Intelligent storage environment management method and system based on artificial intelligence

Abstract

The invention relates to the technical field of intelligent storage environment management, in particular to an intelligent storage environment management method and system based on artificial intelligence, and the method comprises the steps: calculating a theoretical humidity value of an internal cavity of closed power equipment according to an external environment parameter and an equipment sealing grade; obtaining cavity response characteristics including a shell vibration spectrum and microwave dielectric response data; inputting the cavity response features into a pre-trained humidity inversion neural network to obtain an actually measured humidity value of the internal cavity; according to the absolute deviation value of the theoretical humidity value and the actually measured humidity value, a cavity risk index is calculated in combination with the key grade coefficient of the closed power equipment; and when the cavity risk index exceeds a preset risk threshold, generating an active intervention instruction for the closed power equipment. And the risk that the closed power equipment is affected with damp due to humidity change in a storage environment is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehouse environment management technology, specifically an intelligent warehouse environment management method and system based on artificial intelligence. Background Technology

[0002] With the increasing demand for power equipment storage, ensuring the stable maintenance of equipment performance in the storage environment has become crucial. This is especially true for enclosed power equipment, where the environmental conditions within its internal cavities directly affect its long-term reliability and safety. During storage, enclosed power equipment is often affected by fluctuations in external temperature and humidity. Through the equipment's own "breathing effect"—changes in internal gas volume due to temperature variations—it exchanges gases with the outside world, allowing humid air to intrude into the equipment cavity. This can cause moisture damage to internal insulation materials, corrosion of metal components, and other hidden dangers, seriously threatening the stability and safety of the equipment's subsequent operation.

[0003] Currently, traditional warehouse environment monitoring mainly relies on temperature and humidity sensors deployed in open spaces. While these sensors can reflect the overall environmental conditions of the warehouse, they cannot directly detect changes in the microenvironment within enclosed equipment cavities. Taking transformers as an example, their oil conservator space, as a typical enclosed cavity, is highly susceptible to drawing in humid air due to temperature fluctuations during diurnal temperature variations or seasonal changes. This process often occurs silently even when the overall warehouse environmental parameters appear normal. Due to the lack of direct monitoring and precise control methods for the internal environment of enclosed cavities, existing technologies struggle to effectively prevent potential faults such as decreased insulation performance and increased partial discharge caused by internal moisture, increasing the risk of equipment damage during storage.

[0004] Therefore, this invention proposes an intelligent warehouse environment management method based on artificial intelligence, which realizes real-time perception of the internal cavity environment of sealed electrical equipment, thereby effectively avoiding performance degradation and safety hazards caused by internal moisture, and significantly improving the intelligence level of warehouse environment management and the security of equipment storage. Summary of the Invention

[0005] (1) Technical problems to be solved

[0006] The purpose of this invention is to provide an intelligent warehouse environment management method and system based on artificial intelligence to solve the problem of moisture risk to enclosed electrical equipment caused by humidity changes in the warehouse environment.

[0007] (2) Technical solution

[0008] To achieve the above objectives, in one aspect, the present invention provides an intelligent warehouse environment management method based on artificial intelligence, the method comprising:

[0009] S1. Real-time acquisition of external environmental parameters of enclosed electrical equipment in the warehouse environment, including equipment casing temperature, ambient temperature and ambient humidity.

[0010] S2. Obtain the equipment sealing level of the sealed power equipment, and calculate the theoretical humidity value of the internal cavity of the sealed power equipment based on the external environmental parameters and the equipment sealing level.

[0011] S3. Obtain cavity response characteristics by using a non-contact sensor attached to the casing of the sealed power equipment. The cavity response characteristics include the casing vibration spectrum and microwave dielectric response data. Input the cavity response characteristics into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity.

[0012] S4. Based on the absolute deviation between the theoretical humidity value and the measured humidity value, and combined with the criticality level coefficient of the sealed electrical equipment, the cavity risk index is calculated. When the cavity risk index exceeds the preset risk threshold, an active intervention command is generated for the sealed electrical equipment. The active intervention command includes activating the directional dehumidification device and adjusting the storage location of the sealed electrical equipment.

[0013] Furthermore, the method for calculating the theoretical humidity value of the internal cavity of the sealed electrical equipment based on the external environmental parameters and the equipment sealing level includes:

[0014] The temperature difference rate between the equipment casing and the ambient temperature is calculated in real time. When the temperature difference rate exceeds a preset sensitivity threshold, the gas exchange volume per unit time is calculated by a linear-nonlinear hybrid regression algorithm using the temperature difference rate, equipment cavity volume parameters, and basic exchange coefficient. The basic exchange coefficient is determined by mapping the equipment sealing level.

[0015] The gas exchange volume and ambient humidity are used to calculate the number of water vapor moles entering the equipment cavity using the ideal gas law; the number of water vapor moles is converted into water vapor mass and then divided by the total mass of the dry gas in the cavity to obtain the original humidity ratio; the original humidity ratio is then smoothed and filtered through a time series to obtain the theoretical humidity value.

[0016] The total mass of the drying gas in the cavity is dynamically calibrated by multiplying the cavity volume by the standard dry air density, and the dynamic calibration process incorporates an altitude compensation factor.

[0017] Furthermore, the method of inputting the cavity response features into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity includes:

[0018] Energy is extracted from the vibration spectrum of the outer shell in a specific frequency band to separate a subset of vibration features that are strongly correlated with the motion of gas molecules inside the cavity; the microwave dielectric response data is subjected to time-frequency joint transformation to generate a dielectric loss feature map.

[0019] The vibration feature subset and dielectric loss feature map are fused across modes to obtain a high-dimensional fused feature vector. The high-dimensional fused feature vector is then extracted by a dual-branch feature extraction method of a humidity inversion neural network to obtain a dual-branch output feature vector. The dual-branch output feature vector is then reconstructed by attention weighting to generate a latent space representation of the cavity humidity distribution. The first branch output feature vector of the dual-branch output feature vector is used to extract local spatial features through a convolutional neural network, and the second branch output feature vector is used to capture dynamic evolution features through a temporal coding network.

[0020] The latent space representation of the humidity distribution in the cavity is input into a fully connected neural network layer and mapped to the measured humidity value of the internal cavity.

[0021] Furthermore, the method for fusing the vibration feature subset with the dielectric loss feature spectrum across modes to obtain a high-dimensional fused feature vector includes:

[0022] A modal alignment reference is constructed based on the thickness of the metal shell of the sealed power equipment cavity and the distribution pattern of the insulating medium; based on the modal alignment reference, the vibration feature subset is compensated for through penetration attenuation, and the dielectric loss feature spectrum is corrected for multipath reflection to generate standardized vibration features and standardized dielectric loss features.

