embodiment of the present application; Figure 2 This is a schematic diagram of an embodiment of an intelligent turnover cabinet data analysis system with integrated intelligent sensors in an embodiment of the present application. DETAILED DESCRIPTION

[0010] The embodiments of the present application provide a method and system for analyzing data of an intelligent turnover cabinet with integrated intelligent sensors. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application 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 data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0011] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the intelligent turnover cabinet data analysis method integrating intelligent sensors includes: Step S101, performing distributed set member filtering processing on multi-source data from intelligent turnover cabinet storage position sensors, RFID reading modules and environmental monitoring sensors to obtain storage position status data and measuring instrument location data; It is understandable that the execution subject of the present application may be a smart turnover cabinet data analysis system integrated with a smart sensor, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0012] Specifically, the RFID reading module, storage position sensor and environmental monitoring sensor in the intelligent turnover cabinet are configured for collection frequency. By setting the collection interval and trigger mechanism of each sensor, the working frequency of different types of sensors is ensured to match the system target, forming a multi-source sensor data collection architecture. Based on the multi-source sensor collection architecture, raw data is collected, wherein the RFID reading module collects electronic tag data, including tag ID, type, calibration date and status code, and records the signal strength information of the tag for subsequent positioning analysis, while the storage position sensor provides pressure value and existence status information. The infrared detection module supplements the judgment of the occupancy of the storage position and combines the photoelectric switch to feedback the cabinet door status. The environmental monitoring sensor obtains the temperature and humidity data and vibration status data inside and outside the cabinet to provide auxiliary information for state estimation. The multi-source raw data is parsed and feature extracted, and the pressure value, existence status and temperature and humidity data of the storage position sensor and environmental monitoring sensor are extracted to form the storage position status raw data, which provides a basis for describing the real-time status of the storage position; at the same time, the electronic tag information collected by the RFID reading module and the corresponding signal strength are extracted to form the location information data of the measuring instrument. The original data of storage status is input into the distributed set membership filtering algorithm for state estimation and measurement constraint processing. The distributed set membership filtering constructs an ellipsoid set of multi-sensor measurement data, obtains a state estimation set by using intersection operation, and approximates it with the outer ellipsoid approximation technology to minimize the error and obtain the optimal state estimation. The distributed set membership filtering maps the storage status and the location information of the measuring instrument to the high-dimensional state space by establishing an ellipsoid set model, and introduces the ellipsoid volume ratio index to judge the data change. When the ellipsoid volume change exceeds the preset threshold, a new round of data collection is triggered to ensure the accuracy and real-time performance of the state estimation. At the same time, when the distributed set membership filtering algorithm performs state estimation in the data intersection, the influence of measurement noise is considered, and the state estimation error is minimized by the outer ellipsoid approximation technology to ensure the accuracy of data fusion. With the continuous update of the original data, the filtering algorithm dynamically adjusts the state estimation matrix and constrains different types of abnormal data, so that the stability and accuracy of the storage status data are significantly improved. Especially in the case of missing or incomplete data, the outer ellipsoid approximation method is used to make reasonable compensation to ensure the reliability of the data. After being processed by the distributed set membership filtering algorithm, the storage location status data is finally obtained, reflecting the changes in storage location occupancy status and environmental parameters, while the location information of the measuring instrument forms a complete data link through RFID data matching.

[0013] Step S102, calculating the ellipsoid volume ratio index using the storage position status data and the measuring instrument position data, and dynamically adjusting the sensor acquisition frequency based on the ellipsoid volume ratio index to obtain a sensor data stream; Specifically, the storage position status data includes the occupancy status, pressure information, temperature and humidity data of the storage position, and the location information of the items in the cabinet; at the same time, the measuring instrument location data is the location information data obtained by the RFID reading module and the signal strength of the electronic tag information inferred. In order to realize the dynamic data collection and control, the ellipsoid volume ratio is calculated for these data. Based on the state estimation ellipsoid model, the storage position status data and the measuring instrument location data are represented as an ellipsoid set, and the ellipsoid volume ratio index is obtained by calculating the ratio of the ellipsoid volume change at the current moment to the previous moment, reflecting the degree of change of the storage position and the location of the measuring instrument. After obtaining the ellipsoid volume ratio index, the threshold is allocated according to the importance of the storage position in the storage position status data. Different trigger thresholds are set for storage positions of different importance to achieve differentiated control, where the ellipsoid volume ratio threshold of the key storage position is lower, while the threshold of the ordinary storage position is higher. For example, the trigger threshold of the key storage position is set to 1.15, while the threshold of the ordinary storage position is set to 1.35. Through the differentiated threshold setting mechanism, the changes of the key storage position are monitored at a higher frequency, while the monitoring frequency of the ordinary storage position is relatively low. Based on the differentiated trigger threshold, the trigger conditions of the storage status data and the measuring instrument location data are judged. When the ellipsoid volume ratio index is greater than or less than the inverse of the trigger threshold, the system triggers a new data collection cycle. In order to avoid data redundancy and increased energy consumption caused by high-frequency triggering, the trigger signal is processed with time threshold constraints to ensure that the time interval between any two data collection cycles is not less than the set maximum interval. For example, the maximum interval of the RFID module is set to 3 minutes, and the maximum interval of the storage sensor is set to 5 minutes. If the current trigger time is insufficient from the last collection time, the trigger signal will be suppressed. This time constraint mechanism can effectively balance the timeliness and energy consumption of data collection. The RFID reading module is divided into multiple groups, and each group of RFID modules is set to time-sharing polling. The access interval between different groups is set to 100ms to avoid RFID reading conflicts and ensure that RFID reading modules in different areas can work together efficiently. At the same time, according to the characteristics of different types of sensors, differentiated trigger parameters are set. For example, the trigger condition of the storage pressure sensor is triggered when the pressure change exceeds 2.5g. The RFID reading module is triggered according to the ellipsoid volume ratio index, while the temperature and humidity sensor uses a fixed period trigger with an interval of 10 minutes. Based on these configurations, the sensor acquisition control instructions are generated and sent to various sensors to start the dynamic data acquisition control process. The sensor completes data acquisition according to the acquisition control instructions and transmits the collected data stream back to the system to form a sensor data stream, which includes storage position data and measuring instrument location information, as well as temperature and humidity changes, vibration status and other multi-dimensional data.

[0014] Step S103, performing sensor behavior feature analysis on the sensor data stream to obtain abnormal behavior detection results; Specifically, RFID tag data and barcode scan data are extracted from the sensor data stream, wherein the RFID tag data contains the identification information, type, calibration date and status code of the measuring instrument, and the barcode scan data records the item number, batch information and location information. After format conversion and preprocessing of these two types of data, a measurement output vector in a unified format is formed. The measurement output vector is parameterized to generate a parameter vector to be identified, and these parameters include key data such as the relevant attributes of the measuring instrument, location information and status code, and these data are input into the specified time least squares processing unit for parameter update. In the least squares identification process, the identification parameters are dynamically adjusted by using a time-varying gain method to accelerate the parameter update process. In order to ensure that the parameter identification is completed within a fixed time, a forced convergence time is set, and a nonlinear gain function is introduced for calculation acceleration so that the parameter identification can be completed within a predetermined time. In this process, the gain function is dynamically adjusted to effectively speed up the identification speed and ensure the accuracy of the identification result. After obtaining the identification result of the measuring instrument parameter, the missing data in the identification process is detected, and the missing data is linearly interpolated to compensate for the missing data to obtain a more complete identification result. When data missing is detected, the system performs linear interpolation through the previous valid data and uses the historical data trend to extrapolate the current data to supplement the missing data points and avoid recognition errors caused by data missing. The confidence of the recognition result after data compensation is calculated, and the reliability of the RFID recognition result is judged according to the preset confidence target value. If the recognition confidence is lower than the preset target value, the system automatically starts the barcode recognition data processing module, supplements the RFID data with the barcode scanning data, and combines the RFID data with the barcode scanning data through the weighted fusion processing mechanism to form more accurate measuring instrument information. In the weighted fusion process, weights are assigned according to the confidence levels of the two types of data to ensure the reliability of the final recognition result. The sensor behavior characteristic analysis of Markov jump is performed on the measuring instrument information and storage position status data. Markov jump characteristic analysis divides the system behavior into multiple different modes, including normal operation mode, storage position error mode, tag reading error mode and measuring instrument missing mode. On this basis, by comparing and analyzing the measurement data under different modes, it is determined whether the current system is in an abnormal state. At the same time, by building a residual detection mechanism, the difference between the measured data and the estimated data is calculated, and the type of abnormality is determined by the residual change trend. When the system detects abnormal behaviors such as "random handling and placement", "code reading error" or "missing measuring instruments", the corresponding processing flow is triggered according to different abnormal types, including rescanning, locking storage locations, starting asset inventory and other operations.

