Buffer bin dehumidification system

Abstract

The invention provides a surge bin dehumidification system, which relates to the field of intelligent dehumidification and comprises a water tower, a dehumidifier, a cooling water pipe, a humidity sensor and a central control system. The central control system obtains a time queue of real-time humidity values, compares the real-time humidity values with preset expected humidity values, and then carries out segmentation and local time sequence mode coding, so that brake starting judgment of the dehumidifier is achieved according to humidity deviation dynamic causal transmission representation between humidity expected deviation local time sequence mode characteristics. Therefore, the accurate expected humidity value is set, and the time sequence trend and mode of the actual humidity deviation change can be accurately identified, so that an over-sensitive response mechanism is avoided, and the control precision is improved. And meanwhile, misjudgment caused by noise is reduced, so that more efficient, more accurate and more energy-saving humidity control is realized.

Description

Technical Field

[0001] The present application relates to the field of intelligent dehumidification, and more particularly, in an embodiment of the present application, to a buffer bin dehumidification system. Background Art

[0002] In modern warehousing and logistics, buffer warehouses are used to temporarily store various materials, which have strict requirements on the humidity of the storage environment. Too high humidity may cause the product to spontaneously combust, get damp, or mildew, affecting its use effect and safety; while too low humidity may cause the product to dry out and crack, which is also not conducive to product preservation. Therefore, it is very important to maintain an appropriate humidity range, which can not only ensure product quality and extend shelf life, but also improve the safety of the storage environment and ensure the best condition of materials.

[0003] However, in the traditional control mode, when the humidity exceeds the standard, the control system will immediately start the dehumidifier. This highly sensitive response mechanism causes the dehumidifier to start and stop frequently, which not only increases energy consumption, but also may significantly shorten the service life of the equipment. In addition, the humidity sensor may collect noise due to environmental interference (such as temperature fluctuations, electromagnetic interference, etc.). These noises may cause the system to misjudge the humidity status, thereby triggering unnecessary dehumidification operations. These misoperations not only increase equipment wear and energy waste, but may also cause humidity fluctuations, which have a negative impact on the quality and safety of stored items.

[0004] Therefore, an optimized buffer bin dehumidification solution is desired. Summary of the invention

[0005] In order to solve the above-mentioned technical problems, the present application is proposed. An embodiment of the present application provides a buffer bin dehumidification system, which first includes a water tower, a dehumidifier, a cooling water pipe, a humidity sensor and a central control system. The central control system obtains the time queue of the real-time humidity value and compares it with the preset expected humidity value, and then performs segmentation and local timing pattern encoding, so as to realize the braking start judgment of the dehumidifier according to the dynamic causal transmission representation of the humidity deviation between the local timing pattern characteristics of each humidity expected deviation. In this way, by setting a precise expected humidity value and being able to accurately identify the timing trend and pattern of the actual humidity deviation change, an overly sensitive response mechanism is avoided and the control accuracy is improved. At the same time, the misjudgment caused by noise is reduced, thereby achieving more efficient, accurate and energy-saving humidity control.

[0006] According to one aspect of the present application, a buffer silo dehumidification system is provided, comprising a water tower, a dehumidifier connected to the water tower through a cooling water pipe, the water tower is used to provide a cooling water source for the dehumidifier, a humidity sensor deployed in the buffer silo, and a central control system, wherein the central control system comprises:

[0007] An expected humidity value setting module, used for setting an expected humidity value;

[0008] A real-time humidity value acquisition module, used to obtain a time queue of real-time humidity values ​​acquired by the humidity sensor;

[0009] A humidity expected deviation value calculation module, used for calculating the difference between the real-time humidity value at each time point in the time queue of the real-time humidity value and the expected humidity value to obtain a time queue of the humidity expected deviation value;

[0010] A humidity expectation deviation local encoding module, used for performing sequence segmentation and local temporal pattern feature extraction on the time queue of the humidity expectation deviation value to obtain a set of humidity expectation deviation local temporal pattern feature encoding vectors;

[0011] A humidity local time series dynamic transfer module is used to perform a humidity local time series deviation dynamic causal accumulation analysis on the set of humidity expected deviation local time series pattern feature coding vectors to obtain a humidity deviation time series pattern transfer coding vector;

[0012] The dehumidifier control module is used to transmit the coding vector based on the humidity deviation timing mode to obtain a control instruction indicating whether to start the dehumidifier.

[0013] Compared with the prior art, the present application provides a buffer bin dehumidification system, which first includes a water tower, a dehumidifier, a cooling water pipe, a humidity sensor and a central control system. The central control system obtains the time queue of the real-time humidity value and compares it with the preset expected humidity value, and then performs segmentation and local timing pattern encoding, so as to realize the braking start judgment of the dehumidifier according to the dynamic causal transmission representation of the humidity deviation between the local timing pattern characteristics of each humidity expected deviation. In this way, by setting a precise expected humidity value and being able to accurately identify the timing trend and pattern of the actual humidity deviation change, an overly sensitive response mechanism is avoided and the control accuracy is improved. At the same time, the misjudgment caused by noise is reduced, thereby achieving more efficient, accurate and energy-saving humidity control. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0015] Figure 1 It is a block diagram of the central control system in the buffer bin dehumidification system according to an embodiment of the present application.

[0016] Figure 2 This is a schematic diagram of data flow of a central control system in a buffer bin dehumidification system according to an embodiment of the present application.

[0017] Figure 3 It is a block diagram of a humidity expected deviation local encoding module in a buffer bin dehumidification system according to an embodiment of the present application.

[0018] Figure 4 It is a block diagram of a humidity local timing dynamic transmission module in a buffer bin dehumidification system according to an embodiment of the present application.

[0019] Figure 5 It is a block diagram of a humidity expectation deviation causal correlation topological feature construction unit in a buffer bin dehumidification system according to an embodiment of the present application.

[0020] Figure 6 It is a front structural schematic diagram of a buffer bin dehumidification system according to another embodiment of the present application.

[0021] Figure 7 Schematic diagram of the internal structure of a buffer bin dehumidification system according to another embodiment of the present application.

[0022] Figure 8 It is a side structural schematic diagram of a buffer bin dehumidification system according to another embodiment of the present application.

[0023] In the figure: 1. Dehumidifier No. 1; 2. Water tower No. 1; 3. Warehouse inlet air valve; 4. Air duct; 5. Warehouse air outlet; 6. Dehumidifier No. 2; 7. Water tower No. 2; 8. Cooling tower; 9. Cooling water pipe No. 2; 10. Cooling water pipe No. 1; 11. Dehumidifier air valve; 12. Warehouse sensor; 13. Central control system. DETAILED DESCRIPTION

[0024] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0025] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0026] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0027] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0028] In modern warehousing and logistics, material storage has very strict requirements on humidity. Too high or too low humidity can cause damage to items. Too high humidity may cause the product to spontaneously combust, get damp, or become moldy, affecting its use and safety; while too low humidity may cause the product to dry out and crack, which is also not conducive to product preservation. Traditional buffer warehouse dehumidification methods usually use ventilation, heating, or the use of special dehumidification equipment, but these methods have problems such as high energy consumption and unstable effects, which do not meet the needs of modern warehouse management for energy conservation, emission reduction, and intelligent control.