[0023] The modal correlation weight matrix is ​​generated by training the acoustic-electric coupling simulation dataset of the equipment cavity; the standardized vibration features and standardized dielectric loss features are dynamically fused to obtain a high-dimensional fused feature vector, and the feature contribution weights of the standardized vibration features and standardized dielectric loss features are assigned according to the modal correlation weight matrix during the dynamic feature fusion process.

[0024] Furthermore, the method for allocating the feature contribution weights of standardized vibration features and standardized dielectric loss features according to the modal correlation weight matrix during dynamic feature fusion includes:

[0025] The coordinates of key areas of the cavity are analyzed based on the three-dimensional structural model of the sealed power equipment. The key areas include the installation position of the sealing ring and the core distribution area of ​​the insulating medium. A regional coordinate index table corresponding to the physical space of the cavity is constructed, and the coordinates of the key areas are converted into index identifiers that can be addressed by the modal correlation weight matrix. The vibration characteristic weight coefficients of the sealing ring area and the dielectric characteristic weight coefficients of the insulating medium area are generated based on the historical fault database.

[0026] The vibration characteristic weight coefficient of the sealing ring region and the dielectric characteristic weight coefficient of the insulating medium region are written into the corresponding index positions of the modal correlation weight matrix to form a region-differentiated weight allocation matrix.

[0027] The feature contribution weights are assigned to the standardized vibration characteristics and standardized dielectric loss characteristics based on the regional differential weight allocation matrix.

[0028] Furthermore, the method for capturing dynamic evolutionary features through a temporal coding network includes:

[0029] Based on the time-varying characteristics of the vibration feature subset and dielectric loss feature spectrum, periodic feature waveforms synchronized with the breathing effect of the equipment cavity are identified; key evolution frames within each breathing cycle are extracted, the key evolution frames including gas inhalation features during the temperature drop phase and gas exhalation features during the temperature recovery phase.

[0030] After performing cross-cycle consistency alignment on the key evolution frames, the dynamic correlation pattern between the key evolution frames is learned through a gated recurrent unit network, generating a temporal coding vector that is strongly correlated with the cavity humidity change and denoted as the output feature vector of the second branch.

[0031] Furthermore, the method for calculating the cavity risk index based on the absolute deviation between the theoretical humidity value and the measured humidity value, combined with the criticality level coefficient of the sealed electrical equipment, includes:

[0032] The continuous growth trend of the absolute deviation is calculated, and the risk accumulation factor is activated when the absolute deviation shows a unidirectional increase over multiple consecutive monitoring periods; a key correction coefficient is generated based on the voltage level and structural type of the enclosed power equipment.

[0033] The risk base value of the equipment body is calculated based on the risk accumulation factor and the critical correction coefficient; the risk index of the cavity body is calculated by combining the risk base value of the equipment body with the environmental acceleration deterioration coefficient through a nonlinear coupling function.

[0034] Furthermore, the method for calculating the base risk value of the equipment itself based on the risk accumulation factor and the criticality correction coefficient includes:

[0035] An aging attenuation factor is generated based on the service life of the sealing ring of the sealed electrical equipment, and the key correction coefficient is dynamically corrected based on the aging attenuation factor.

[0036] When the rate of change of temperature and humidity in the storage environment exceeds the equipment's breathing compensation capability threshold, the credibility verification of the risk accumulation factor is initiated, and the fusion weight of the risk accumulation factor is dynamically reduced based on the breathing compensation capability coefficient; the critical correction coefficient after timeliness correction is weighted and fused with the risk accumulation factor to obtain the fusion result value.

[0037] By matching the current humidity deviation curve with the moisture similarity of typical moisture-induced accident modes using a historical fault case database of enclosed power equipment, a risk enhancement coefficient is generated when the moisture similarity is greater than a preset similarity threshold. The fusion result value and the risk enhancement coefficient are then used to obtain the equipment's risk base value through nonlinear mapping.

[0038] Furthermore, the method for obtaining the environmental acceleration deterioration coefficient includes:

[0039] The environmental temperature and humidity change rate data are collected in real time through the warehouse microenvironment monitoring network. The current environmental deterioration intensity value is calculated based on the environmental temperature and humidity change rate data. Historical plum rain erosion records of the sealed power equipment storage area are retrieved and combined with meteorological early warning information to generate a seasonal deterioration correction coefficient.

[0040] The degradation intensity value is sensitively calibrated according to the equipment sealing level; the environmental accelerated degradation coefficient is calculated by combining the calibrated degradation intensity value with the seasonal degradation correction factor.

[0041] On the other hand, based on the same inventive concept, the present invention also provides an intelligent warehouse environment management system based on artificial intelligence. The system includes: an external environment parameter acquisition module, a theoretical humidity calculation module, a measured humidity inversion module, and a cavity risk assessment and active intervention instruction generation module, with each module connected in sequence.

[0042] The external environment parameter acquisition module is used to collect the external environment parameters of enclosed electrical equipment in the warehouse environment in real time. The external environment parameters include the equipment casing temperature, ambient temperature and ambient humidity.

[0043] The theoretical humidity calculation module is used to obtain the equipment sealing level of the sealed electrical equipment and calculate the theoretical humidity value of the internal cavity of the sealed electrical equipment based on the external environmental parameters and the equipment sealing level.

[0044] The measured humidity inversion module is used to acquire cavity response characteristics through a non-contact sensor attached to the casing of a sealed electrical equipment. The cavity response characteristics include the casing vibration spectrum and microwave dielectric response data. The cavity response characteristics are input into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity.

[0045] The cavity risk assessment and active intervention command generation module is used to calculate the cavity risk index based on the absolute deviation between the theoretical humidity value and the measured humidity value, combined with the criticality level coefficient of the sealed electrical equipment; when the cavity risk index exceeds the preset risk threshold, an active intervention command is generated for the sealed electrical equipment, the active intervention command including activating the directional dehumidification device and adjusting the storage location of the sealed electrical equipment.

[0046] (3) Beneficial effects

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] 1. By combining external environmental parameters with the equipment's sealing level to calculate theoretical humidity values, and by using non-contact sensors to acquire cavity response characteristics and using a humidity inversion neural network to obtain measured humidity values, accurate monitoring of the humidity inside the sealed power equipment's cavity is achieved. This can more comprehensively and accurately reflect the internal humidity status of the equipment, effectively avoiding performance degradation and safety hazards caused by internal moisture.

[0049] 2. Based on accurate monitoring, and according to the absolute deviation between theoretical and measured humidity values, combined with the criticality level coefficient of the sealed electrical equipment, the equipment cavity risk index is accurately calculated. Furthermore, based on the cavity risk index, active intervention commands are generated for the sealed electrical equipment to ensure that the equipment is always in suitable storage conditions, effectively reducing the risk of equipment damage and improving the level of intelligence in warehouse management.