[0015] The information of measuring instruments and storage location status data are divided into modes, and the operating status of the system is divided into four working modes: normal operation mode, storage location error mode, tag reading error mode and measuring instrument missing mode. The normal operation mode means that the storage location status fully matches the measuring instrument information and the sensor data is updated normally, while the storage location error mode refers to the state feedback of the storage location sensor being inconsistent with the measuring instrument information, such as the wrong access sequence or the wrong item placement; the tag reading error mode is usually due to the failure of the RFID reading module to correctly identify the tag data, resulting in the failure of the measuring instrument location identification; the measuring instrument missing mode means that the item should be in the storage location but the tag or barcode data is not detected, so it is determined that the item is lost or missing. After completing the classification of the working modes, the state equation and measurement equation of Markov jump transition are constructed according to the classification data of the working mode, and the jump probability matrix is ​​set to describe the transition probability between each working mode to form a system state jump model. The jump probability matrix defines the state transition probability between different modes, such as the normal operation mode jumps to the storage location error mode or the tag reading error mode with a higher probability, but the probability of jumping to the measuring instrument missing mode is low. Since the state transition of the system has time correlation, the system state jump model is subjected to time-varying delay compensation. The compensation mechanism takes into account the data delay between the RFID reading module and the storage position sensor, and performs time alignment on the measurement data to ensure the time consistency of the data between different sensors, thus obtaining a measurement equation with time-varying delay characteristics. Based on the information of the measuring instrument and the storage position state data, the state residual analysis is performed. By comparing the difference between the actual observation data and the system state prediction data, the system behavior residual sequence is obtained, which reflects the behavior characteristics of the system under different working modes. Based on the system behavior residual sequence, the storage position error feature matrix, the tag reading error feature matrix and the measuring instrument missing feature matrix are constructed. These feature matrices extract the feature information in the residual sequence to judge the working mode of the current system, and calculate the feature distance by residual projection to quantify the distance between the current state and various abnormal modes, so as to determine the most likely abnormal type. After obtaining the judgment results of various types of abnormalities, the judgment results of multiple sensors are fused and analyzed, and a credibility weight is assigned to each sensor. These weights are dynamically adjusted according to the accuracy and historical performance of different sensors in various types of abnormality detection. Sensors with higher weights will occupy a larger proportion in the final judgment. Based on the maximum likelihood criterion, the probability of each working mode is calculated, the probability value of the observed data under different modes is maximized, and the final detection result of abnormal behavior is obtained. The maximum likelihood criterion can make the best judgment on abnormal behavior according to the probability distribution of different modes, thereby improving the accuracy of abnormal detection. Through the comprehensive analysis of the probability of each mode, abnormal behaviors such as storage location errors, label reading errors, and missing measuring instruments are identified, and the corresponding alarm mechanism or remedial measures are triggered.

[0016] The data change characteristics of the storage position sensor are extracted from the system behavior residual sequence, including the storage position pressure change rate and the number of existence state changes. The storage position pressure change rate is to judge the stability of the storage position state by comparing the change rate of the storage position pressure data at different times, while the number of existence state changes is to analyze the frequency of use and abnormal conditions of the storage position by recording the number of switch state changes of the storage position sensor in a short period of time. When the storage position state is abnormal, the pressure change rate fluctuates violently and the number of existence state changes will also increase significantly. Therefore, these characteristic data constitute the storage position error characteristic parameter set. The principal component analysis is performed on the storage position error characteristic parameter set, and the high-dimensional feature data is projected into the low-dimensional space. By extracting the feature vector with the largest variance, the storage position error feature matrix is ​​formed, which contains the main feature information of the storage position state change. At the same time, the RFID data in the system behavior residual sequence is feature extracted, and the signal strength fluctuation index and the reading success rate are calculated. The signal strength fluctuation index reflects the stability of the RFID tag signal in different time periods. When the signal strength fluctuates greatly, it means that the tag reading may be wrong. The reading success rate measures the working performance of the RFID reading module by calculating the ratio of the number of successful tag readings to the total number of readings in a specific time period. These two features together constitute the tag reading error feature parameter set. By performing covariance analysis on the tag reading error feature parameter set, the tag reading error feature matrix is ​​constructed. Covariance analysis can quantify the correlation between different feature variables and map high-dimensional data into the feature matrix, making the feature information of tag reading errors more recognizable. Based on the metrological instrument information data in the system behavior residual sequence, the data integrity and consistency indicators are calculated. The data integrity reflects whether there is data loss in a specific time period, while the consistency index measures the degree of matching between the metrological instrument data and the storage position status data. When the integrity or consistency of the metrological instrument information data deviates, it indicates that the metrological instrument is missing, so these feature data are used to construct the metrological instrument missing feature matrix. After the construction is completed, the first discrimination threshold, the second discrimination threshold and the third discrimination threshold are set for the storage error feature matrix, the tag reading error feature matrix and the measuring instrument missing feature matrix respectively. These thresholds are used to determine whether the feature distance exceeds the normal range, so as to judge whether an abnormality occurs. After the feature matrix and threshold setting are completed, the system behavior residual sequence is projected onto the storage error feature matrix, the tag reading error feature matrix and the measuring instrument missing feature matrix respectively, and the projection distance is calculated. The projection distance reflects the matching degree between the current system state and various abnormal feature matrices. The larger the distance, the higher the probability of abnormality.The storage error characteristic distance value, tag reading error characteristic distance value and measuring instrument missing characteristic distance value are compared with the corresponding discrimination thresholds. When the storage error characteristic distance value exceeds the first discrimination threshold, the system determines that a storage error occurs; when the tag reading error characteristic distance value exceeds the second discrimination threshold, the system determines that an abnormality occurs in the RFID tag reading; and when the measuring instrument missing characteristic distance value exceeds the third discrimination threshold, the system determines that the measuring instrument is missing. Different types of abnormal behaviors can be distinguished through the method of multidimensional feature matrix and projection distance calculation, and abnormal judgment results can be generated in a timely manner.

[0017] Step S104: construct a storage location adaptation matrix according to the abnormal behavior detection result and the measuring instrument information, and generate an access control instruction based on the storage location adaptation matrix.