[0029] In view of the above problems, the present application proposes a buffer silo dehumidification system, including a water tower, a dehumidifier connected to the water tower through a cooling water pipe, the water tower is used to provide cooling water for the dehumidifier, a humidity sensor deployed in the buffer silo, and a central control system. It should be understood that in the technical solution of the present application, a large amount of water is usually stored inside the water tower. After cooling, the water can be transported to the cooling system of the dehumidifier through a pipeline to absorb the heat generated when the dehumidifier is working. The water tower is connected to the dehumidifier through a cooling water pipe to ensure that the performance of the dehumidifier will not be affected by overheating during operation. Through the continuous circulation of the cooling water source, the dehumidifier is helped to maintain a suitable operating temperature to prevent it from malfunctioning due to excessive temperature during long-term operation. The cooling water pipe is responsible for transporting the cooled water source to the dehumidifier to ensure the stable operation of the system. Further, the dehumidifier is responsible for reducing the humidity level in the buffer silo. The dehumidifier dissipates heat through the cooling water provided by the water tower to ensure that it can still work stably in a high temperature environment. Based on the principle of freezing and dehumidification, the dehumidifier uses a compression refrigeration system to forcefully cool the air entering the dehumidifier, so that a large amount of water vapor in it condenses into droplets, which are separated and discharged by a special gas-liquid separator. The dew point of the finished gas can reach 5-10°C. Furthermore, the humidity sensor is a monitoring device in the system, which is deployed in the buffer warehouse to monitor the humidity changes in the warehouse in real time. The function of the humidity sensor is to sense the moisture content in the air in the warehouse and feed the data back to the central control system. Through this monitoring, the system can start or adjust the operating status of the dehumidifier when the humidity reaches the set threshold, so as to keep the humidity in the warehouse within the set range and ensure the stability of the environmental conditions. The central control system is the brain of the dehumidification system, which is responsible for receiving signals from the humidity sensor and controlling the switch, operation mode and cooling water flow of the dehumidifier according to the actual humidity conditions. The central control system intelligently adjusts the dehumidifier according to the set humidity range to ensure that the environment is always in the best humidity state.

[0030] Correspondingly, in the central control system, the technical concept of the present application is to set the expected humidity value, and obtain the time queue of the real-time humidity value collected by the humidity sensor, and then calculate the difference between each real-time humidity value and the expected humidity value to obtain the time queue of the expected humidity deviation value, and use artificial intelligence-based data processing and timing coding technology to segment and encode the time queue of the expected humidity deviation value, so as to realize the braking start judgment of the dehumidifier according to the dynamic causal transmission representation of the humidity deviation between the local timing pattern characteristics of each humidity expected deviation. The present application avoids an overly sensitive response mechanism and improves control accuracy by setting a precise expected humidity value and accurately identifying the timing trend and pattern of the actual humidity deviation change. At the same time, it reduces misjudgments caused by noise, thereby achieving more efficient, accurate and energy-saving humidity control.

[0031] Figure 1 It is a block diagram of the central control system in the buffer bin dehumidification system according to an embodiment of the present application. Figure 2 Schematic diagram of data flow in the central control system of the buffer silo dehumidification system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the central control system 100 of the buffer bin dehumidification system of the embodiment of the present application, it includes: an expected humidity value setting module 110, which is used to set the expected humidity value; a real-time humidity value acquisition module 120, which is used to obtain the time queue of the real-time humidity value collected by the humidity sensor; a humidity expected deviation value calculation module 130, which is used to calculate the difference between the real-time humidity value at each time point in the time queue of the real-time humidity value and the expected humidity value to obtain the time queue of the humidity expected deviation value; a humidity expected deviation local encoding module 140, which is used to perform sequence segmentation and local temporal pattern feature extraction on the time queue of the humidity expected deviation value to obtain a set of humidity expected deviation local temporal pattern feature encoding vectors; a humidity local temporal dynamic transfer module 150, which is used to perform humidity local temporal deviation dynamic causal accumulation analysis on the set of the humidity expected deviation local temporal pattern feature encoding vectors to obtain a humidity deviation temporal pattern transfer encoding vector; a dehumidifier control module 160, which is used to obtain a control instruction for indicating whether to start the dehumidifier based on the humidity deviation temporal pattern transfer encoding vector.

[0032] In the buffer dehumidification system 100, the expected humidity value setting module 110 is used to set the expected humidity value. It should be understood that, considering that different materials have different sensitivity and suitable range to humidity, food, medicine and other moisture-sensitive items require a lower humidity environment to prevent mildew or chemical reactions; while wood, paper products and the like require a higher humidity to avoid drying and cracking. By setting the expected humidity value, the expected humidity value can be flexibly set according to the needs of specific storage materials, thereby meeting the refined management requirements in different scenarios.

[0033] In the above buffer dehumidification system 100, the real-time humidity value acquisition module 120 is used to obtain the time queue of the real-time humidity value collected by the humidity sensor. It should be understood that the humidity sensor has the characteristics of fast response, high resolution and strong anti-interference ability to ensure that the subtle changes in humidity in the buffer can be accurately captured. In addition, in the technical solution of the present application, the sensors are evenly distributed in the buffer, and try to avoid the vents, heating equipment and other locations that may cause local temperature and humidity fluctuations, so as to obtain more representative humidity data. Among them, considering that in a high humidity environment, the sensor may have a measurement deviation due to the influence of condensed water, it is necessary to select a model with an anti-condensation design; and in an industrial environment with strong electromagnetic interference, sensors with shielding function should be given priority. Next, a reasonable data acquisition frequency needs to be set. The selection of the acquisition frequency needs to balance the relationship between data accuracy and system load. If the acquisition frequency is too low, important humidity change information may be missed, resulting in the time queue being unable to truly reflect the environmental state; while a too high acquisition frequency will increase the pressure of data transmission and storage, and may introduce more noise interference. Specifically, for highly sensitive items that are susceptible to humidity, a higher acquisition frequency (such as once per second) can be set to capture humidity fluctuations in a timely manner; for items with higher stability, a lower frequency (such as once per minute) can be selected to reduce the system burden. Then, after the data acquisition is completed, the sensor data is transmitted to the central control system through the communication protocol. Modern humidity sensors usually support a variety of communication methods, including wired communication (such as RS-485, Modbus) and wireless communication (such as Wi-Fi, Zigbee). In actual deployment, in situations where wiring is convenient and real-time requirements are high, wired communication becomes the first choice due to its stability and low latency; in large-scale distributed warehousing environments, wireless communication has more advantages due to its flexibility and scalability. It is worth mentioning that the raw data collected by the sensor often contains a certain amount of noise components, which may come from environmental interference (such as temperature fluctuations, electromagnetic radiation) or measurement errors of the device itself. In order to improve the accuracy of the time queue, the technical solution of the present application preprocesses and filters the noise of the raw data. Commonly used noise filtering methods include sliding average filtering, Kalman filtering, and wavelet transform. Among them, the sliding average filter can effectively eliminate the influence of random noise by averaging the data within a certain period of time; while the Kalman filter uses the prediction and correction mechanism to show stronger adaptability in a dynamically changing environment. Through these technical means, the smoothness and credibility of the time queue can be significantly improved. The noise-filtered data will be further organized into a time queue and stored in the central control system. The time queue is a data structure arranged in chronological order, which can clearly show the trend of humidity changes over time. Among them, the time queue usually records a humidity value at a fixed time interval (such as every second or every minute) with a corresponding timestamp.This structured storage method is not only convenient for query and backtracking, but also provides basic support for subsequent sequence segmentation and feature extraction. In this way, the time queue for obtaining the real-time humidity value collected by the humidity sensor covers multiple aspects of work, from sensor selection to data acquisition frequency setting, to communication protocol selection, noise filtering, and data storage. Through scientific planning and meticulous operation, a high-quality time queue can be generated.