[0050] 3. Since the sealing quality of enclosed electrical equipment is particularly important, by comprehensively considering multiple factors such as the service life of the equipment's sealing rings, deterioration characteristics, the rate of change of temperature and humidity in the storage environment, and historical plum rain erosion records, the environmental acceleration deterioration coefficient is dynamically calculated, and the risk assessment is optimized accordingly. This improves the accuracy of risk assessment and ensures the safety of equipment storage in the storage environment management. Attached Figure Description

[0051] Figure 1 This is a flowchart of an intelligent warehouse environment management method based on artificial intelligence, according to Embodiment 1 of the present invention.

[0052] Figure 2 This is a schematic diagram of the module composition of an artificial intelligence-based intelligent warehouse environment management system according to Embodiment 2 of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Before giving examples, it is necessary to explain the application scenarios of the present invention. The present invention is an intelligent warehousing environment management method and system based on artificial intelligence. It is applied to the storage of enclosed electrical equipment. Due to fluctuations in the temperature and humidity of the external environment, the equipment itself has a "breathing effect" that causes gas exchange with the outside world. This allows humid air to enter the equipment cavity, which will cause hidden dangers such as moisture absorption of internal insulation materials and corrosion of metal parts. The environmental state of the internal cavity of the enclosed electrical equipment directly affects the long-term reliability and safety of the equipment.

[0055] Example 1: As Figure 1 As shown in the figure, this embodiment provides an intelligent warehouse environment management method based on artificial intelligence, the method including:

[0056] S1. Real-time acquisition of external environmental parameters of enclosed electrical equipment in the storage environment, including equipment casing temperature, ambient temperature, and ambient humidity; for example, when a 10kV oil-immersed distribution transformer is stored in a warehouse, external environmental parameters are acquired in real time through a network of environmental sensors distributed around the equipment. These sensors acquire data every 5 minutes, including the casing temperature measured by a temperature sensor attached to the surface of the transformer casing, the ambient temperature measured by an ambient temperature sensor inside the warehouse, and the ambient relative humidity measured by a humidity sensor.

[0057] S2. Obtain the sealing rating of the sealed power equipment. Calculate the theoretical humidity value of the internal cavity of the sealed power equipment based on the external environmental parameters and the sealing rating. Obtain the sealing rating information of the transformer from the equipment file database. The transformer belongs to the IP54 protection level, and the corresponding sealing performance parameter is a medium sealing level.

[0058] S3. The cavity response characteristics are acquired using non-contact sensors attached to the casing of the sealed power equipment. These characteristics include the casing vibration spectrum and microwave dielectric response data. The cavity response characteristics are input into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity. Cavity response characteristics are also acquired using an array of non-contact sensors attached to different locations on the transformer casing. Vibration sensors capture micro-vibration signals in the oil conservator area, and the casing vibration spectrum is obtained through fast Fourier transform analysis, focusing on extracting vibration characteristics related to the motion of gas molecules inside the cavity within the 15-150Hz frequency band. Microwave sensors emit electromagnetic waves of a specific frequency that penetrate the transformer casing, receive the reflected signals, and analyze their dielectric response. A dielectric loss characteristic spectrum is generated through time-frequency joint transformation.

[0059] S4. Based on the absolute deviation between the theoretical humidity value and the measured humidity value, and combined with the criticality level coefficient of the sealed electrical equipment, a cavity risk index is calculated. When the cavity risk index exceeds a preset risk threshold, an active intervention command is generated for the sealed electrical equipment. The active intervention command includes activating a directional dehumidification device and adjusting the storage location of the sealed electrical equipment. The directional dehumidification device closest to the equipment is activated, with a target humidity set at 45% and running continuously for 1 hour. Simultaneously, an instruction to adjust the storage location is issued, suggesting that the equipment be moved from the current high-humidity area to a relatively dry area within the warehouse to ensure that the humidity of the equipment cavity returns to a safe range, effectively preventing the risk of decreased insulation performance and equipment damage due to internal moisture.

[0060] The method for calculating the theoretical humidity value of the internal cavity of the sealed power equipment based on the external environmental parameters and the equipment sealing level includes:

[0061] The system calculates the rate of change of the temperature difference between the equipment casing and the ambient temperature in real time. When the rate of change exceeds a preset sensitivity threshold, the gas exchange volume per unit time is calculated using a linear-nonlinear hybrid regression algorithm based on the rate of change, the equipment cavity volume parameter, and the basic exchange coefficient. The basic exchange coefficient is determined by mapping the equipment's sealing level. A real-time monitoring mechanism for the rate of change of temperature difference is established, calculating the rate of change of the temperature difference between the equipment casing and the ambient temperature every minute. For example, when the transformer's equipment casing temperature changes from 32.5℃ to 33.2℃ and the ambient temperature changes from 28.0℃ to 28.1℃, the rate of change of temperature difference is 0.6℃ / min, exceeding the set sensitivity threshold of 0.4℃ / min, and the gas exchange volume calculation is immediately initiated. Using the transformer's technical data, the oil conservator cavity volume parameter is obtained as 15 liters, and the basic exchange coefficient of 0.012 m³ / h•℃ is determined from a preset mapping table based on the IP54 sealing level. The gas exchange volume per unit time was calculated using a hybrid linear-nonlinear regression algorithm. The linear part considered the direct proportionality between the temperature difference driving force and the cavity volume, while the nonlinear part incorporated the effects of sealing aging and the quadratic term of the temperature gradient. The calculation yielded a gas exchange volume of 0.081 m³ per hour at the current rate of temperature change.

[0062] The gas exchange volume and ambient humidity are used to calculate the number of water vapor moles entering the equipment cavity using the ideal gas law; the number of water vapor moles is converted into water vapor mass and then divided by the total mass of the dry gas in the cavity to obtain the original humidity ratio; the original humidity ratio is then smoothed and filtered through a time series to obtain the theoretical humidity value.

[0063] The total mass of the drying gas in the cavity is dynamically calibrated by multiplying the cavity volume by the standard dry air density, and the dynamic calibration process incorporates an altitude compensation factor.

[0064] Substituting the gas exchange volume of 0.081 m³ and the current ambient humidity of 65% into the ideal gas law PV=nRT, where the pressure P is taken as standard atmospheric pressure of 101.325 kPa, the temperature T as absolute temperature of 301.15 K (28℃), and the gas constant R as 8.314 J / (mol•K), the calculated number of water vapor moles entering the oil conservator cavity is 3.26 mol. Converting this number of water vapor moles to water vapor mass, multiplying by the molar mass of water of 18.015 g / mol, yields a water vapor mass of 58.7 g. The dry air density corresponding to a 15-liter oil conservator cavity volume under standard conditions is 1.225 kg / m³. However, considering that the warehouse is located at an altitude of 780 meters, an altitude compensation factor of 0.912 is needed for dynamic calibration. The final total mass of the dry gas in the cavity is 15 × 1.225 × 0.912 = 16.8 kg. The original humidity ratio was 58.7g / 16800g = 0.00349, or 0.349%. Due to fluctuations in the instantaneous calculated value, a time-series smoothing filter was used. The original humidity ratio of 12 consecutive sampling points was filtered using the exponential smoothing method, with a filter coefficient set to 0.3, ultimately yielding a theoretical humidity value of 69.8%. The theoretical humidity value reflects the humidity level that the transformer oil conservator cavity should theoretically reach under the current external environmental conditions and equipment sealing state.