[0018] Specifically, the storage location status in the intelligent turnover cabinet is divided into four types, including idle, occupied and available, occupied but locked, and occupied and reserved. Among them, the idle state means that there is no item stored in the storage location and it can be allocated at any time. The occupied and available state means that there are items in the storage location but the access operation can be performed. The locked state is triggered by the abnormal behavior detection result, which is used to mark the storage location where the abnormality is detected and prohibit any access operation. The occupied and reserved state means that the storage location has been allocated to a specific measuring instrument and is waiting for the access operation. By analyzing the abnormal behavior detection results, the abnormal storage location is identified in time, and the lock mark is set for it. The latest storage location state vector is updated to reflect the status information of all storage locations in the system. According to the key attributes such as ID, type, calibration date and status code in the measuring instrument information, the attribute of the measuring instrument is extracted to form a measuring instrument attribute vector, where the ID identifies the unique number of the measuring instrument, the type indicates the category of the item, the calibration date reflects the service life of the item, and the status code records the current status of the item, including normal, pending inspection, invalid, etc. By calculating the matching degree between the storage position state vector and the metering instrument attribute vector, the storage position adaptation matrix is ​​obtained. Each element of the matrix represents the degree of adaptation between a storage position and a metering instrument. The higher the adaptation degree, the higher the rationality and safety of the metering instrument stored in the corresponding storage position. At the same time, in order to optimize the access order, the priority is calculated according to the information of the metering instrument, and the priority sequence is set according to the principle of "first in, first out" or "first in, first out". Among them, the metering instrument with the closer verification date and the more urgent use status has the higher priority. The storage position adaptation matrix and the metering instrument priority sequence are input into the quadratic programming solver for solution. By constructing an optimization problem with storage position selection and path planning as the goal, and converting the constraints into the form of linear matrix inequalities, the optimal storage position allocation scheme and access path planning results are obtained. The storage position allocation scheme ensures that the metering instrument is stored in the optimal storage position, thereby maximizing the utilization of the storage space, and the access path planning result is optimally controlled according to the motion trajectory and path cost of the access mechanism, so that the system can achieve the goals of shortest path, optimal time and lowest energy consumption during the access process. The motion control parameters of the access mechanism are set for the storage allocation plan and the access path planning results. By accurately controlling the motion trajectory, acceleration, speed change and other parameters of the access mechanism, the execution sequence of the access operation is generated. In order to improve the dynamic response capability of the system, the dynamic time window mechanism is introduced to optimize the execution sequence. The dynamic time window mechanism dynamically adjusts the time window length according to the real-time operating status of the system to ensure that the system can maintain a high operating efficiency under high load conditions and avoid system delays or congestion caused by task accumulation. After scheduling optimization, the access control instructions finally generated are sent to the control system of the intelligent turnover cabinet, thereby realizing efficient access to measuring instruments and safe management of abnormal storage locations.

[0019] In the embodiments of the present application, multi-source sensor data is processed by distributed set member filtering technology, which overcomes the data island problem of the traditional system and provides high-precision storage status and metering instrument position monitoring capabilities; the dynamic adjustment mechanism of sensor acquisition frequency based on the ellipsoid volume ratio index reduces unnecessary data acquisition and transmission, and reduces the overall energy consumption of the system; the specified time least squares recognition algorithm ensures the rapid recognition of metering instrument information, and solves the problems of slow recognition speed and low accuracy of the traditional system; the abnormal behavior detection model based on Markov jump characteristics can adapt to the time-varying delay environment of sensor reading, and realizes the accurate recognition of abnormal behaviors such as "random picking and placing" and "code reading errors"; the access control strategy optimized by linear matrix inequality takes into account both the "first-in, first-out" and "first-in, first-out" principles, greatly optimizing the efficiency of storage location selection and path planning; the present application also integrates historical data and real-time data, and establishes an adaptive learning mechanism, so that the intelligent turnover cabinet system can continuously optimize parameter configuration during long-term operation and maintain an efficient and stable operating state, thereby comprehensively improving the metering asset management level of the power marketing system.

[0020] In a specific embodiment, the process of executing step S101 may specifically include the following steps: Configure the acquisition frequency of the RFID reading module, storage position sensor and environmental monitoring sensor in the intelligent turnover cabinet to obtain a multi-source sensor acquisition architecture; Collect multi-sensor raw data through a multi-source sensor acquisition architecture; According to the multi-sensor raw data, the pressure value, existence state and environmental parameters of the storage position sensor and the environmental monitoring sensor are extracted to obtain the storage position state raw data; According to the multi-sensor raw data, the electronic tag information and signal strength collected by the RFID reading module are extracted to obtain the measuring instrument location data; The original data of storage state are input into the distributed set member filtering algorithm for state estimation and measurement constraint processing to obtain storage state data.

[0021] Specifically, differentiated collection frequency configurations are performed according to the functional characteristics of the RFID reading module, storage location sensor, and environmental monitoring sensor. The collection frequency configuration of the RFID reading module is set according to the flow frequency of the items in the cabinet. When the system is in a state of high-frequency placement and placement of items, the collection interval of the RFID reading module is shortened to ensure the real-time update of the measuring instrument information. In the low-flow state, the collection interval is appropriately increased to reduce data redundancy and energy consumption. The collection frequency configuration of the storage location sensor is dynamically adjusted in combination with the storage location occupancy state. When the storage location state changes, the sensor needs to immediately feedback the current pressure value and existence status information. Therefore, when an item is placed or taken away, the collection interval needs to be shortened, and when the storage location remains stationary, the collection interval can be extended. For environmental monitoring sensors, including temperature and humidity sensors and triaxial acceleration sensors, their collection frequency should be flexibly configured according to the rate of environmental change. The update frequency of temperature and humidity data is set to once every 10 minutes to capture the long-term change trend of the environment in the turnover cabinet, and the collection frequency of the acceleration sensor is set at 1kHz to monitor the vibration state of the cabinet in real time to ensure that the system can perceive abnormal vibration or displacement in time. After completing the configuration of the acquisition frequency of the multi-source sensor, a multi-source sensor acquisition architecture is formed. This architecture connects various sensors to the field controller through the RS485 bus, uses the Modbus-RTU protocol for communication, and sets the baud rate to 9600bps to ensure the stability and reliability of data transmission. Through this acquisition architecture, the system realizes real-time acquisition and parallel processing of raw data from multiple sensors. During the data acquisition process, the RFID reading module reads the electronic tag information from each storage location, including the ID, type, calibration date and status code of the measuring instrument, and records the signal strength data of the tag; the storage location sensor collects the pressure value and existence status information of the storage location, where the pressure value reflects the weight change of the item placed, and the existence status is used to determine whether the storage location is idle or occupied. At the same time, the environmental monitoring sensor continuously monitors the temperature and humidity parameters inside the cabinet and the external vibration status, and after the data acquisition is completed, the raw data of the multi-source sensor is transmitted to the data processing module of the system for further analysis. The raw data of the multi-source sensor is parsed and feature extracted, and the original information of the storage location status, including pressure value, existence status and environmental parameters, is extracted from the data of the storage location sensor and the environmental monitoring sensor. The change in pressure value is used to judge the taking and placing behavior of items. When the pressure value changes significantly, the system infers the change of the storage items, and the information of the existence status is used to judge whether the storage location is currently free or occupied. These two types of data together constitute the original data of the storage location status. The environmental parameters collected by the temperature and humidity sensor reflect the changes in the storage environment, including the impact of temperature changes on the storage status of items and the potential threat of humidity changes to the storage environment. These environmental parameters are also an important part of the storage location status data, providing environmental background data for subsequent state estimation and anomaly detection.At the same time, the electronic tag information and signal strength are extracted from the data collected by the RFID reading module to obtain the location data of the measuring instrument. The electronic tag information includes the tag ID, item type, verification date and status code. This information is used to determine the uniqueness of the item and evaluate the usage status of the item in combination with the verification date to ensure that the item is properly managed within the service life. The signal strength information is used to determine the spatial position of the item. By comparing and analyzing the signal strengths read by different antennas, the relative position of the item in the storage location is calculated to form the location data of the measuring instrument. The original data of the storage location status is input into the distributed set membership filtering algorithm for state estimation and measurement constraint processing to obtain more accurate storage location status data. The distributed set membership filtering algorithm is a state estimation method based on the ellipsoid set theory. By constructing an ellipsoid set model of the storage location status, multi-sensor data is fused and optimized. The algorithm represents the sensor measurement value as a bounded ellipsoid set, and the estimated value of the storage location status data is used as the center position of the ellipsoid set, while the measurement noise and data error are quantified by the shape matrix of the ellipsoid. By performing intersection operations on the data sets of different sensors, the state estimation set after multi-sensor data fusion is obtained. Due to the complexity of the intersection calculation of multi-source data, the outer ellipsoid approximation method is used to approximate the intersection result, and the state estimation error is optimized by minimizing the ellipsoid matrix trace method, and finally more accurate storage position state data is obtained. At the same time, the distributed set member filtering algorithm performs dynamic constraint processing on the measurement data. When the system detects data anomalies or noise interference, it automatically adjusts the measurement weight to reduce the impact of abnormal data on the state estimation results, ensuring that the system can still maintain a high state estimation accuracy in the case of incomplete or abnormal data. Through multi-level data processing and optimization methods, the storage position state data finally obtained includes the real-time occupancy status, pressure changes and environmental parameters of the storage position, and combines the RFID tag reading information and signal strength data to achieve accurate positioning of the measuring instrument.