[0034] In the above buffer dehumidification system 100, the humidity expected deviation value calculation module 130 is used to calculate the difference between the real-time humidity value at each time point in the time queue of the real-time humidity value and the expected humidity value to obtain the time queue of the humidity expected deviation value. It should be understood that in order to more clearly quantify the gap between the current environmental humidity and the target humidity, so as to better understand and analyze the time series dynamic change of the humidity real-time deviation value, in the technical solution of the present application, the difference between the real-time humidity value at each time point in the time queue of the real-time humidity value and the expected humidity value is calculated to obtain the time queue of the humidity expected deviation value. That is, by calculating the difference, the deviation of the humidity can be quantified, and a specific value can be used to clearly show the degree of deviation of the actual humidity at each time point compared with the expected humidity, so as to reflect the change of the humidity deviation over time. Among them, the real-time humidity value is obtained by continuous monitoring by the humidity sensor, reflecting the actual humidity in the environment, and the expected humidity value is the set ideal humidity value, which is usually set according to work requirements or environmental conditions. The difference between the two, that is, the humidity expected deviation value, can reflect the accuracy and efficiency of system regulation. If the humidity expected deviation value is large, it means that the gap between the current ambient humidity and the target value is large, and it may be necessary to increase the working intensity of the dehumidifier or adjust other control parameters to ensure that the humidity can be restored to the expected value as soon as possible. On the contrary, if the deviation value is small, it means that the humidity control system has achieved good stability. In this way, the calculation of the humidity expected deviation value can help understand the degree of deviation between the current humidity level and the target humidity level.

[0035] In the buffer dehumidification system 100, the humidity expected deviation local encoding module 140 is used to perform sequence segmentation and local time series pattern feature extraction on the time queue of the humidity expected deviation value to obtain a set of humidity expected deviation local time series pattern feature encoding vectors. It should be understood that the time queue of the humidity expected deviation value usually contains complex dynamic changes in the humidity control process. A single global analysis may not reveal the underlying laws and changing trends. Therefore, it is necessary to capture the local time series pattern through segmentation and feature extraction, extract more meaningful time series features from the humidity deviation data, and help the system respond and adjust better.

[0036] Figure 3FIG. 4 is a block diagram of a humidity expectation deviation local encoding module in a buffer silo dehumidification system according to an embodiment of the present application. Figure 3 As shown, in an embodiment of the present application, the humidity expected deviation local encoding module 140 includes: a humidity expected deviation equal-length sequence segmentation unit 141, which is used to perform equal-length sequence segmentation on the time queue of the humidity expected deviation value to obtain a set of humidity expected deviation local time series sequences; a humidity expected deviation local time series sequence feature extraction unit 142, which is used to input each humidity expected deviation local time series sequence in the set of humidity expected deviation local time series sequences into a humidity deviation time series pattern feature extractor based on a forward LSTM model to obtain a set of humidity expected deviation local time series pattern feature encoding vectors.

[0037] Specifically, the humidity expected deviation equal-time sequence segmentation unit 141 is used to segment the time queue of the humidity expected deviation value into equal-time sequence to obtain a set of humidity expected deviation local time series sequences. It should be understood that considering that the time queue of the humidity expected deviation value is usually a continuous dynamic data stream, it may contain complex fluctuations and change patterns, which will increase the complexity of the data and make it difficult to capture the local change rules. In other words, different time series pattern feature information exists in different local time segments, such as local periodicity or burst characteristics. Therefore, in order to be able to analyze and process each local time segment more carefully, in the technical solution of the present application, the time queue of the humidity expected deviation value is segmented into equal-time sequence to convert the long sequence data into a relatively short local sequence set, and obtain a set of humidity expected deviation local time series sequences. In this way, each subsequence can have a similar time scale, thereby ensuring that the change trend of the humidity deviation data in each local sequence is more stable, which is convenient for extracting regularity features. Specifically, in a specific embodiment of the present application, a time queue of the humidity expected deviation value is first received, wherein each element contains the humidity expected deviation value and its corresponding timestamp. Then, in order to achieve equal-length segmentation, the unit needs to set a fixed window length and divide the time queue into several subsequences to ensure that the time span of each subsequence is exactly the same. Among them, if the time length of the last subsequence is less than the window length, you can choose to truncate or fill it. Then, in order to improve the flexibility and adaptability of segmentation, the technical solution of the present application supports the function of dynamically adjusting the window length. This design enables the system to select different granularities for analysis according to actual needs. For example, when it is necessary to pay attention to short-term fluctuations, a shorter window length can be set; when studying long-term trends, a longer window length can be selected. In addition, the technical solution of the present application also supports a sliding window mechanism, which generates partially overlapping subsequences by setting step parameters (such as moving 1 time point each time) to capture more detailed information. It is worth noting that during the segmentation process, timestamps may be missing or discontinuous. At this time, it is necessary to adopt interpolation or remove outliers to repair them to ensure the quality of each local time series. In this way, by segmentation, complex time series data can be decomposed into small fragments that are easy to analyze, providing basic support for subsequent feature extraction and pattern recognition.