[0065] The method of inputting the cavity response features into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity includes:

[0066] Energy extraction was performed on the vibration spectrum of the outer shell at specific frequency bands to separate a subset of vibration features strongly correlated with the motion of gas molecules inside the cavity. A time-frequency joint transformation was performed on the microwave dielectric response data to generate a dielectric loss feature map. The raw vibration signal collected by the vibration sensor contains broadband information from 0-800Hz. Using a specific frequency band energy extraction algorithm, vibration features strongly correlated with the motion of gas molecules inside the cavity within the 15-150Hz frequency band were specifically separated. This is because the vibration in this frequency band is mainly caused by pressure fluctuations due to changes in gas density inside the oil conservator, and is directly related to humidity changes. A 96-dimensional subset of vibration features was extracted through bandpass filtering and energy density analysis. The microwave sensor operates at 2.4GHz, transmitting an 8mW electromagnetic wave signal that penetrates the transformer shell, and the amplitude and phase information of the reflected signal are received. The received microwave dielectric response data is subjected to time-frequency joint transformation. The time-domain signal is converted into a frequency-domain representation using short-time Fourier transform. Then, multi-scale features are extracted by combining wavelet transform to generate a 48×48 pixel dielectric loss feature map, which can reflect the spatial distribution and time-varying characteristics of the dielectric constant of the medium inside the oil conservator. Increased humidity will lead to a significant increase in dielectric loss.

[0067] The vibration feature subset and dielectric loss feature map are fused across modes to obtain a high-dimensional fused feature vector. The high-dimensional fused feature vector is then extracted by a dual-branch feature extraction method of a humidity inversion neural network to obtain a dual-branch output feature vector. The dual-branch output feature vector is then reconstructed by attention weighting to generate a latent space representation of the cavity humidity distribution. The first branch output feature vector of the dual-branch output feature vector is used to extract local spatial features through a convolutional neural network, and the second branch output feature vector is used to capture dynamic evolution features through a temporal coding network.

[0068] The latent space representation of the humidity distribution in the cavity is input into a fully connected neural network layer and mapped to the measured humidity value of the internal cavity.

[0069] The humidity inversion neural network adopts a dual-branch architecture. The first branch uses a three-layer convolutional neural network to process the dielectric loss feature map, with kernel sizes of 5×5, 3×3, and 3×3. Dimensionality is gradually reduced through pooling layers to extract local spatial features and output a 192-dimensional feature vector. The second branch uses a gated recurrent unit network to process the temporal information of vibration features. This network contains 48 hidden units and can capture the dynamic evolution pattern of the cavity breathing effect, outputting a 96-dimensional temporal feature vector. The output feature vectors of the two branches are fused through an attention-weighted reconstruction mechanism. The attention weights are dynamically calculated based on the current monitoring conditions, with a spatial feature weight of 0.6 and a temporal feature weight of 0.4. After fusion, a 288-dimensional latent space representation of the cavity humidity distribution is generated, containing the spatial distribution information and temporal evolution trend of humidity within the cavity. Finally, a three-layer fully connected neural network with 192, 96, and 1 nodes, using ReLU and Sigmoid activation functions, maps the latent space representation of the cavity humidity distribution to a specific humidity value of 74.6%, which is the measured humidity value of the transformer oil conservator cavity.

[0070] The method for fusing the vibration feature subset with the dielectric loss feature map across modes to obtain a high-dimensional fused feature vector includes:

[0071] A modal alignment benchmark is constructed based on the thickness of the metal shell and the distribution pattern of the insulating medium in the cavity of the sealed power equipment. Based on this benchmark, penetration attenuation compensation is applied to a subset of vibration characteristics, and multipath reflection correction is performed on the dielectric loss characteristic spectrum to generate standardized vibration characteristics and standardized dielectric loss characteristics. For example, a transformer shell thickness of 5mm and the relatively simple distribution pattern of the internal insulating oil, insulating paper, and other media constitute the benchmark reference for modal alignment. Vibration signals experience significant attenuation when penetrating the shell. Based on the mechanical impedance characteristics of steel, frequency-by-frequency penetration attenuation compensation is applied to the vibration characteristics in the 15-150Hz frequency band. Low-frequency signals experience relatively less attenuation, while high-frequency signals experience more severe attenuation, with compensation coefficients ranging from 1.5 to 2.8 times. Multipath reflection occurs when microwave signals propagate inside the transformer. Electromagnetic waves are reflected at multiple interfaces such as the oil tank wall, insulator surface, and metal connectors, resulting in a superposition effect of multiple propagation paths in the received signal. A multipath reflection model based on ray tracing was established to calculate the time delay and amplitude attenuation of the main reflection paths. This model was then used to correct the dielectric loss characteristic spectrum, eliminating false high dielectric loss regions and preserving the true characteristics reflecting humidity distribution. To eliminate dimensional differences between different characteristics, the vibration and dielectric loss characteristics needed to be standardized.

[0072] A modal correlation weight matrix was generated by training on a simulation dataset of acoustic-electric coupling in a device cavity. The standardized vibration features and standardized dielectric loss features were dynamically fused to obtain a high-dimensional fused feature vector. During the dynamic feature fusion process, the feature contribution weights of the standardized vibration features and standardized dielectric loss features were assigned according to the modal correlation weight matrix. The simulation dataset of acoustic-electric coupling in a device cavity was generated using simulation technology and is a dataset used to describe the behavioral characteristics of a device cavity under acoustic and electromagnetic coupling. The modal correlation weight matrix was generated by training the simulation dataset. During the simulation, different humidity distributions were set in the oil conservator area of ​​the transformer's 3D model, with 13 gradients from 30% to 90% relative humidity. For each humidity gradient, 40 different spatial distribution patterns were set, forming 520 simulation samples. For each sample, the vibration response and electromagnetic response were calculated simultaneously, and a modal correlation weight matrix including spatial location, frequency characteristics, and amplitude relationships was established. The modal correlation weight matrix describes the correlation strength between vibration features and dielectric features at different spatial locations and frequencies. The standardized vibration features and standardized dielectric loss features were fused into a high-dimensional fused feature vector using a weighted average method.