[0022] In a specific embodiment, the process of executing step S102 may specifically include the following steps: Calculate the ellipsoid volume ratio of the storage position data and the measuring instrument position data to obtain the ellipsoid volume ratio index; According to the importance of the reservoir in the reservoir status data, a threshold value is assigned to the ellipsoid volume ratio index to obtain a differentiated trigger threshold value; The trigger condition judgment is performed on the storage position status data and the measuring instrument position data based on the differential trigger threshold, and when the ellipsoid volume ratio index is greater than or less than the inverse of the differential trigger threshold, a sensor trigger signal is obtained; Performing time threshold constraint processing on the sensor trigger signal to obtain a time constraint trigger signal; Divide the RFID reading modules into multiple groups, perform time-sharing polling settings for each group of RFID reading modules, and set differentiated trigger parameters according to different sensor characteristics to obtain sensor acquisition control instructions; Based on the sensor acquisition control instruction, the sensor dynamic acquisition control is performed to obtain the sensor data stream.

[0023] Specifically, key feature information is extracted from storage status data and measuring instrument location data. Storage status data includes pressure change, existence status, and environmental parameters, while the measuring instrument location data comes from the electronic tag information and signal strength of the RFID reading module. The ellipsoid set model of the system state is constructed through these data. The ellipsoid set model represents the current system state with an ellipsoid, where the center of the ellipsoid represents the state estimation value, and the shape matrix of the ellipsoid reflects the uncertainty of the system state. After the multi-sensor data is collected and fused, the ellipsoid volume is calculated according to the state estimation covariance matrix at different time points, and the ellipsoid volume ratio index is obtained by comparing the ratio of the current ellipsoid volume to the ellipsoid volume at the previous moment, reflecting the degree of change of the system state between two acquisitions. The change of the ellipsoid volume ratio can judge the dynamic change of the storage status and the location of the measuring instrument. If the ellipsoid volume ratio is large, it means that the system state has changed significantly, and data collection needs to be re-triggered to update the system state. If the ellipsoid volume ratio is small, the system state change is not obvious, and data collection is delayed to reduce energy consumption. After obtaining the ellipsoid volume ratio index, differentiated trigger thresholds are allocated according to the importance of the storage location in the storage location status data. The importance of the storage location is comprehensively evaluated based on factors such as item type, verification date, frequency of use, and item value. Different trigger thresholds are set for storage locations of different importance. For example, the trigger threshold of key storage locations is lower, set to 1.15, so that data collection can be quickly triggered when the state changes, while the trigger threshold of ordinary storage locations is higher, set to 1.35, and data update is triggered only when the state changes significantly. The differentiated trigger threshold allocation mechanism can effectively balance the frequency of data collection and system energy consumption, ensuring that key storage locations receive more frequent data updates, while ordinary storage locations delay collection when the state changes are small, thereby improving the overall efficiency of the system. The trigger conditions of the storage location status data and the measuring instrument location data are judged based on differentiated trigger thresholds. When the ellipsoid volume ratio index is greater than the set trigger threshold, or less than the reciprocal of the trigger threshold, the system triggers a new round of data collection. This mechanism ensures that the data is updated in time when the system state changes significantly, and prevents system misjudgment caused by the failure to capture the state change in time. The trigger condition judgment mechanism effectively improves the system's sensitivity to abnormal changes and can automatically adjust the data collection strategy according to changes in the system state. In order to prevent frequent triggering from causing data redundancy and increased energy consumption, the sensor trigger signal is processed with a time threshold constraint to ensure that the minimum time interval is maintained between two data collection cycles. Even if the trigger condition is met, data collection will not be triggered repeatedly in a short period of time. The time threshold of the RFID reading module is set to 3 minutes, and the time threshold of the storage sensor is set to 5 minutes. Even if the ellipsoid volume ratio changes significantly, a new round of data collection can only be carried out after the minimum time interval is reached.Through the time constraint mechanism, the data acquisition frequency is effectively reduced, avoiding resource waste caused by high-frequency acquisition in the system, while maintaining the accuracy of system state estimation. After completing the time constraint processing, the RFID reading modules are divided into multiple groups, and time-sharing polling settings are performed on each group of RFID modules. Since the number of RFID reading modules is often large in a large-scale turnover cabinet system, if all modules perform data acquisition simultaneously, it will lead to data conflicts and increased energy consumption. Therefore, the RFID modules are divided into multiple groups, and time-sharing polling is performed on different groups of RFID modules. The polling time interval for each group of modules is set to 100 milliseconds. Through the polling mechanism, the reading conflicts between RFID reading modules are avoided, ensuring that each module can successfully complete data acquisition. At the same time, different triggering parameters are set according to the characteristics of different types of sensors. For example, the triggering parameter of the storage position pressure sensor is triggered when the pressure change exceeds 2.5g, while the triggering parameter of the environmental monitoring sensor is triggered when the temperature change exceeds 0.5°C or the humidity change exceeds 2%RH. The differential triggering parameter setting mechanism can perform refined control according to the data characteristics of different sensors, thereby further improving the data acquisition efficiency of the system. After completing the time-sharing polling and differential triggering parameter settings, sensor acquisition control instructions are generated. These control instructions include the acquisition frequency control, triggering condition setting, and data transmission commands for the RFID reading modules, storage position sensors, and environmental monitoring sensors. The system performs dynamic acquisition control on the sensors according to the control instructions. The dynamic acquisition control mechanism automatically adjusts the acquisition frequency according to the change of the system state. When the system is in a stable state, the acquisition frequency is maintained at a low level to save energy consumption, while when the system detects an abnormal change, the acquisition frequency will increase rapidly to ensure the timeliness of data update. Through the dynamic acquisition control mechanism, while ensuring the accuracy of data acquisition, the optimal management of the sensor data stream is achieved.

[0024] In a specific embodiment, the process of executing step S103 may specifically include the following steps: Extract RFID tag data and barcode scanning data from the sensor data stream, and perform format conversion and preprocessing on the RFID tag data and barcode scanning data to obtain a measurement output vector; Perform parameter setting on the measurement output vector to obtain a parameter vector to be recognized; Input the measurement output vector and the parameter vector to be recognized into a prescribed time least squares processing unit, calculate the parameter update amount in a time-varying gain manner, and obtain the measurement instrument parameter recognition result; Set the forced convergence time according to the measurement instrument parameter recognition result, and use a non-linear gain function for calculation acceleration to obtain the recognition output information within a prescribed time; Detect the data missing points in the sensor data stream according to the recognition output information within the specified time, and perform linear interpolation compensation on the data missing points to obtain the corrected recognition result; Calculate the confidence of the corrected recognition result. When the RFID recognition confidence is lower than the preset target value, automatically start the barcode recognition data processing, and perform weighted fusion processing on the RFID tag data and the barcode scanning data to obtain the measuring instrument information. The sensor behavior characteristic analysis of Markov jump is performed on the measuring instrument information and storage position status data to obtain the abnormal behavior detection result.