[0038] Specifically, the humidity expected deviation local time series feature extraction unit 142 is used to input each humidity expected deviation local time series sequence in the set of the humidity expected deviation local time series sequence into the humidity deviation time series pattern feature extractor based on the forward LSTM model to obtain a set of humidity expected deviation local time series pattern feature encoding vectors. It should be understood that each humidity expected deviation local time series sequence represents the change of humidity deviation in a specific time period. By analyzing these local sequences, specific patterns of humidity changes in different time periods can be mined, such as local periodic fluctuations, mutation points, etc. For example, the discovery of similar periodic humidity deviation changes in several local sequences may mean that there is a factor that periodically affects humidity in the buffer bin. Based on this, the present application inputs each humidity expected deviation local time series sequence in the set of the humidity expected deviation local time series sequence into the humidity deviation time series pattern feature extractor based on the forward LSTM model to utilize the gating mechanism (input gate, forget gate, output gate) of the forward LSTM model to effectively capture long-term dependencies, thereby mining the hidden nonlinear patterns in the data, such as the periodic changes of humidity deviation, sudden rising or falling trends, and other features, to obtain a set of humidity expected deviation local time series pattern feature encoding vectors. The obtained features can reveal the changing characteristics of humidity in different time periods, such as the rate of change and stability of humidity deviation, to accurately represent the local pattern of humidity change.

[0039] In the above-mentioned buffer dehumidification system 100, the humidity local time series dynamic transfer module 150 is used to perform a humidity local time series deviation dynamic causal accumulation analysis on the set of the humidity expected deviation local time series pattern feature coding vectors to obtain a humidity deviation time series pattern transfer coding vector. It should be understood that, considering that each humidity expected deviation local time series pattern feature coding vector reflects the local time series characteristics of the humidity deviation in each local time period, the independent local characteristics cannot reflect the dynamic causal relationship between the humidity deviations in different time periods. For example, the sudden increase in humidity deviation at a certain moment may be the result of the accumulation of slow changes in humidity in the previous time periods. Therefore, in order to reveal this hidden causal relationship between different local time periods and fully understand the evolution of humidity deviation over time, the present application performs a humidity local time series deviation dynamic causal accumulation analysis on the set of the humidity expected deviation local time series pattern feature coding vectors to obtain a humidity deviation time series pattern transfer coding vector. That is, through this dynamic causal accumulation analysis mechanism, the inherent driving relationship and key information of the complex characteristics of humidity deviation can be captured more efficiently to reduce noise interference, and at the same time, it no longer emphasizes only the explicit local pattern, but also fully captures the hidden global dependency, making humidity control more accurate and reliable.

[0040] Figure 4FIG. 1 is a block diagram of a humidity local time series dynamic transmission module in a buffer bin dehumidification system according to an embodiment of the present application. Figure 4 As shown, in an embodiment of the present application, the humidity local temporal dynamic transfer module 150 includes: a humidity expected deviation local temporal pattern implicit feature mining unit 151, which is used to perform implicit feature mining on each humidity expected deviation local temporal pattern feature coding vector in the set of the humidity expected deviation local temporal pattern feature coding vector to obtain a set of humidity expected deviation local temporal depth implicit feature coding vectors; a humidity expected deviation causal association topological feature construction unit 152, which is used to construct the causal association topological features of the set of the humidity expected deviation local temporal depth implicit feature coding vectors to obtain a humidity expected deviation semantic causal association topological feature matrix; a humidity expected deviation context dynamic walking fusion unit 153, which is used to use the humidity expected deviation semantic causal association topological feature matrix as the semantic causal association structure information, and perform context dynamic walking fusion on the set of the humidity expected deviation local temporal pattern feature coding vectors and the set of the humidity expected deviation local temporal depth implicit feature coding vectors to obtain the humidity deviation temporal pattern transfer coding vector.

[0041] Specifically, the humidity expected deviation local time series pattern implicit feature mining unit 151 is used to perform implicit feature mining on each humidity expected deviation local time series pattern feature coding vector in the set of humidity expected deviation local time series pattern feature coding vectors to obtain a set of humidity expected deviation local time series depth implicit feature coding vectors, which is expressed by the implicit feature mining formula:

[0042] O={x1,x2,...,x i ,...,x n}

[0043] v i =Sigmoid[Conv 1×1 (x i )]

[0044] D={v1,v2,...,v i ,...,v n}

[0045] Wherein, O is the set of the humidity expected deviation local temporal pattern feature encoding vectors, x1, x2, x i and x n are the first, second, i-th and n-th humidity expected deviation local time series pattern feature encoding vectors in the set of humidity expected deviation local time series pattern feature encoding vectors, Conv 1×1 is the point convolutional coding, Sigmoid is the convolutional coding activation function, v1, v2, v i, v j and v n are the first, second, i-th, j-th and n-th humidity expected deviation local time series depth implicit feature encoding vectors in the set of the humidity expected deviation local time series depth implicit feature encoding vectors, and D is the set of the humidity expected deviation local time series depth implicit feature encoding vectors. It should be understood that although the humidity expected deviation local time series pattern feature encoding vector has captured the explicit features of some time series through sequence segmentation and feature extraction, these features are often limited to the surface level and cannot fully reflect the internal mechanism of humidity change. By mining the implicit features of the encoding vector, these potential correlations and regularities can be discovered, thereby more accurately describing the essence of humidity change. Among them, in the initial stage, deep semantic features are extracted from the set of input humidity expected deviation local time series pattern feature encoding vectors through implicit feature mining. The essence of implicit modeling is to alleviate the noise problem and obvious limitations of explicit features, and to generate a more extensive semantic representation by mapping the original features to the latent space, in which important information is efficiently compressed and redundant information is eliminated.

[0046] Figure 5 FIG. 1 is a block diagram of a humidity expectation deviation causal correlation topological feature construction unit in a buffer silo dehumidification system according to an embodiment of the present application. Figure 5 As shown, in an embodiment of the present application, the humidity expectation deviation causal association topological feature construction unit 152 includes: a humidity expectation deviation semantic causal association factor calculation subunit 1521, which is used to calculate the semantic causal association factor between any two humidity expectation deviation local temporal depth implicit feature coding vectors in the set of humidity expectation deviation local temporal depth implicit feature coding vectors to obtain a humidity expectation deviation semantic causal association topological matrix composed of multiple humidity expectation deviation semantic causal association factors; a humidity expectation deviation semantic causal triggering subunit 1522, which is used to input the humidity expectation deviation semantic causal association topological matrix into a causal triggering network based on a gated activation function to obtain the humidity expectation deviation semantic causal association topological feature matrix.