[0073] The method for allocating feature contribution weights for standardized vibration features and standardized dielectric loss features based on the modal correlation weight matrix during dynamic feature fusion includes:

[0074] The coordinates of key areas of the cavity are analyzed based on the three-dimensional structural model of the sealed power equipment. These key areas include the installation positions of the sealing rings and the core distribution area of ​​the insulating medium. A regional coordinate index table corresponding to the physical space of the cavity is constructed, converting the coordinates of the key areas into index identifiers addressable by the modal correlation weight matrix. Vibration characteristic weight coefficients for the sealing ring area and dielectric characteristic weight coefficients for the insulating medium area are generated based on a historical fault database. For example, in the process of accurately allocating characteristic contribution weights, the coordinates of key areas of the oil conservator cavity are identified by analyzing the three-dimensional CAD model of a 10kV transformer. The sealing ring installation positions include four locations: the sealing ring at the top flange of the oil conservator (coordinates X: 800mm, Y: 600mm, Z: 1400mm) and the sealing ring at the breather connection (coordinates X: 850mm, Y: 550mm, Z: 1450mm). The core distribution area of ​​the insulating medium is mainly concentrated in the lower part of the oil conservator (coordinate range X: 750-900mm, Y: 500-700mm, Z: 1200-1350mm). This area has the highest insulating oil density and is most sensitive to humidity changes. A regional coordinate index table corresponding one-to-one with the physical space of the cavity was constructed, converting the three-dimensional physical coordinates into two-dimensional index identifiers addressable by the modal correlation weight matrix. The physical coordinates were meshed into 18×18 grids with an 8mm step size, with each grid corresponding to an index position in the weight matrix. The index identifiers corresponding to the sealing ring area are (10,7), (11,6), etc., while the index identifiers corresponding to the insulating medium area are multiple positions within the range of (9-11,6-8). Statistical analysis of 800 cases of moisture-induced faults in 10kV transformers collected from the historical fault database revealed that moisture-induced accidents caused by sealing ring failure accounted for 58%, while accidents caused by direct moisture absorption of the insulating medium accounted for 35%. Based on these statistical results, and combined with the sensor characteristic analysis at the time of the fault, a vibration characteristic weighting coefficient of 0.76 was generated for the sealing ring region and a dielectric characteristic weighting coefficient of 0.8 was generated for the insulating medium region. These weighting coefficients reflect the degree of contribution of different regions and different sensing modes to the accuracy of humidity monitoring.

[0075] The vibration characteristic weight coefficient of the sealing ring region and the dielectric characteristic weight coefficient of the insulating medium region are written into the corresponding index positions of the modal correlation weight matrix to form a region-differentiated weight allocation matrix.

[0076] The feature contribution weights are assigned to the standardized vibration characteristics and standardized dielectric loss characteristics based on the regional differential weight allocation matrix.

[0077] When forming the regional differentiated weight allocation matrix, the calculated weight coefficients are written into the corresponding index positions of the modal correlation weight matrix. For the index position (10,7) of the sealing ring area, the vibration feature weight is set to 0.76, and the dielectric feature weight is adjusted accordingly to 0.24, ensuring that the sum of the weights is 1.0. For the index position (10,7) of the insulating medium area, the dielectric feature weight is set to 0.8, and the vibration feature weight is 0.2. For other non-critical areas, the vibration and dielectric feature weights are both set to 0.5 to maintain a balanced distribution. Based on the physical location of the current monitoring point, the regional differentiated weight allocation matrix is ​​queried, and corresponding feature contribution weights are assigned to the standardized vibration feature and standardized dielectric loss feature. For example, when detecting near the top sealing ring of the transformer oil conservator, the contribution weight of the vibration feature is 0.76, and the contribution weight of the dielectric feature is 0.24; when detecting in the insulating medium area in the middle of the oil conservator, the contribution weight of the dielectric feature is 0.8, and the contribution weight of the vibration feature is 0.2. This differentiated weight allocation mechanism based on physical location and historical experience significantly improves the accuracy and reliability of humidity monitoring.

[0078] The method for capturing dynamic evolutionary features through a time-series coding network includes:

[0079] Based on the time-varying characteristics of vibration feature subsets and dielectric loss feature maps, periodic characteristic waveforms synchronized with the breathing effect of the equipment cavity were identified. Key evolution frames within each breathing cycle were extracted, including gas intake characteristics during the temperature drop phase and gas discharge characteristics during the temperature rise phase. The implementation of the time-series coding network focuses on capturing the breathing effect pattern of the transformer oil conservator cavity. During diurnal temperature variations, this 10kV transformer exhibits obvious periodic breathing characteristics: gas expands and discharges during daytime temperature increases, and gas contracts and is drawn in during nighttime temperature decreases. By analyzing vibration feature subsets and dielectric loss feature maps over a continuous 24-hour period, periodic characteristic waveforms synchronized with the breathing effect of the equipment cavity were identified. The breathing cycle is approximately 8 hours, roughly consistent with one-third of the local diurnal temperature variation cycle. Within each breathing cycle, five key evolution frames were extracted as representative features: the initial equilibrium state frame (temperature stabilization period), the gas expansion frame during the temperature rise phase, the gas discharge frame at the highest temperature, the gas contraction frame during the temperature fall phase, and the gas intake frame at the lowest temperature. During the rapid temperature drop phase, the gas intake is characterized by an increase in energy in the vibration spectrum at low frequencies, and the dielectric loss characteristic spectrum shows a decrease in the dielectric constant within the cavity; during the temperature recovery phase, the gas exhaust exhibits the opposite trend.

[0080] After performing cross-cycle consistency alignment on the key evolutionary frames, a gated recurrent unit network (GRN) is used to learn the dynamic correlation patterns between the key evolutionary frames, generating a temporal encoding vector strongly correlated with changes in cavity humidity, which is then denoted as the output feature vector of the second branch. To ensure the comparability of features during different respiratory cycles, cross-cycle consistency alignment is performed on the key evolutionary frames. Using the inflection point of temperature change as the time reference, evolutionary frames from different cycles are aligned on the time axis to eliminate time offsets caused by differences in the rate of change of ambient temperature. The length of the aligned evolutionary frame sequence is uniformly 5 frames. The specific implementation of the GRN network includes 48 hidden units, each of which includes three components: an update gate, a reset gate, and a new memory unit. The network input is the aligned key evolutionary frame sequence. The update gate controls how much memory from the previous moment is retained, the reset gate determines how much past information is discarded, and the new memory unit integrates the current input and the filtered historical information. For example, in the monitoring of this transformer, the network learned a strong correlation between the increase in low-frequency energy of vibration characteristics during the gas inhalation phase and a subsequent 0.6% increase in humidity over 1.5 hours, and a correspondence between the high-frequency attenuation of dielectric characteristics during the gas exhalation phase and a subsequent 0.4% decrease in humidity over 1 hour. After training for six historical breathing cycles, the gated recurrent unit network outputs a 96-dimensional temporal coding vector as the second branch output feature vector. The second branch output feature vector contains the dynamic evolution law of cavity humidity changes and can predict the humidity change trend in the next 1-2 hours under the current breathing phase.