[0025] Specifically, multi-source data is obtained from the sensor data stream, including the tag data collected by the RFID reading module and the barcode information captured by the barcode scanning device. The RFID tag data contains the core attributes of the measuring instrument, such as ID, type, calibration date, and status code, while the barcode scanning data contains key information such as item number, batch information, and storage location information. Then, the RFID tag data and the barcode scanning data are format converted and preprocessed, and the two types of data are standardized and uniformly converted into the measurement output vector format. The reading results of the RFID tag data have problems such as multi-tag conflict and tag signal strength fluctuation. Therefore, the tag data is preprocessed by deduplication, signal strength filtering, and outlier removal. At the same time, the barcode scanning data is partially missing or inconsistent in format due to the scanning angle and barcode quality, so format verification and data completion are performed to ensure the integrity and consistency of the final measurement output vector. Through format conversion and preprocessing, data from different sources are fused into a unified measurement output vector. The measurement output vector is parameterized and mapped to a parameter vector to be identified. The vector contains key attribute parameters of the measuring instrument, such as the unique identification ID, type, calibration date, current status, etc. of the measuring instrument. At the same time, the historical data accumulated during the operation of the system is introduced to model the state trend of the measuring instrument, and these parameters are input as variables to be identified into the specified time least squares processing unit for processing. The specified time least squares processing unit uses the recursive least squares method to implement parameter update, dynamically adjusts the parameter update rate by means of time-varying gain, and continuously corrects the parameter estimation value according to the change of the observed data to improve the identification accuracy. The specified time least squares method accelerates parameter update through the time-varying gain matrix. When new data is input, the gain matrix dynamically adjusts the weight according to the confidence of the data, so that the system can quickly converge to the optimal solution within a limited time and obtain more accurate measurement instrument parameter identification results. The mandatory convergence time is set according to the measurement instrument parameter identification results, and the nonlinear gain function is introduced to accelerate the calculation process. The mandatory convergence time is used to ensure that parameter identification is completed within the specified time. By introducing a nonlinear gain function, the system increases the gain coefficient when it is close to convergence, thereby accelerating the convergence speed of the identification process. The nonlinear gain function dynamically adjusts the gain value according to the current identification error, so that the system can quickly approach the identification result within the specified time and obtain more efficient identification output information. After obtaining the identification output information within the specified time, the data missing points in the sensor data stream are detected. The data missing is caused by RFID reading failure, barcode scanning angle offset or environmental interference. In order to improve the identification accuracy, the data missing points are compensated by linear interpolation. Linear interpolation compensation uses the previous valid data for linear prediction to deduce the approximate value of the missing data, and inserts the compensated data into the measurement output vector to obtain the corrected identification result.The confidence of the corrected recognition result is calculated to determine whether the recognition result meets the reliability requirements of the system. The confidence calculation quantifies the reliability of the recognition result by comparing the matching degree of the recognition result with the historical data. If the confidence is lower than the preset target value, the system automatically triggers the barcode recognition data processing module, and the barcode scanning data will be weighted fused with the RFID tag data as supplementary information. In the weighted fusion process, weights are assigned according to the confidence of the two types of data, and the data with high confidence accounts for a larger proportion, thereby improving the accuracy of the final recognition result. Through the multi-source data fusion mechanism, the inaccurate recognition problem caused by a single data source is effectively compensated, and more reliable metrological instrument information is generated. The metrological instrument information is associated with the storage position status data, and the sensor behavior characteristics are analyzed based on the Markov jump model. The Markov jump characteristic model divides the different working modes of the system into multiple states such as normal operation mode, storage position error mode, tag reading error mode and measuring instrument missing mode, and the state jump model is modeled according to the probability matrix of the system state change. By analyzing the matching between the storage position status data and the measuring instrument information, it is determined whether the current working mode is abnormal. When the system behavior characteristics deviate from the normal mode, the abnormal behavior detection mechanism is triggered. Combined with the changing trend of the system behavior residual, various abnormal behavior characteristics are analyzed, including abnormal situations such as random picking and placing of items, tag recognition failure, and loss of measuring instruments. The probability of abnormal behavior is calculated based on the maximum likelihood criterion to obtain accurate abnormal behavior detection results.

[0026] In a specific embodiment, the execution step performs Markov jump sensor behavior feature analysis on the measuring instrument information and the storage position status data to obtain the abnormal behavior detection result, which may specifically include the following steps: The measuring instrument information and storage position status data are divided into modes to obtain working mode classification data, the working mode classification data including normal operation mode, storage position error mode, label reading error mode and measuring instrument missing mode; According to the classification data of the working mode, the state equation and measurement equation of Markov jump are constructed, the jump probability matrix is ​​set, the system state jump model is obtained, and the system state jump model is subjected to time-varying delay compensation processing to obtain the measurement equation with time-varying delay characteristics; Based on the information of measuring instruments and storage position status data, the state residual analysis is carried out to obtain the system behavior residual sequence; According to the system behavior residual sequence, the storage error feature matrix, label reading error feature matrix and measuring instrument missing feature matrix are constructed, and the feature distance calculation of the residual projection is performed to obtain the judgment results of various types of abnormalities; According to the various types of abnormal judgment results of multiple sensors, a credibility weight is assigned to each sensor, and the probability of each working mode is calculated based on the maximum likelihood criterion to obtain the abnormal behavior detection result.

[0027] Specifically, feature extraction and pattern analysis are performed on the multi-source data collected by the system. The measuring instrument information includes key attributes such as the ID, type, calibration date, and status code of the item, while the storage location status data includes data such as storage location occupancy status, pressure change, temperature and humidity parameters, and RFID tag signal strength. By performing feature analysis on these data, the storage status of the measuring instrument and the storage location usage are extracted, and the working mode classification rules are established in combination with historical data. The normal operation mode means that the storage location status fully matches the measuring instrument information, the item access behavior conforms to the system logic, the tag is read successfully, and the measuring instrument does not have any abnormality; the storage location error mode reflects that the data fed back by the storage location sensor is inconsistent with the RFID tag information or the barcode scanning result, which is caused by the wrong placement of the item or the disorder of the access order; the tag reading error mode mainly occurs when the RFID reading module fails to correctly identify the tag data, resulting in the inability to accurately match the position of the measuring instrument; and the measuring instrument missing mode indicates that the item data that should be detected in the storage location is lost, including the item being taken away but not recorded, the tag is lost, or the RFID reading fails. By real-time analysis of the metering instrument information and storage position status data, the current operating state of the system is automatically divided into modes, and the working mode classification data is generated. The state equation and measurement equation of Markov jump transition are constructed according to the working mode classification data, and the jump probability matrix is ​​set to establish the system state jump model. The Markov jump model is a discrete state space model in which the system state jumps between different modes, and the probability of the jump is determined by the state transition matrix. The system state equation is constructed, the storage position state data and metering instrument information are mapped into state vectors, and the state jump matrix is ​​defined in combination with the system operation law. Through statistical analysis of historical data, the state transition probability between each mode is calculated to generate the jump probability matrix. The measurement equation is constructed, the storage position state data and metering instrument information are used as observation data, and the measurement noise covariance matrix is ​​established in combination with the noise characteristics of the system to ensure the robustness and accuracy of the model. After the state equation and measurement equation are constructed, the system state jump model is subjected to time-varying delay compensation processing to correct the time delay problem of different sensor data during the system operation. Due to the different working frequencies and data update cycles of the RFID reading module, storage position sensor and barcode scanning equipment, data asynchrony occurs. Therefore, the system state jump model is compensated for time-varying delays, and the data of different sensors are aligned to the same time dimension to obtain a measurement equation with time-varying delay characteristics. State residual analysis is performed based on the measuring instrument information and storage position state data to obtain the system behavior residual sequence. State residual analysis determines whether the current behavior of the system is abnormal by comparing the difference between the system's observed data and the state prediction data. When the storage position state data does not match the measuring instrument information, the system generates a large residual, while the residual in the normal operation mode is usually small.The system behavior residual sequence is a time series representation of residual data in different modes. By performing feature analysis on the residual sequence, the characteristics of abnormal behaviors such as storage errors, tag reading errors and missing measuring instruments are extracted. According to the system behavior residual sequence, the storage error feature matrix, tag reading error feature matrix and missing measuring instrument feature matrix are constructed, and the feature distance calculation of the residual projection is performed to obtain the judgment results of each type of abnormality. The storage error feature matrix is ​​constructed by analyzing the difference characteristics between the storage state data and the measuring instrument information, including characteristic variables such as pressure change rate and number of state changes. These characteristics can characterize the characteristics of storage error behavior. The tag reading error feature matrix is ​​constructed based on characteristic data such as RFID signal strength fluctuation and reading success rate. By capturing the signal interference and reading failure that occur during the reading process of the RFID module, the tag reading error behavior is identified. The missing measuring instrument feature matrix is ​​modeled by analyzing the matching of the storage state and the measuring instrument, as well as the residual changes caused by the loss of items or data anomalies. The system behavior residual sequence is projected onto the storage error feature matrix, the tag reading error feature matrix, and the measuring instrument missing feature matrix, and the feature distance is calculated. The feature distance is compared with the set discrimination threshold to determine whether the corresponding type of abnormal behavior occurs. A credibility weight is assigned to each sensor according to the various types of abnormal judgment results of multiple sensors, and the probability of each working mode is calculated based on the maximum likelihood criterion. The credibility weight is dynamically adjusted according to the historical performance and data reliability of different sensors. Sensors with high confidence will occupy a greater weight in the final judgment, thereby improving the reliability of the abnormal detection results. The probability of each working mode is calculated by the maximum likelihood criterion, the probability value of the observed data under different modes is maximized, and the optimal abnormal behavior detection result is obtained. The maximum likelihood criterion can combine the observed data and abnormal judgment results of different sensors to calculate the probability of different working modes, so as to accurately identify the current mode of the system and generate abnormal behavior detection results.