[0047] In an embodiment of the present application, the humidity expected deviation semantic causal association factor calculation subunit 1521 is used to: calculate the association matrix between any two humidity expected deviation local temporal depth implicit feature coding vectors in the set of humidity expected deviation local temporal depth implicit feature coding vectors to obtain a set of humidity expected deviation association matrices; calculate the semantic causal association factor of each humidity expected deviation association matrix in the set of humidity expected deviation association matrices to obtain the humidity expected deviation semantic causal association topological matrix composed of multiple humidity expected deviation semantic causal association factors, and the humidity expected deviation semantic causal association factor is calculated by the mean, variance, maximum value and causal association bias value of the humidity expected deviation association matrix; wherein, in response to the variance of the humidity expected deviation association matrix being greater than or equal to a predetermined threshold, the weighted average of the distances between any two humidity expected deviation local temporal depth implicit feature coding vectors in the set of humidity expected deviation local temporal depth implicit feature coding vectors is used as the causal association bias value; in response to the variance of the humidity expected deviation association matrix being less than the predetermined threshold, the weighted value of the variance of the humidity expected deviation association matrix is ​​used as the causal association bias value.

[0048] Specifically, the humidity expectation deviation semantic causal association factor calculation subunit is expressed as follows:

[0049]

[0050] Among them, v j T Yes j The transposed vector, M i-j Yes i and v j The humidity expectation deviation correlation matrix between is the matrix multiplication, σ 2 (M i-j ) is M i-j The variance of max(M i-j ) is to take M i-j The maximum value in μ(M i-j ) is M i-j The mean of , λ is the causal association bias value, d(v i ,v j ) is v i and v j The distance between them, L is the number of vectors in D, ε is the predetermined threshold, α and β are weighted hyperparameters, t i-j It is M i-j The corresponding humidity expectation deviation semantic causal correlation factor, t 1-1 , t 1-n , t n-1 and t n-nare respectively the humidity expectation deviation semantic causal association factors of each position in the humidity expectation deviation semantic causal association topological matrix, and T is the humidity expectation deviation semantic causal association topological matrix.

[0051] It should be understood that after the implicit feature mining is completed, the model enters the semantic causal association calculation stage. The goal of this step is to derive the causal association between each pair of features from the set of humidity expected deviation local temporal depth implicit feature coding vectors. In the specific implementation, this process can be implemented in combination with causal modeling technology. In particular, in the technical solution of the present application, it constructs a humidity expected deviation association matrix between any two humidity expected deviation local temporal depth implicit feature coding vectors in the set of humidity expected deviation local temporal depth implicit feature coding vectors, and uses a causal association energy measurement function to express the causal association between any two humidity expected deviation local temporal depth implicit feature coding vectors in the set of humidity expected deviation local temporal depth implicit feature coding vectors in explicit quantitative coding to obtain the humidity expected deviation semantic causal association factor. More specifically, by calculating the association matrix, the similarity and degree of mutual influence of humidity changes in different time periods can be quantified.

[0052] Next, by calculating the semantic causal correlation factor of the humidity expectation deviation correlation matrix, the causal relationship between humidity expectation deviations can be further revealed. The calculation of the semantic causal correlation factor takes into account the statistical characteristics of the humidity expectation deviation correlation matrix, such as the mean, variance, maximum value, and causal correlation bias value. These indicators can effectively describe the causal relationship and trend changes of humidity change patterns in time series.

[0053] In particular, the introduction of causal association bias values ​​can adjust the potential bias in the data, making the analysis of causal relationships more accurate and reliable. Specifically, in the analysis of humidity expectation deviation, the variance of the humidity expectation deviation association matrix reflects the amplitude and volatility of the changes between humidity sequences. An association matrix with a large variance indicates that the changes between humidity expectation deviation sequences are more significant, which usually means that the pattern of humidity change is more complex or unstable, so a more detailed analysis of these changes is required. In this case, in order to adjust the causal relationship analysis model of humidity expectation deviation, by considering the low-level causal associations in complex systems as molecular-level relationships inferred based on statistical correlation, the intervention prediction of causal association energy can be further performed on the basis of the global fine-grained statistical association representation, so as to study the fine-grained structure of causal association and its dynamic regulation based on the high-dimensional and heterogeneous representation of causal omics. Among them, when the aggregation distribution representation of the causal graph is greater than the predetermined threshold, the bias is performed when integrating the source data based on the matrix graph node effect representation of the semantic causal association factor of the humidity expectation deviation, and when the aggregation distribution representation of the causal graph is less than the predetermined threshold, the condensed structure modeling can be directly performed through the integration and compression of the feature pattern. In this way, not only can the causal energy in the description system be encoded, but also the implicit causal intervention prediction results can be condensed, thereby obtaining a more efficient disclosure of key causal relationships. Therefore, by processing the humidity expectation deviation association matrix, a set of humidity expectation deviation semantic causal association factors is obtained, and these factors are organized into a humidity expectation deviation semantic causal association topological matrix. This topological matrix is ​​based on causal relationships and shows the temporal sequence and causal relationship between humidity expectation deviations.

[0054] Specifically, the humidity expectation deviation semantic causal trigger subunit 1522 is used to input the humidity expectation deviation semantic causal association topological matrix into the causal trigger network based on the gated activation function to obtain the humidity expectation deviation semantic causal association topological feature matrix, which is expressed as the causal trigger formula:

[0055]

[0056] Among them, softmax is a nonlinear activation function, τ is a normalized threshold, and f trigger(T) is the gated activation processing of T, and M is the semantic causal association topological feature matrix of humidity expectation deviation. It should be understood that the semantic causal association topological matrix of humidity expectation deviation has revealed the causal association and temporal sequence between humidity expectation deviation sequences through previous analysis. However, relying solely on the original form of the association matrix may not be able to fully capture the complex dynamic causal mechanism. Therefore, by inputting the topological matrix into the causal trigger network based on the gated activation function, the potential patterns of these causal relationships can be deeply understood and mined from multiple dimensions through the nonlinear transformation ability of the activation function. The role of the causal trigger network based on the gated activation function is to perform more detailed feature extraction and modeling of causal topological relationships. Its core is the dynamic gating mechanism and nonlinear activation function to dynamically trigger the topological matrix representation. The gating mechanism allows the model to distinguish key causal paths in a dynamic context, thereby strengthening important associations and reducing noise interference. At the same time, the activation function in the network enhances the model's expressiveness by introducing nonlinearity, capturing high-order regularities hidden in complex causal structures. In this way, through the gating mechanism, the model can adaptively adjust according to the input causal topology matrix, actively capture those key time dependencies and causal paths in the humidity change process, so as to effectively control and adjust the transmission of information flow, screen out the features most relevant to the causal relationship of humidity expectation deviation, and enhance the prediction ability of the dynamic process of humidity change.