[0081] The method for calculating the cavity risk index based on the absolute deviation between the theoretical humidity value and the measured humidity value, combined with the criticality level coefficient of the sealed electrical equipment, includes:

[0082] The continuous growth trend of the absolute deviation is calculated. When the absolute deviation shows a unidirectional increase over multiple consecutive monitoring periods, the risk accumulation factor is activated. A critical correction coefficient is generated based on the voltage level and structural type of the enclosed electrical equipment. The calculation of the cavity risk index needs to comprehensively consider the severity and persistence of the humidity deviation. During the monitoring of the 10kV transformer, the absolute deviation between the theoretical humidity value of 69.8% and the measured humidity value of 74.6% was 4.8%. Continuous monitoring revealed that in the past three consecutive monitoring periods (5 minutes per period), the absolute deviations were 4.2%, 4.5%, and 4.8%, respectively, showing a clear unidirectional increasing trend, with a deviation growth rate of 0.3% per period. When the absolute deviation shows a unidirectional increase over three consecutive monitoring periods, the risk accumulation factor is activated. The risk accumulation factor is a parameter used to quantify the degree of risk accumulation; its value increases with the continuous increase of the deviation, reflecting the cumulative effect of risk. For example, the risk accumulation factor is 1.8. A critical correction coefficient is generated based on the voltage level and structural type of the 10kV transformer. The basic criticality factor for a 10kV voltage level is 1.2, and the correction factor for an oil-immersed structure is 1.05. Therefore, the criticality correction factor is 1.2 × 1.05 = 1.26, which reflects the impact of the equipment's importance and structural complexity in the power distribution system on risk assessment.

[0083] The equipment body risk baseline value is calculated based on the aforementioned risk accumulation factor and critical correction coefficient; the cavity risk index is then calculated by combining the equipment body risk baseline value with the environmental acceleration factor through a nonlinear coupling function. For example, ; and This is a weighting coefficient used to adjust the contribution of the equipment's inherent risk baseline and the environmental acceleration factor to the cavity risk index. It is adjusted based on actual risk assessment needs and historical data. The adjustment factor is a constant used to fine-tune the entire cavity risk index to ensure it falls within a reasonable range, such as between 0 and 1. The adjustment factor can be set based on experience or historical data, for example, to 0.1, 0.5, or 1.

[0084] The method for calculating the basic risk value of the equipment body based on the risk accumulation factor and the criticality correction coefficient includes:

[0085] An aging attenuation factor is generated based on the service life of the sealing rings in enclosed electrical equipment. This attenuation factor is then used to dynamically adjust the critical correction coefficients. The aging attenuation factor is a parameter used to quantify the degree of performance degradation of equipment components over time. Since the aging of the sealing rings affects the overall sealing performance of the equipment, potentially increasing risks such as internal moisture ingress, dynamic aging correction of the critical correction coefficients is necessary to more accurately reflect the current risk status of the equipment. The introduction of the aging attenuation factor allows the critical correction coefficients to dynamically adjust as the sealing rings age. More severe sealing ring aging leads to a higher risk level.

[0086] When the rate of change of temperature and humidity in the storage environment exceeds the equipment's breathing compensation capability threshold, the credibility verification of the risk accumulation factor is initiated. The fusion weight of the risk accumulation factor is dynamically adjusted based on the breathing compensation capability coefficient. The time-adjusted critical correction coefficient is then weighted and fused with the risk accumulation factor to obtain the fusion result value. The rate of change of temperature and humidity in the storage environment refers to the speed at which temperature and humidity change in the storage environment, reflecting the dynamic nature of environmental conditions. Equipment has a certain self-regulation (or "breathing") capability when responding to changes in environmental temperature and humidity. The breathing compensation capability threshold represents the upper limit of the rate of change of temperature and humidity that the equipment can normally cope with environmental changes without causing significant risks. When the rate of change of temperature and humidity in the storage environment exceeds the equipment's breathing compensation capability threshold, it means that the equipment may not be able to adapt to environmental changes in a timely manner, leading to abnormal fluctuations in internal parameters such as humidity. At this time, the previously calculated risk accumulation factor may no longer be completely accurate, therefore credibility verification needs to be initiated. The breathing compensation capability coefficient is a parameter that quantifies the equipment's breathing compensation capability; the smaller the value, the worse the equipment's breathing compensation capability. The dynamic reduction of the weight of the risk accumulation factor in the subsequent weighted fusion, based on the respiratory compensation capacity coefficient, is intended to reduce the undue influence of the risk accumulation factor on the final risk assessment result when the environment changes too rapidly, because the risk accumulation factor may lose some accuracy due to sudden environmental changes.

[0087] The current humidity deviation curve is matched with typical moisture-related accident patterns using a historical fault case database of enclosed electrical equipment. A risk enhancement coefficient is generated when the similarity exceeds a preset similarity threshold. The fusion result and the risk enhancement coefficient are then mapped nonlinearly to obtain the equipment's risk baseline value. The historical fault case database contains past fault cases of the equipment, recording various environmental parameters, equipment status, and fault modes at the time of the fault. The current humidity deviation curve represents the change in the deviation between the current humidity state and the normal or expected humidity state over time. Typical moisture-related accident patterns are extracted from the historical fault case database and represent typical fault patterns related to equipment moisture, along with their corresponding humidity deviation curve characteristics. Algorithms (such as dynamic time warping algorithms) are used to calculate the similarity between the current humidity deviation curve and typical moisture-related accident patterns to assess the correlation between the current equipment state and historical moisture-related faults. For example, among 800 historical faults, 12 cases are highly similar to the current deviation pattern: the deviation is in the range of 4-5% and increases continuously for more than 3 cycles. The similarity to dampness calculated using the dynamic time warping algorithm is 0.83, exceeding the preset similarity threshold of 0.8, resulting in a risk amplification coefficient of 1.15 and a fusion result value of 1.441. The risk amplification coefficient enhances the accuracy of the current risk assessment, especially when the equipment condition is highly similar to historical failure modes. Increasing the risk amplification coefficient reflects this potential high risk. For example, the following nonlinear mapping function (a variant of the Sigmoid function) is used to calculate the equipment's intrinsic risk baseline. This nonlinear mapping function, through the Sigmoid property, nonlinearly amplifies the influence of the fusion result value and the risk amplification coefficient, allowing the risk baseline value to rise rapidly when the risk is high, thus more accurately reflecting the actual risk of the equipment. , The adjustment coefficients, used to control the influence of the fusion result value and the risk enhancement coefficient on the final risk baseline value, are determined based on actual risk assessment needs and historical data. For example, If the equipment's risk baseline is close to 1, then the equipment's risk baseline value is approximately 0.995. A risk baseline value close to 1 indicates that the equipment's current risk level is relatively high.