[0028] In a specific embodiment, the execution step constructs a storage error feature matrix, a tag reading error feature matrix, and a measuring instrument missing feature matrix according to the system behavior residual sequence, and performs feature distance calculation of residual projection to obtain various types of abnormality judgment results. The process may specifically include the following steps: According to the change characteristics of the reservoir sensor data in the system behavior residual sequence, the reservoir pressure change rate and the number of existing state changes are calculated to obtain the reservoir error characteristic parameter set; Perform principal component analysis on the storage error characteristic parameter set to obtain the storage error characteristic matrix; The feature extraction of RFID data in the system behavior residual sequence is performed, the signal strength fluctuation index and the reading success rate are calculated, the tag reading error feature parameter set is obtained, and the tag reading error feature matrix is ​​constructed through covariance analysis; Based on the integrity and consistency indicators of the measuring instrument information data in the system behavior residual sequence, a measuring instrument missing feature matrix is ​​constructed, and the first discrimination threshold of the storage error feature matrix, the second discrimination threshold of the tag reading error feature matrix, and the third discrimination threshold of the measuring instrument missing feature matrix are set respectively; The system behavior residual sequence is projected onto the storage position error feature matrix, the tag reading error feature matrix and the measuring instrument missing feature matrix respectively, and the projection distance is calculated to obtain the storage position error feature distance value, the tag reading error feature distance value and the measuring instrument missing feature distance value; The storage location error characteristic distance value is compared with the first discrimination threshold, the tag reading error characteristic distance value is compared with the second discrimination threshold, and the measuring instrument missing characteristic distance value is compared with the third discrimination threshold to obtain various types of abnormal judgment results.

[0029] Specifically, the data change characteristics of the storage location sensor are extracted from the system behavior residual sequence, including pressure data and existence state information. The storage location pressure change rate reflects the speed of change of the weight of the items in the storage location. When the items are taken and placed, the pressure data will fluctuate significantly. The pressure change rate is calculated by comparing the pressure changes at different times. The number of existence state changes refers to the number of state switching times of the storage location within a certain period of time, such as the number of state transitions from idle to occupied, from occupied to locked, etc. The count is performed by detecting the time series of storage location state changes. These two characteristic variables together constitute the storage location error feature parameter set. The storage location error feature parameter set is subjected to principal component analysis to extract the most representative feature information. Principal component analysis is a data dimensionality reduction technology. By performing covariance matrix decomposition on the storage location error feature parameter set, the main characteristic components are extracted, and the eigenvalues ​​are sorted. The first few principal components with the largest variance are selected as feature vectors to obtain the storage location error feature matrix. At the same time, feature extraction is performed on the RFID data in the system behavior residual sequence, focusing on the two core features of signal strength fluctuation index and reading success rate. The signal strength fluctuation index is used to measure the stability of RFID tag signals in different time periods. If the signal strength fluctuates greatly, it indicates that there is interference or signal attenuation in the tag reading process, which leads to reading failure. The reading success rate measures the working performance of the RFID reading module by calculating the ratio of the number of successful tag readings to the total number of attempts in a specific time period. These two features together constitute the tag reading error feature parameter set. After obtaining the tag reading error feature parameter set, the tag reading error feature matrix is ​​constructed through covariance analysis. Covariance analysis can measure the correlation between different feature variables. By performing feature decomposition on the covariance matrix of the tag reading error feature parameter set, key feature components are extracted, and these feature components are mapped into the tag reading error feature matrix, thereby effectively characterizing the abnormal behavior occurring during the RFID tag reading process. The missing feature matrix of the measuring instrument is constructed based on the integrity and consistency indicators of the measuring instrument information data in the system behavior residual sequence. Data integrity reflects whether there is data loss in a specific time period, while the consistency index measures the degree of matching between the measuring instrument information data and the storage position status data. If there is a deviation in data integrity and consistency, it means that the measuring instrument data is abnormal or lost. By calculating the changing trends of integrity and consistency indicators, the missing feature matrix of measuring instruments is constructed to effectively characterize the feature information of missing or abnormal status of measuring instruments. After completing the construction of the feature matrix, the first discrimination threshold, the second discrimination threshold and the third discrimination threshold are set for the storage error feature matrix, the tag reading error feature matrix and the missing feature matrix of measuring instruments. These thresholds are used to determine whether the feature distance exceeds the normal range, so as to judge whether the corresponding type of abnormal behavior occurs.Based on the historical data and statistical analysis results of the system, the threshold is set. By measuring the characteristic distance of the system in the normal state for many times, the average value and standard deviation of the characteristic distance are calculated, and the abnormal discrimination threshold is set as the mean value plus a certain multiple of the standard deviation, so as to ensure that the early warning mechanism can be triggered in time when an abnormality occurs. After completing the setting of the characteristic matrix and the discrimination threshold, the system behavior residual sequence is projected onto the storage error characteristic matrix, the label reading error characteristic matrix, and the measuring instrument missing characteristic matrix, and the projection distance is calculated. The projection distance quantifies the degree of matching between the current state and various abnormal modes by calculating the distance between the current residual sequence and the characteristic matrix. The larger the projection distance, the greater the degree to which the current state of the system deviates from the normal mode, indicating that the probability of abnormality is higher. By calculating the storage error characteristic distance value, the label reading error characteristic distance value, and the measuring instrument missing characteristic distance value, it is judged whether the system has abnormal behaviors such as storage error, label reading error, or measuring instrument missing. After obtaining the characteristic distance values ​​of each type, these distance values ​​are compared with the preset discrimination thresholds. When the storage error characteristic distance value exceeds the first discrimination threshold, the system determines that a storage error has occurred; when the tag reading error characteristic distance value exceeds the second discrimination threshold, the system determines that an abnormality has occurred in the RFID tag reading; and when the measuring instrument missing characteristic distance value exceeds the third discrimination threshold, the system determines that the measuring instrument information is missing or abnormal. Through the abnormal detection mechanism based on the comparison of the characteristic matrix projection distance and the discrimination threshold, abnormal behaviors such as storage errors, tag reading errors, and missing measuring instruments are identified, and the corresponding abnormal handling process is triggered in time.

[0030] In a specific embodiment, the process of executing step S104 may specifically include the following steps: The storage location status in the intelligent turnover cabinet is divided into four types: free, occupied and available, occupied but locked, and occupied and reserved. The abnormal storage location is locked and marked according to the abnormal behavior detection result to obtain the storage location status vector. According to the ID, type, verification date and status code in the measuring instrument information, the measuring instrument attributes are extracted to obtain the measuring instrument attribute vector; The storage position state vector and the measuring instrument attribute vector are matched to obtain the storage position adaptation matrix, and the measuring instrument information is prioritized to obtain the measuring instrument priority sequence; The storage location adaptation matrix and the priority sequence of the measuring instruments are input into the quadratic programming solver, and the constraints are converted into the form of linear matrix inequalities to obtain the storage location allocation plan and the access path planning results; The access mechanism motion control parameters are set for the storage allocation plan and access path planning results, an execution sequence is generated, and the execution sequence is scheduled and optimized according to the dynamic time window mechanism to obtain the access control instructions.