[0057] In an embodiment of the present application, the humidity expectation deviation context dynamic walking fusion unit 153 is used to: input the humidity expectation deviation semantic causal association topological feature matrix and the set of humidity expectation deviation local temporal pattern feature coding vectors into the feature sequence dynamic walking encoder based on the graph convolutional neural network model to obtain the humidity expectation deviation surface context dynamic walking semantic coding vector; input the humidity expectation deviation semantic causal association topological feature matrix and the set of humidity expectation deviation local temporal depth implicit feature coding vectors into the feature sequence dynamic walking encoder based on the graph convolutional neural network model to obtain the humidity expectation deviation hidden context dynamic walking semantic coding vector; fuse the humidity expectation deviation hidden context dynamic walking semantic coding vector and the humidity expectation deviation surface context dynamic walking semantic coding vector to obtain the humidity deviation temporal pattern transfer coding vector.

[0058] Specifically, the set of the humidity expectation deviation semantic causal association topological feature matrix and the humidity expectation deviation local temporal pattern feature encoding vector is input into the feature sequence dynamic walk encoder based on the graph convolutional neural network model to obtain the humidity expectation deviation surface context dynamic walk semantic encoding vector, which is expressed as the humidity expectation deviation surface formula:

[0059]

[0060] Among them, GCN is a graph convolution process, H surface It is the dynamic walk semantic encoding vector of the surface context of humidity expectation deviation. It should be understood that after the causal trigger network based on the gated activation function deeply optimizes the associated topological structure, the dynamic walk encoder of the feature sequence based on the graph convolutional neural network model generates the dynamic walk semantic encoding vector of the surface context of humidity expectation deviation and the dynamic walk semantic encoding vector of the hidden context of humidity expectation deviation in a hierarchical manner through the dynamic graph representation learning based on the graph convolutional neural network (GCN). Among them, the graph convolutional neural network (GCN) has a powerful graph structure data processing capability, can effectively propagate information through the adjacency matrix, and establish complex relationships between graph nodes, capturing the potential timing and causal relationships between each humidity expectation deviation sequence. The dynamic walk jumps and transmits along the edges between nodes in the graph structure, capturing the time-varying relationship and context information between nodes. In this way, in order to generate the semantic representation of the surface context of humidity expectation deviation, the dynamic walk mechanism simulates the propagation of features in the topological structure, recursively aggregates the explicit semantics between nodes from local to global.

[0061] Specifically, the set of the humidity expectation deviation semantic causal association topological feature matrix and the humidity expectation deviation local temporal depth implicit feature encoding vector is input into the feature sequence dynamic walk encoder based on the graph convolutional neural network model to obtain the humidity expectation deviation hidden layer context dynamic walk semantic encoding vector, which is expressed as the humidity expectation deviation hidden layer formula:

[0062]

[0063] Among them, H hidden is the dynamic walk semantic encoding vector of the hidden context of humidity expectation deviation. It should be understood that the humidity expectation deviation semantic causal association topological feature matrix can capture the complex causal relationship and interaction pattern between humidity changes, while the humidity expectation deviation local temporal depth implicit feature encoding vector focuses on describing the local characteristics of humidity changes in the time dimension. Combining the two and inputting them into the feature sequence dynamic walk encoder model based on the graph convolutional neural network model can simultaneously utilize topological structure information and time series features to characterize the essential characteristics of humidity changes at a higher level. In this way, for the dynamic semantic encoding of the hidden context of humidity expectation deviation, the dynamic walk mechanism conducts a deeper semantic exploration of implicit features. This hidden layer representation provides greater generalization ability for feature expression by capturing profound temporal dependencies and complex semantic patterns.

[0064] Specifically, the humidity expectation deviation hidden layer context dynamic walk semantic coding vector and the humidity expectation deviation surface layer context dynamic walk semantic coding vector are fused to obtain the humidity deviation temporal mode transfer coding vector, which is expressed as a fusion formula:

[0065] H final =γ·H surface +(1-γ)·H hidden

[0066] Among them, γ is the fusion weight parameter, H final is the humidity deviation time series pattern transfer encoding vector. It should be understood that the humidity expectation deviation surface context dynamic walk semantic encoding vector focuses on describing the explicit features of humidity changes. The humidity expectation deviation hidden context dynamic walk semantic encoding vector mainly captures the deep dynamic laws of humidity changes, including causal relationships, long-term dependencies, and complex interaction patterns hidden in the data. The semantic representations of the surface and hidden layers each capture explicit and implicit semantic information. Considering that these two feature dimensions are complementary, feature fusion is required in the final stage. In specific implementations, the importance distribution of surface and hidden features in the fusion process can be dynamically adjusted through strategies such as weighted summation, attention mechanism, or multi-head fusion to ensure that the comprehensive features have stronger distinguishing ability and representational integrity.

[0067] In the above-mentioned buffer bin dehumidification system 100, the dehumidifier control module 160 is used to transfer the coding vector based on the humidity deviation timing pattern to obtain a control instruction for indicating whether to start the dehumidifier. It should be understood that traditional control strategies usually use fixed rules or thresholds to decide whether to start the dehumidifier. This approach often appears too rigid when facing complex and changeable storage environments. The humidity deviation timing pattern transfer coding vector contains long-term trends and causal relationships extracted from historical data, as well as the immediate characteristics of the current humidity state. This comprehensive information fusion enables the system to more accurately determine whether the current humidity deviates from the expected range and predict possible future changes. The control instructions generated based on this coding vector no longer rely on a single point in time humidity value or a simple threshold comparison, but take into account the overall dynamic characteristics of humidity changes. That is, the humidity deviation time series pattern transfer coding vector obtained by dynamic causal accumulation of the set of humidity expected deviation local time series pattern feature coding vectors is classified and processed, so that the dehumidifier control module built based on the classifier can accurately judge whether the current humidity condition requires the dehumidifier to be started according to the information carried by the humidity deviation time series pattern transfer coding vector. In particular, it uses the existing classification rules and training models to map the complex humidity feature information to the two clear control decision categories of "start" or "do not start". This avoids unnecessary or untimely start of the dehumidifier due to unreasonable judgment, ensures timely adjustment when the humidity exceeds the expected range, maintains the humidity stability in the buffer bin, and protects the items in the bin from the influence of abnormal humidity.

[0068] In an embodiment of the present application, the dehumidifier control module 160 is used to: input the humidity deviation timing pattern transfer coding vector into the classifier-based dehumidifier control module to obtain the control instruction, and the control instruction is used to indicate whether to start the dehumidifier. Specifically, in an embodiment of the present application, the dehumidifier control module 160 is used to: use the fully connected layer of the classifier-based dehumidifier control module to fully connect the humidity deviation timing pattern transfer coding vector to obtain a humidity deviation fully connected coding feature vector; pass the humidity deviation fully connected coding feature vector through the Softmax classification function of the classifier-based dehumidifier control module to obtain a first probability belonging to starting the dehumidifier and a second probability belonging to not starting the dehumidifier; and, based on the comparison between the first probability and the second probability, determine the control instruction. It should be understood that traditional control methods usually rely on fixed rules and are difficult to cope with complex dynamic environmental changes. The classifier can extract key features from the coding vector and establish a mapping relationship through training and learning, and then accurately determine whether the current humidity state requires the dehumidifier to be started. Compared with simple threshold judgment, this method is more adaptable to complex dynamic environmental changes and avoids misjudgment or delayed response caused by a single indicator.