[0088] The method for obtaining the environmental acceleration deterioration coefficient includes:

[0089] The warehouse microenvironment monitoring network collects real-time data on the rate of change of ambient temperature and humidity, and calculates the current environmental deterioration intensity based on this data. Historical plum rain erosion records for the enclosed power equipment storage area are retrieved, and a seasonal deterioration correction coefficient is generated by combining this data with meteorological early warning information. Obtaining the environmental acceleration deterioration coefficient requires comprehensive monitoring of the warehouse microenvironment. The warehouse is equipped with a microenvironment monitoring network consisting of 20 sensor nodes, spaced 15 meters apart, covering the entire storage area. Real-time collection of ambient temperature and humidity change rate data at each node revealed that near the 10kV transformer storage location, the temperature change rate was 1.0℃ / h, and the humidity change rate was 3.2% / h, while the warehouse average change rates were 0.7℃ / h and 1.9% / h, respectively, indicating that the environmental changes in this area were more severe than the overall environment. The current environmental degradation intensity value is calculated using a weighted average method: Degradation Intensity Value = (Temperature Change Rate / Baseline Temperature Change Rate) × 0.6 + (Humidity Change Rate / Baseline Humidity Change Rate) × 0.4, where the baseline change rate is the upper limit of normal environmental fluctuations (temperature 0.4℃ / h, humidity 1.2% / h). The calculated degradation intensity value is 2.568. Historical plum rain erosion records for the past four years of the transformer storage area were retrieved, revealing that the average humidity in this area during the plum rain season is 12% higher than normal months, and the average duration of the plum rain season is 38 days. Combined with current weather warnings indicating continuous rainy weather for the next 8 days, with relative humidity remaining above 72%, a seasonal degradation correction coefficient is generated. The seasonal correction coefficient is calculated as: Base coefficient 1.0 + Plum rain intensity factor 0.38 + Duration factor 0.22 = 1.6.

[0090] Sensitivity calibration of the degradation intensity value was performed based on the equipment's sealing rating. The environmental accelerated degradation coefficient was calculated by combining the calibrated degradation intensity value with a seasonal degradation correction factor. Since this 10kV transformer has an IP54 sealing rating (medium sealing), its sensitivity to changes in ambient humidity is moderate. According to the sealing rating sensitivity calibration table, sensitivity calibration was performed on the degradation intensity value: the calibration factor was 1.05 (0.9 for IP65, 1.25 for IP44). The calibrated degradation intensity value was 2.696. The final environmental accelerated degradation coefficient was calculated by combining the calibrated degradation intensity value with the seasonal degradation correction factor: Environmental accelerated degradation coefficient = Calibrated degradation intensity value × Seasonal degradation correction factor = 2.696 × 1.6 = 4.314.

[0091] Example 2: Based on the same inventive concept, such as Figure 2 As shown, this embodiment also provides an intelligent warehouse environment management system based on artificial intelligence. The system includes: an external environment parameter acquisition module, a theoretical humidity calculation module, a measured humidity inversion module, and a cavity risk assessment and active intervention instruction generation module. The modules are connected in sequence for communication.

[0092] The external environment parameter acquisition module is used to collect the external environment parameters of enclosed electrical equipment in the warehouse environment in real time. The external environment parameters include the equipment casing temperature, ambient temperature and ambient humidity.

[0093] The theoretical humidity calculation module is used to obtain the equipment sealing level of the sealed electrical equipment and calculate the theoretical humidity value of the internal cavity of the sealed electrical equipment based on the external environmental parameters and the equipment sealing level.

[0094] The measured humidity inversion module is used to acquire cavity response characteristics through a non-contact sensor attached to the casing of a sealed electrical equipment. The cavity response characteristics include the casing vibration spectrum and microwave dielectric response data. The cavity response characteristics are input into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity.

[0095] The cavity risk assessment and active intervention command generation module is used to calculate the cavity risk index based on the absolute deviation between the theoretical humidity value and the measured humidity value, combined with the criticality level coefficient of the sealed electrical equipment; when the cavity risk index exceeds the preset risk threshold, an active intervention command is generated for the sealed electrical equipment, the active intervention command including activating the directional dehumidification device and adjusting the storage location of the sealed electrical equipment.

[0096] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0097] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent warehouse environment management based on artificial intelligence, characterized in that, The method includes: Real-time acquisition of external environmental parameters of enclosed electrical equipment in the warehouse environment, including equipment casing temperature, ambient temperature and ambient humidity; The sealing rating of the sealed power equipment is obtained, and the theoretical humidity value of the internal cavity of the sealed power equipment is calculated based on the external environmental parameters and the sealing rating. The cavity response characteristics are obtained by non-contact sensors attached to the casing of a sealed electrical equipment. The cavity response characteristics include the casing vibration spectrum and microwave dielectric response data. The cavity response characteristics are then input into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity. Based on the absolute deviation between the theoretical humidity value and the measured humidity value, and combined with the criticality level coefficient of the sealed electrical equipment, the cavity risk index is calculated. When the cavity risk index exceeds the preset risk threshold, an active intervention command is generated for the sealed electrical equipment. The active intervention command includes activating the directional dehumidification device and adjusting the storage location of the sealed electrical equipment.

2. The intelligent warehouse environment management method based on artificial intelligence according to claim 1, characterized in that, The method for calculating the theoretical humidity value of the internal cavity of the sealed power equipment based on the external environmental parameters and the equipment sealing level includes: The temperature difference rate between the equipment casing temperature and the ambient temperature is calculated in real time. When the temperature difference rate exceeds a preset sensitivity threshold, the gas exchange volume per unit time is calculated by a linear-nonlinear hybrid regression algorithm using the temperature difference rate, equipment cavity volume parameters, and basic exchange coefficient. The basic exchange coefficient is determined by mapping the equipment sealing level. The gas exchange volume and ambient humidity are used to calculate the number of water vapor moles entering the equipment cavity using the ideal gas law; the number of water vapor moles is converted into water vapor mass and then divided by the total mass of the dry gas in the cavity to obtain the original humidity ratio; the original humidity ratio is then smoothed and filtered through a time series to obtain the theoretical humidity value. The total mass of the drying gas in the cavity is dynamically calibrated by multiplying the cavity volume by the standard dry air density, and the dynamic calibration process incorporates an altitude compensation factor.

3. The intelligent warehouse environment management method based on artificial intelligence according to claim 1, characterized in that, The method of inputting the cavity response features into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity includes: Energy in specific frequency bands is extracted from the vibration spectrum of the outer shell to separate a subset of vibration features strongly correlated with the motion of gas molecules inside the cavity; the microwave dielectric response data is subjected to time-frequency joint transformation to generate a dielectric loss feature map. The vibration feature subset and dielectric loss feature map are fused across modes to obtain a high-dimensional fused feature vector. The high-dimensional fused feature vector is then extracted by a dual-branch feature extraction method of a humidity inversion neural network to obtain a dual-branch output feature vector. The dual-branch output feature vector is then reconstructed by attention weighting to generate a latent space representation of the cavity humidity distribution. The first branch output feature vector of the dual-branch output feature vector is used to extract local spatial features through a convolutional neural network, and the second branch output feature vector is used to capture dynamic evolution features through a temporal coding network. The latent space representation of the humidity distribution in the cavity is input into a fully connected neural network layer and mapped to the measured humidity value of the internal cavity.