[0031] Specifically, the storage location status is classified, and the storage location changes are monitored in real time. The four states of the storage location reflect the current usage status of the storage location. The idle state means that the storage location is not occupied and items can be accessed at any time; the occupied and available state means that there are items stored in the storage location, but normal access operations can be performed; the occupied but locked state means that although the storage location is occupied, the system has locked the storage location due to abnormal behavior detection, and access operations are prohibited; and the occupied and reserved state means that the storage location has been allocated to a specific measuring instrument, and the system reserves the storage location for subsequent access tasks. Based on the classification of storage location status, the abnormal storage location is locked and marked according to the abnormal behavior detection results. The abnormal behavior detection results are derived from the Markov jump characteristic analysis and the characteristic distance calculation of the residual projection. When the system detects abnormal behaviors such as storage location errors, label reading errors, or missing measuring instruments, the locking mechanism will be automatically triggered, and the corresponding storage location status will be marked as "occupied but locked" to prevent continued access operations, thereby ensuring that the abnormal storage location will not affect the normal operation of the system. By dynamically locking and marking abnormal storage locations, the storage location status is quickly adjusted when an abnormality occurs, and the storage location status is updated to a storage location status vector. Attributes are extracted based on key attributes such as ID, type, verification date, and status code in the measuring instrument information to form a measuring instrument attribute vector. ID is used to identify the uniqueness of the measuring instrument, type represents the classification information of the item, verification date reflects the service life of the item, and status code records the current status of the item, such as normal, pending inspection, and invalid. By extracting these attribute information, a measuring instrument attribute vector is formed. The storage location state vector and the measuring instrument attribute vector are matched to calculate the degree of fit to obtain a storage location fitness matrix. The matching degree calculation is based on a comprehensive evaluation of the degree of fit between the item attributes and the storage location status, where the elements of the storage location fitness matrix represent the degree of fit between a storage location and a measuring instrument. The higher the degree of fit, the more reasonable it is to store the measuring instrument in the corresponding storage location. In the process of matching degree calculation, factors such as the idle state of the storage location, the type of item, the verification date, and the priority of the item are comprehensively considered, and the matching degree is quantified through multi-parameter weighted calculation to ensure the rationality of storage location allocation. At the same time, the priority calculation of the measuring instrument information is performed to generate the priority sequence of the measuring instrument. The setting of the priority follows the principle of "first in, first out" or "first in, first out". The measuring instrument with a higher priority should be given priority for access operations. For example, the measuring instrument with a verification date approaching expiration or a status code marked as "pending inspection" is set to a higher priority to ensure the timely verification and safe use of the items. The priority sequence of the measuring instrument is dynamically sorted according to these factors and provides a decision basis for the subsequent storage location allocation. The storage location fitness matrix and the priority sequence of the measuring instrument are input into the quadratic programming solver for solution, and the constraints are converted into linear matrix inequality form to obtain the storage location allocation plan and access path planning results.The goal of quadratic programming is to maximize the storage location matching degree and optimize the access path, with the objective function of minimizing the access time, reducing energy consumption and maximizing the storage location utilization. At the same time, constraints are set in combination with the storage location state vector and the metering instrument attribute vector, such as ensuring that the locked storage location is not allocated, and that items with high priority are allocated to storage locations with higher matching degrees. By constructing linear matrix inequalities, these constraints are converted into matrix form for solution, and the optimal storage location allocation scheme and access path planning results are obtained. The storage location allocation scheme and access path planning results are set with the motion control parameters of the access mechanism to generate a specific execution sequence. The motion control parameters of the access mechanism include motion speed, acceleration, path trajectory, and stop position. By accurately setting these parameters, it is ensured that the access mechanism reaches the optimal motion state when performing access operations, thereby improving access efficiency and reducing energy consumption. At the same time, in order to optimize the execution sequence, a dynamic time window mechanism is introduced for scheduling optimization. The dynamic time window mechanism dynamically adjusts the time window length of the execution sequence according to the real-time operating status of the system. When the system is under high load, the time window is shortened to speed up the access speed, and when the system load is low, the time window is appropriately extended to reduce system energy consumption. Through the dynamic adjustment mechanism, it is ensured that the system can maintain the best operating state under different workloads and maximize the working efficiency of the access mechanism. The generated execution sequence is converted into specific access control instructions after optimization processing by the dynamic time window mechanism. These control instructions are sent to the control module of the intelligent turnover cabinet, driving the access mechanism to perform access operations on items according to the optimal path and sequence, thereby realizing the automatic allocation and efficient management of storage locations.

[0032] The above describes the data analysis method of the intelligent turnover cabinet integrated with the intelligent sensor in the embodiment of the present application. The following describes the data analysis system of the intelligent turnover cabinet integrated with the intelligent sensor in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the intelligent turnover cabinet data analysis system integrating intelligent sensors includes: The processing module 201 is used to perform distributed set member filtering processing on multi-source data from the intelligent turnover cabinet storage position sensor, RFID reading module and environmental monitoring sensor to obtain storage position status data and measuring instrument position data; A calculation module 202 is used to calculate an ellipsoid volume ratio index using the storage position status data and the measuring instrument position data, and dynamically adjust the sensor acquisition frequency based on the ellipsoid volume ratio index to obtain a sensor data stream; The feature analysis module 203 is used to perform sensor behavior feature analysis on the sensor data stream to obtain abnormal behavior detection results; The generation module 204 is used to construct a storage location adaptation matrix according to the abnormal behavior detection result and the measuring instrument information, and generate an access control instruction based on the storage location adaptation matrix.

[0033] Through the coordinated cooperation of the above-mentioned components, the multi-source sensor data is processed by distributed set member filtering technology, which overcomes the data island problem of the traditional system and provides high-precision storage status and metering instrument position monitoring capabilities; the dynamic adjustment mechanism of sensor acquisition frequency based on the ellipsoid volume ratio index reduces unnecessary data collection and transmission, and reduces the overall energy consumption of the system; the specified time least squares recognition algorithm ensures the rapid recognition of metering instrument information, and solves the problems of slow recognition speed and low accuracy of the traditional system; the abnormal behavior detection model based on Markov jump characteristics can adapt to the time-varying delay environment of sensor reading, and realizes the accurate recognition of abnormal behaviors such as "random picking and placing" and "code reading errors"; the access control strategy optimized by linear matrix inequality takes into account both the "first-in, first-out" and "first-in, first-out" principles, greatly optimizing the efficiency of storage location selection and path planning; this application also integrates historical data and real-time data, and establishes an adaptive learning mechanism, so that the intelligent turnover cabinet system can continuously optimize parameter configuration during long-term operation and maintain an efficient and stable operating state, thereby comprehensively improving the metering asset management level of the power marketing system.

[0034] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0035] 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 all 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 an intelligent turnover cabinet data analysis device (which can be a personal computer, server, or network device, etc.) with an integrated intelligent sensor 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 codes.

[0036] 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 data analysis method for an intelligent turnover cabinet integrating an intelligent sensor, characterized in that: include: Perform distributed set member filtering on multi-source data from intelligent turnover cabinet storage sensors, RFID reading modules and environmental monitoring sensors to obtain storage status data and measuring instrument location data; Calculating an ellipsoid volume ratio index using the storage position status data and the measuring instrument position data, and dynamically adjusting the sensor acquisition frequency based on the ellipsoid volume ratio index to obtain a sensor data stream; Performing sensor behavior feature analysis on the sensor data stream to obtain an abnormal behavior detection result; A storage location adaptation matrix is ​​constructed according to the abnormal behavior detection result and the measuring instrument information, and an access control instruction is generated based on the storage location adaptation matrix.

2. The method for analyzing data of intelligent turnover cabinets with integrated intelligent sensors according to claim 1 is characterized in that: The distributed set member filtering process is performed on the multi-source data from the intelligent turnover cabinet storage position sensor, the RFID reading module and the environmental monitoring sensor to obtain the storage position status data and the measuring instrument position data, including: Configure the acquisition frequency of the RFID reading module, storage position sensor and environmental monitoring sensor in the intelligent turnover cabinet to obtain a multi-source sensor acquisition architecture; Collecting multi-sensor raw data through the multi-source sensor acquisition architecture; Extracting the pressure value, existence state and environmental parameters of the storage position sensor and the environmental monitoring sensor according to the multi-sensor raw data to obtain storage position state raw data; Extract the electronic tag information and signal strength collected by the RFID reading module according to the multi-sensor raw data to obtain the measuring instrument position data; The original data of the storage position state is input into a distributed set member filtering algorithm for state estimation and measurement constraint processing to obtain storage position state data.