[0069] In summary, a buffer bin dehumidification system 100 based on an embodiment of the present application is explained, which includes a water tower, a dehumidifier, a cooling water pipe, a humidity sensor and a central control system. The central control system obtains the time queue of the real-time humidity value and compares it with the preset expected humidity value, and then performs segmentation and local timing pattern encoding, so as to realize the braking start judgment of the dehumidifier according to the dynamic causal transmission representation of the humidity deviation between the local timing pattern characteristics of each humidity expected deviation. In this way, by setting a precise expected humidity value and being able to accurately identify the timing trend and pattern of the actual humidity deviation change, an overly sensitive response mechanism is avoided and the control accuracy is improved. At the same time, the misjudgment caused by noise is reduced, thereby achieving more efficient, accurate and energy-saving humidity control.

[0070] As described above, the buffer silo dehumidification system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the buffer silo dehumidification system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the buffer silo dehumidification system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the buffer silo dehumidification system 100 can also be one of the many hardware modules of the terminal device.

[0071] Alternatively, in another example, the buffer silo dehumidification system 100 and the terminal device may also be separate devices, and the buffer silo dehumidification system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0072] In particular, in another embodiment of the present application, Figures 6 to 8 As shown, a buffer silo dehumidification system includes: No. 1 dehumidifier 1, No. 1 water tower 2, inlet air valve 3, air duct 4, in-store air outlet 5, No. 2 dehumidifier 6, No. 2 water tower 7, No. 2 cooling tower 8, No. 2 cooling water pipe 9, No. 1 cooling water pipe 10, dehumidifier air valve 11, in-store sensor 12, and central control system 13. Among them, No. 1 dehumidifier 1 is connected to No. 1 water tower 2 through No. 1 cooling water pipe 10, and No. 1 water tower 2 provides cooling water for it; external air enters No. 1 dehumidifier 1 through inlet air valve 3 and air duct 4 for dehumidification treatment, and the treated dry air is then transported to the in-store air outlet 5 through air duct 4 and discharged into the buffer silo. At the same time, it is also equipped with No. 2 dehumidifier 6, which is connected to No. 2 water tower 7 through No. 2 cooling water pipe 9 for dehumidification treatment; two cooling towers are connected to two dehumidifiers through cooling water pipes, which are used to circulate cooling water to support the operation of the refrigeration system. The dehumidifier air valve 11 is arranged between the air duct 4 and the dehumidifier, and is used to adjust the air volume and wind direction, and realize the interlocking control with the dehumidifier. The internal sensor 12 is deployed in the buffer warehouse, and is used to monitor the humidity in the warehouse in real time and transmit the signal to the central control system 13. The central control system 13 is electrically connected with the No. 1 dehumidifier 1, the No. 1 water tower 2, the No. 2 dehumidifier 6, the No. 2 water tower 7, the warehouse inlet air valve 3, the dehumidifier air valve 11, the warehouse outlet 5 and the internal sensor 12, respectively, to realize the unified control and automatic adjustment of each component.

[0073] Specifically, the water temperature and water level sensor signals of the two cooling towers are connected to the PLC to receive the hot water sent by the cooling pump, and provide cooling water for the dehumidifier after cooling. The water temperature and water level alarm is displayed on the control system screen, and the interlocking start and stop control of the cooling pump under the alarm state is realized. The air valve of the buffer bin dehumidification pipeline realizes the opening angle control, and the air valve opening signal is connected to the PLC to realize the interlocking control with the dehumidification system. The control system automatically opens and closes the dehumidification system according to the humidity conditions of each granary in the buffer bin. The dehumidifier pressure gauge data is connected to the PLC, and the pressure value display and alarm value setting functions are realized in the control system. Among them, a dehumidifier contains two compressors. The compressors are automatically turned on and off according to the required moisture content in the warehouse. The two compressors automatically select the compressor that needs to be put into use first according to the running time. It is worth mentioning that in the dehumidifier, the high-temperature and high-humidity air first enters the heat exchanger and is cooled by low-temperature air, so that the intake air temperature is cooled by about 3-8°C and some moisture is removed. The pre-cooled wet air then enters the evaporator for forced cooling, so that the wet air temperature drops to about 1-10°C. Most of the moisture is removed by a special gas-liquid separator. At this time, the water content in the air is very small, about 7-10g / m 3 After the dehumidified low-temperature air is heated by the heat exchanger and heater, the temperature is about 20-30℃ and the relative humidity is ≤60%, becoming dry finished gas. Under extreme working conditions, the entire dehumidification process can remove about 80% of the moisture.

[0074] In particular, the air pressure sensors and sound and light alarm systems in the four granaries of the buffer compartment set alarm values. When the air pressure in the warehouse reaches a certain pressure, the sound and light alarm system at the door of the buffer warehouse automatically starts and controls the air intake and exhaust volume in a chain manner. In this way, a warning can be issued in time when the air pressure in the warehouse is abnormal, avoiding safety hazards caused by excessively high or low air pressure, and no manual intervention is required, which effectively improves the automation level of the system and reduces delays and errors in human operations.

[0075] The operation control of the dehumidifier and related valve groups is uniformly connected to the automatic control platform of the main engine in the central control room. The automatic control platform can display the dehumidifier equipment operation status, temperature and humidity control, pressure control, data alarm and other data, and realize the equipment interlocking control start and stop, dehumidification system gear frequency conversion control and other functions. The equipment operation status, temperature and humidity control, pressure control, data alarm, equipment interlocking control start and stop and other functions are all centralized in the automatic control platform and can be clearly displayed. In this way, the centralized management of equipment operation status, temperature and humidity control, pressure control, data alarm and other functions is realized, which improves work efficiency.

[0076] In this way, the dehumidification system is automatically adjusted, and the dehumidifier is automatically started and stopped according to the temperature and humidity in the warehouse based on the existing sensors. The temperature and humidity sensor control point in the warehouse or the average temperature and humidity in the warehouse can be arbitrarily selected to control the start and stop of the dehumidification unit. The compressor calculates the compressor energy adjustment input according to the suction pressure (the suction pressure corresponds to the dew point temperature) to control the dehumidification amount. After dehumidification, the low-temperature and low-humidity gas is electrically heated and the input amount is automatically calculated according to the outlet air temperature. The moisture content is automatically calculated according to the temperature and humidity in the warehouse. When the moisture content in the warehouse is too high, the dehumidification unit is automatically started. When the moisture content in the warehouse is low, the dehumidification unit is automatically stopped to realize the temperature and humidity control of the four granaries in the buffer cabin separately. The system has the characteristics of precise control, energy saving and environmental protection, easy maintenance and high automation, which can effectively improve the efficiency of warehouse management and the stability of product quality.