4. The intelligent warehouse environment management method based on artificial intelligence according to claim 3, characterized in that, The method for fusing the vibration feature subset with the dielectric loss feature map across modes to obtain a high-dimensional fused feature vector includes: A modal alignment reference is constructed based on the thickness of the metal shell and the distribution pattern of the insulating medium of the sealed power equipment cavity; based on the modal alignment reference, the vibration feature subset is compensated for through penetration attenuation, and the dielectric loss feature spectrum is corrected for multipath reflection to generate standardized vibration features and standardized dielectric loss features. The modal correlation weight matrix is ​​generated by training the acoustic-electric coupling simulation dataset of the equipment cavity; the standardized vibration features and standardized dielectric loss features are dynamically fused to obtain a high-dimensional fused feature vector, and the feature contribution weights of the standardized vibration features and standardized dielectric loss features are assigned according to the modal correlation weight matrix during the dynamic feature fusion process.

5. The intelligent warehouse environment management method based on artificial intelligence according to claim 4, characterized in that, The method for allocating feature contribution weights for standardized vibration features and standardized dielectric loss features based on the modal correlation weight matrix during dynamic feature fusion includes: The coordinates of key areas of the cavity are analyzed based on the three-dimensional structural model of the sealed power equipment. The key areas include the installation position of the sealing ring and the core distribution area of ​​the insulating medium. A regional coordinate index table corresponding to the physical space of the cavity is constructed, and the coordinates of the key areas are converted into index identifiers that can be addressed by the modal correlation weight matrix. The vibration characteristic weight coefficients of the sealing ring area and the dielectric characteristic weight coefficients of the insulating medium area are generated based on the historical fault database. The vibration characteristic weight coefficient of the sealing ring region and the dielectric characteristic weight coefficient of the insulating medium region are written into the corresponding index positions of the modal correlation weight matrix to form a region-differentiated weight allocation matrix. The feature contribution weights are assigned to the standardized vibration characteristics and standardized dielectric loss characteristics based on the regional differential weight allocation matrix.

6. The intelligent warehouse environment management method based on artificial intelligence according to claim 3, characterized in that, The method for capturing dynamic evolutionary features through a time-series coding network includes: Based on the time-varying characteristics of the vibration feature subset and dielectric loss feature spectrum, periodic feature waveforms synchronized with the breathing effect of the equipment cavity are identified; key evolution frames within each breathing cycle are extracted, the key evolution frames including gas intake features during the temperature drop phase and gas exhaust features during the temperature recovery phase. After performing cross-cycle consistency alignment on the key evolution frames, the dynamic correlation pattern between the key evolution frames is learned through a gated recurrent unit network, generating a temporal coding vector that is strongly correlated with the cavity humidity change and denoted as the output feature vector of the second branch.

7. The intelligent warehouse environment management method based on artificial intelligence according to claim 1, characterized in that, The method for calculating the cavity risk index based on the absolute deviation between the theoretical humidity value and the measured humidity value, combined with the criticality level coefficient of the sealed electrical equipment, includes: The continuous growth trend of the absolute deviation is calculated, and the risk accumulation factor is activated when the absolute deviation shows a unidirectional increase over multiple consecutive monitoring periods; a key correction coefficient is generated based on the voltage level and structural type of the enclosed power equipment. The risk base value of the equipment body is calculated based on the risk accumulation factor and the critical correction coefficient; the risk index of the cavity body is calculated by combining the risk base value of the equipment body with the environmental acceleration deterioration coefficient through a nonlinear coupling function.

8. The intelligent warehouse environment management method based on artificial intelligence according to claim 7, characterized in that, The method for calculating the basic risk value of the equipment body based on the risk accumulation factor and the criticality correction coefficient includes: An aging attenuation factor is generated based on the service life of the sealing ring of the sealed electrical equipment, and the key correction coefficient is dynamically corrected based on the aging attenuation factor. When the rate of change of temperature and humidity in the storage environment exceeds the equipment's breathing compensation capability threshold, the credibility verification of the risk accumulation factor is initiated, and the fusion weight of the risk accumulation factor is dynamically reduced based on the breathing compensation capability coefficient; the critical correction coefficient after timeliness correction is weighted and fused with the risk accumulation factor to obtain the fusion result value. By matching the current humidity deviation curve with the moisture similarity of typical moisture-induced accident modes using a historical fault case database of enclosed power equipment, a risk enhancement coefficient is generated when the moisture similarity is greater than a preset similarity threshold. The fusion result value and the risk enhancement coefficient are then used to obtain the equipment's risk base value through nonlinear mapping.

9. The intelligent warehouse environment management method based on artificial intelligence according to claim 7, characterized in that, The method for obtaining the environmental acceleration deterioration coefficient includes: The warehouse microenvironment monitoring network collects real-time data on the rate of change of ambient temperature and humidity, and calculates the current environmental deterioration intensity value based on the data. Historical plum rain erosion records of the enclosed power equipment storage area are retrieved and combined with meteorological early warning information to generate a seasonal deterioration correction coefficient. The degradation intensity value is sensitively calibrated according to the equipment sealing level; the environmental accelerated degradation coefficient is calculated by combining the calibrated degradation intensity value with the seasonal degradation correction factor.

10. An intelligent warehouse environment management system based on artificial intelligence, characterized in that, The system includes: an external environment parameter acquisition module, a theoretical humidity calculation module, a measured humidity inversion module, and a cavity risk assessment and active intervention command generation module, with each module connected in a sequential communication manner; An external environment parameter acquisition module is used to collect external environment parameters of enclosed electrical equipment in the warehouse environment in real time. The external environment parameters include equipment casing temperature, ambient temperature and ambient humidity. The theoretical humidity calculation module is used to obtain the equipment sealing level of the sealed power equipment and calculate the theoretical humidity value of the internal cavity of the sealed power equipment based on the external environmental parameters and the equipment sealing level. The measured humidity inversion module is used to acquire cavity response characteristics through a non-contact sensor attached to the casing of a sealed electrical equipment. The cavity response characteristics include the casing vibration spectrum and microwave dielectric response data. The cavity response characteristics are input into a pre-trained humidity inversion neural network to obtain the measured humidity value of the internal cavity. The cavity risk assessment and active intervention command generation module is used to calculate the cavity risk index based on the absolute deviation between the theoretical humidity value and the measured humidity value, combined with the criticality level coefficient of the sealed electrical equipment; when the cavity risk index exceeds the preset risk threshold, an active intervention command is generated for the sealed electrical equipment, the active intervention command including activating the directional dehumidification device and adjusting the storage location of the sealed electrical equipment.

Images