3. The data analysis method of the intelligent turnover cabinet integrated with the intelligent sensor according to claim 1 is characterized in that: The method of calculating the ellipsoid volume ratio index by using the storage position status data and the measuring instrument position data, and dynamically adjusting the sensor acquisition frequency based on the ellipsoid volume ratio index to obtain the sensor data stream includes: Calculating the ellipsoid volume ratio of the storage position data and the measuring instrument position data to obtain an ellipsoid volume ratio index; assigning a threshold to the ellipsoid volume ratio index according to the storage position importance in the storage position status data to obtain a differentiated trigger threshold; Based on the differential trigger threshold, the storage position state data and the measuring instrument position data are subjected to trigger condition judgment, and when the ellipsoid volume ratio index is greater than or less than the inverse of the differential trigger threshold, a sensor trigger signal is obtained; Performing time threshold constraint processing on the sensor trigger signal to obtain a time constraint trigger signal; The RFID reading modules are divided into multiple groups, time-sharing polling is set for each group of RFID reading modules, and differentiated trigger parameters are set according to different sensor characteristics to obtain sensor acquisition control instructions; Based on the sensor acquisition control instruction, dynamic sensor acquisition control is performed to obtain a sensor data stream.

4. The method for analyzing data of an intelligent turnover cabinet with integrated intelligent sensor according to claim 1 is characterized in that: The performing sensor behavior feature analysis on the sensor data stream to obtain an abnormal behavior detection result includes: Extracting RFID tag data and barcode scan data from the sensor data stream, and performing format conversion and preprocessing on the RFID tag data and the barcode scan data to obtain a measurement output vector; Setting parameters of the measurement output vector to obtain a parameter vector to be identified; Input the measurement output vector and the parameter vector to be identified into a specified time least square processing unit, calculate the parameter update amount by a time-varying gain method, and obtain a parameter identification result of the measuring instrument; According to the measurement instrument parameter identification result, a mandatory convergence time is set, and a nonlinear gain function is used to accelerate the calculation to obtain the identification output information within a specified time; Detecting data missing points in the sensor data stream according to the recognition output information within the specified time, and performing linear interpolation compensation on the data missing points to obtain a corrected recognition result; Calculating the confidence of the corrected recognition result, automatically starting the barcode recognition data processing when the RFID recognition confidence is lower than a preset target value, and performing weighted fusion processing on the RFID tag data and the barcode scan data to obtain the measuring instrument information; A Markov jump sensor behavior feature analysis is performed on the measuring instrument information and the storage position status data to obtain an abnormal behavior detection result.

5. The method for analyzing data of an intelligent turnover cabinet with integrated intelligent sensor according to claim 4 is characterized in that: The performing of Markov jump sensor behavior feature analysis on the measuring instrument information and the storage position status data to obtain abnormal behavior detection results includes: Performing mode division on the measuring instrument information and the storage position status data to obtain working mode classification data, wherein the working mode classification data includes a normal operation mode, a storage position error mode, a tag reading error mode, and a measuring instrument missing mode; Constructing a state equation and a measurement equation of Markov jump according to the working mode classification data, setting a jump probability matrix, obtaining a system state jump model, and performing time-varying delay compensation processing on the system state jump model to obtain a measurement equation with time-varying delay characteristics; Perform state residual analysis based on the measuring instrument information and the storage position state data to obtain a system behavior residual sequence; According to the system behavior residual sequence, a storage position error feature matrix, a tag reading error feature matrix and a measuring instrument missing feature matrix are constructed, and characteristic distance calculation of residual projection is performed to obtain various types of abnormality judgment results; According to the various types of abnormal judgment results of the multi-sensors, a credibility weight is assigned to each sensor, and the probability of each working mode is calculated based on the maximum likelihood criterion to obtain the abnormal behavior detection result.

6. The method for analyzing data of an intelligent turnover cabinet with integrated intelligent sensor according to claim 5 is characterized in that: The storage position error feature matrix, the tag reading error feature matrix and the measuring instrument missing feature matrix are constructed according to the system behavior residual sequence, and the feature distance calculation of the residual projection is performed to obtain various types of abnormality judgment results, including: Calculate the reservoir pressure change rate and the number of existing state changes according to the change characteristics of the reservoir sensor data in the system behavior residual sequence to obtain a reservoir error characteristic parameter set; Performing principal component analysis on the storage position error characteristic parameter set to obtain a storage position error characteristic matrix; Extract features from the RFID data in the system behavior residual sequence, calculate the signal strength fluctuation index and the reading success rate, obtain a tag reading error feature parameter set, and construct a tag reading error feature matrix through covariance analysis; Based on the integrity and consistency index of the measuring instrument information data in the system behavior residual sequence, a measuring instrument missing feature matrix is ​​constructed, and a first discrimination threshold of the storage position error feature matrix, a second discrimination threshold of the tag reading error feature matrix, and a third discrimination threshold of the measuring instrument missing feature matrix are set respectively; Projecting the system behavior residual sequence onto the storage position error feature matrix, the tag reading error feature matrix and the measuring instrument missing feature matrix respectively, calculating the projection distance, and obtaining the storage position error feature distance value, the tag reading error feature distance value and the measuring instrument missing feature distance value; The storage position error characteristic distance value is compared with the first discrimination threshold, the tag reading error characteristic distance value is compared with the second discrimination threshold, and the measuring instrument missing characteristic distance value is compared with the third discrimination threshold to obtain various types of abnormality judgment results.

7. The method for analyzing data of an intelligent turnover cabinet with integrated intelligent sensor according to claim 1, characterized in that: The step of constructing a storage location adaptability matrix according to the abnormal behavior detection result and the measuring instrument information, and generating an access control instruction based on the storage location adaptability matrix, includes: The storage position status in the intelligent turnover cabinet is divided into four types: free, occupied and available, occupied but locked, and occupied and reserved, and the abnormal storage position is locked and marked according to the abnormal behavior detection result to obtain a storage position state vector; Extracting the attributes of the measuring instrument according to the ID, type, verification date and status code in the measuring instrument information to obtain a measuring instrument attribute vector; Performing matching calculation on the storage position state vector and the measuring instrument attribute vector to obtain a storage position adaptation matrix, and performing priority calculation on the measuring instrument information to obtain a measuring instrument priority sequence; Input the storage location adaptability matrix and the metering instrument priority sequence into a quadratic programming solver, convert the constraints into a linear matrix inequality form, and obtain a storage location allocation plan and an access path planning result; The access mechanism motion control parameters are set for the storage location allocation scheme and the access path planning results to generate an execution sequence, and the execution sequence is scheduled and optimized according to a dynamic time window mechanism to obtain access control instructions.

8. An intelligent turnover cabinet data analysis system integrated with intelligent sensors, characterized in that: The method for analyzing data of an intelligent turnover cabinet integrated with an intelligent sensor according to any one of claims 1 to 7 is used, wherein the intelligent turnover cabinet data analysis system integrated with an intelligent sensor comprises: A processing module is used to perform distributed set member filtering processing on multi-source data from the intelligent turnover cabinet storage position sensor, RFID reading module and environmental monitoring sensor to obtain storage position status data and measuring instrument position data; A calculation module, used to calculate an ellipsoid volume ratio index using the storage position status data and the measuring instrument position data, and dynamically adjust the sensor acquisition frequency based on the ellipsoid volume ratio index to obtain a sensor data stream; A feature analysis module, used to perform sensor behavior feature analysis on the sensor data stream to obtain an abnormal behavior detection result; A generation module is used to construct a storage location adaptation matrix according to the abnormal behavior detection result and the measuring instrument information, and to generate an access control instruction based on the storage location adaptation matrix.

Images