Claims

1. A buffer silo dehumidification system, comprising a water tower, a dehumidifier connected to the water tower through a cooling water pipe, the water tower is used to provide cooling water for the dehumidifier, a humidity sensor deployed in the buffer silo and a central control system, characterized in that: The central control system comprises: An expected humidity value setting module, used for setting an expected humidity value; A real-time humidity value acquisition module, used to obtain a time queue of real-time humidity values ​​acquired by the humidity sensor; A humidity expected deviation value calculation module, used for calculating the difference between the real-time humidity value at each time point in the time queue of the real-time humidity value and the expected humidity value to obtain a time queue of the humidity expected deviation value; A humidity expectation deviation local encoding module, used for performing sequence segmentation and local temporal pattern feature extraction on the time queue of the humidity expectation deviation value to obtain a set of humidity expectation deviation local temporal pattern feature encoding vectors; A humidity local time series dynamic transfer module is used to perform a humidity local time series deviation dynamic causal accumulation analysis on the set of humidity expected deviation local time series pattern feature coding vectors to obtain a humidity deviation time series pattern transfer coding vector; The dehumidifier control module is used to transmit the coding vector based on the humidity deviation timing mode to obtain a control instruction indicating whether to start the dehumidifier.

2. The buffer silo dehumidification system according to claim 1, characterized in that: The humidity expected deviation local encoding module includes: A humidity expected deviation equal-time sequence segmentation unit is used to segment the time queue of the humidity expected deviation value into equal-time sequence to obtain a set of humidity expected deviation local time series; The humidity expected deviation local time series sequence feature extraction unit is used to input each humidity expected deviation local time series sequence in the set of humidity expected deviation local time series sequences into the humidity deviation time series pattern feature extractor based on the forward LSTM model to obtain the set of humidity expected deviation local time series pattern feature encoding vectors.

3. The buffer silo dehumidification system according to claim 2, characterized in that: The humidity local time series dynamic transmission module includes: A humidity expectation deviation local time series pattern implicit feature mining unit, used for performing implicit feature mining on each humidity expectation deviation local time series pattern feature coding vector in the set of humidity expectation deviation local time series pattern feature coding vectors to obtain a set of humidity expectation deviation local time series depth implicit feature coding vectors; A humidity expectation deviation causal association topological feature construction unit is used to construct the causal association topological features of the set of humidity expectation deviation local temporal depth implicit feature encoding vectors to obtain a humidity expectation deviation semantic causal association topological feature matrix; The humidity expectation deviation context dynamic walking fusion unit is used to use the humidity expectation deviation semantic causal association topological feature matrix as the semantic causal association structure information, and perform context dynamic walking fusion on the set of the humidity expectation deviation local temporal pattern feature coding vectors and the set of the humidity expectation deviation local temporal depth implicit feature coding vectors to obtain the humidity deviation temporal pattern transfer coding vector.

4. The buffer silo dehumidification system according to claim 3, characterized in that: The humidity expectation deviation causal association topological feature construction unit includes: A humidity expectation deviation semantic causal association factor calculation subunit, used to calculate the semantic causal association factor between any two humidity expectation deviation local temporal depth implicit feature coding vectors in the set of humidity expectation deviation local temporal depth implicit feature coding vectors to obtain a humidity expectation deviation semantic causal association topological matrix composed of multiple humidity expectation deviation semantic causal association factors; The humidity expectation deviation semantic causal trigger subunit is used to input the humidity expectation deviation semantic causal association topological matrix into a causal trigger network based on a gated activation function to obtain the humidity expectation deviation semantic causal association topological feature matrix.

5. The buffer silo dehumidification system according to claim 4, characterized in that: The humidity expectation deviation semantic causal association factor calculation subunit is used to: Calculate the association matrix between any two humidity expected deviation local time series depth implicit feature coding vectors in the set of humidity expected deviation local time series depth implicit feature coding vectors to obtain a set of humidity expected deviation association matrices; Calculating the semantic causal association factor of each humidity expectation deviation association matrix in the set of the humidity expectation deviation association matrices to obtain the humidity expectation deviation semantic causal association topological matrix composed of multiple humidity expectation deviation semantic causal association factors, wherein the humidity expectation deviation semantic causal association factor is calculated from the mean, variance, maximum value and causal association bias value of the humidity expectation deviation association matrix; In which, in response to the variance of the humidity expected deviation association matrix being greater than or equal to a predetermined threshold, a weighted average of the distances between any two humidity expected deviation local temporal depth implicit feature coding vectors in the set of humidity expected deviation local temporal depth implicit feature coding vectors is used as the causal association bias value; In response to the variance of the humidity expectation deviation correlation matrix being smaller than the predetermined threshold, a weighted value of the variance of the humidity expectation deviation correlation matrix is ​​used as the causal correlation bias value.

6. The buffer silo dehumidification system according to claim 5, characterized in that: The humidity expectation deviation context dynamic walking fusion unit is used to: Input the set of the humidity expectation deviation semantic causal association topological feature matrix and the humidity expectation deviation local temporal pattern feature encoding vector into a feature sequence dynamic walk encoder based on a graph convolutional neural network model to obtain a humidity expectation deviation surface context dynamic walk semantic encoding vector; Input the humidity expectation deviation semantic causal association topological feature matrix and the set of the humidity expectation deviation local temporal depth implicit feature encoding vector into the feature sequence dynamic walk encoder based on the graph convolutional neural network model to obtain the humidity expectation deviation hidden layer context dynamic walk semantic encoding vector; The humidity expectation deviation hidden layer context dynamic walking semantic coding vector and the humidity expectation deviation surface layer context dynamic walking semantic coding vector are fused to obtain the humidity deviation time series mode transfer coding vector.

7. The buffer silo dehumidification system according to claim 6, characterized in that: The dehumidifier control module is used to: input the humidity deviation timing pattern transfer coding vector into the classifier-based dehumidifier control module to obtain the control instruction, and the control instruction is used to indicate whether to start the dehumidifier.

8. The buffer silo dehumidification system according to claim 7, characterized in that: The dehumidifier control module is used to: Using the fully connected layer of the classifier-based dehumidifier control module to perform fully connected encoding on the humidity deviation time series pattern transfer encoding vector to obtain a humidity deviation fully connected encoding feature vector; Passing the humidity deviation fully connected encoded feature vector through the Softmax classification function of the classifier-based dehumidifier control module to obtain a first probability attributable to starting the dehumidifier and a second probability attributable to not starting the dehumidifier; Based on a comparison between the first probability and the second probability, the control instruction is determined.

Images