An intelligent storage management system and method for dried orange peel

Abstract

The embodiment of the application provides a Chenpi intelligent warehouse management and control system and method. The system comprises a digital twin visualization module, an environment sensing module, an inference module, a regulation and control module, and a power supply module. The digital twin visualization module realizes scene depth adaptation, the environment sensing module realizes environment intelligent monitoring, the inference module realizes early warning of mildew hidden danger, the regulation and control module realizes automatic regulation and control of equipment, and the power supply module realizes green energy supply of the system. Therefore, the Chenpi warehouse full-process precision, intelligentization, visualization, and low-carbonization management and control are realized, so as to significantly reduce the labor input and energy consumption cost, and improve the operation efficiency and Chenpi aging quality stability.

Description

Technical Field

[0001] This invention relates to the field of tangerine peel storage management technology, and in particular to an intelligent tangerine peel storage management system and method. Background Technology

[0002] Xinhui tangerine peel, a national geographical indication product and a national intangible cultural heritage, is a pillar industry for characteristic agriculture and rural revitalization in Guangdong Province. Warehousing quality control and intelligent upgrading are core elements for ensuring the authenticity of geographical indication agricultural products, enhancing industrial added value, and supporting rural revitalization and high-quality agricultural development.

[0003] However, the current Xinhui tangerine peel storage industry is still mainly based on the traditional manual management model. Existing tangerine peel storage technologies have technical problems such as difficulty in controlling the storage environment, high risk of mold growth, low labor efficiency, and high energy consumption costs. Summary of the Invention

[0004] This invention provides an intelligent warehousing and management system and method for dried tangerine peel. Through a digital twin visualization module, it achieves deep scene adaptation; an environmental perception module achieves intelligent environmental monitoring; an inference module achieves early warning of mold potential; a control module achieves automatic equipment control; and a power supply module achieves green energy supply for the system. This enables precise, intelligent, visualized, and low-carbon management and control of the entire warehousing process of dried tangerine peel, significantly reducing labor input and energy consumption costs, and improving operational efficiency and the stability of the aging quality of dried tangerine peel.

[0005] In a first aspect, embodiments of the present invention provide an intelligent warehousing and management system for dried tangerine peel, comprising: The digital twin visualization module is used to construct a high-fidelity virtual mapping scene of the target tangerine peel warehouse and synchronize the real-time status data of the target tangerine peel warehouse in real time. An environmental sensing module is used to collect real-time status data of the target tangerine peel storage. The real-time status data includes environmental data and tangerine peel image data. The environmental data includes environmental temperature data, environmental humidity data, and gas concentration data. The environmental sensing module includes a temperature sensor, a humidity sensor, a gas sensor, and a camera device. The temperature sensor is used to collect the environmental temperature data, the humidity sensor is used to collect the environmental humidity data, the gas sensor is used to collect the gas concentration data, and the camera device is used to collect the tangerine peel image data. The reasoning module includes a tangerine peel mold risk perception model and a temperature and humidity prediction model. The temperature and humidity prediction model is used to output predicted data on the temperature and humidity change trends in the future period. The reasoning module is used to input the real-time status data and the predicted data into the tangerine peel mold risk perception model to obtain a mold risk index, and determine the mold risk level based on the mold risk index and a preset index threshold range, and then generate control instructions based on the mold risk level. The control module includes ventilation equipment and dehumidification equipment, and the control module is used to adjust the operating status of the ventilation equipment and dehumidification equipment according to the control command; The power supply module includes a photovoltaic energy storage device, which is used to provide power to the system.

[0006] In some embodiments, the digital twin visualization module includes: The 3D modeling unit is used to construct a detailed model of the main building, equipment, sensors, and storage locations of the target tangerine peel warehouse based on the Unity3D engine. The virtual-real synchronization unit is used to connect to the MySQL database through a standardized interface and map the real-time status data to dynamic identifiers in the virtual scene; The interactive rendering unit is used to realize interactive visualization rendering of multiple viewing perspective modes, including scene free roaming mode, first-person inspection mode and AGV trajectory tracking mode.

[0007] In some embodiments, the expression for the mold risk index is as follows:

[0008] Among them, MRI is the risk index for mold growth. For real-time ambient temperature in the warehouse, This is the optimal temperature for mold growth. For ambient relative humidity, The minimum humidity level is critical for mold germination. Duration of the abnormal environment.

[0009] In some embodiments, the tangerine peel mold risk perception model adopts a mechanism fusion RNN architecture, and the mold risk level output by the tangerine peel mold risk perception model includes an intact level, a slightly moldy level, and a moderately moldy level.

[0010] In some embodiments, the tangerine peel mold risk perception model is deployed on an edge device or a host computer. The tangerine peel mold risk perception model uses a target detection method based on YOLOv8 to identify the tangerine peel image data, and extracts key features of the target moldy area in the tangerine peel image data based on an attention mechanism to obtain the detection result. The detection result is then returned to the visualization interface of the digital twin visualization module for annotation and display.

[0011] In some embodiments, the photovoltaic energy storage device includes: Solar panels are installed at a fixed tilt angle on the warehouse roof. Lithium-ion battery charging module for storing photovoltaic energy; The power management module is configured to: when there is sufficient sunlight, power is supplied by the solar panel and the lithium battery charging module is charged; when there is insufficient sunlight, the power supply is switched to the lithium battery charging module.

[0012] Secondly, embodiments of the present invention also provide a method for intelligent warehousing and management of dried tangerine peel, applied to the intelligent warehousing and management system for dried tangerine peel as described in the first aspect, the method comprising: The environmental perception module collects real-time status data of the target tangerine peel storage. The real-time status data and the predicted data are input into the tangerine peel mold risk perception model through the reasoning module to obtain the mold risk index. The reasoning module determines the mold risk level based on the mold risk index and a preset index threshold range. The reasoning module generates control instructions based on the mold risk level. The control module adjusts the operating status of the ventilation equipment and the dehumidification equipment according to the control instructions. The digital twin visualization module visually displays the warehouse status, mold risk level, and equipment operating status.

[0013] In some embodiments, the temperature and humidity prediction model employs a long short-term memory network structure; the training method of the temperature and humidity prediction model includes: Acquire historical environmental data and continuous monitoring environmental data for a preset time period; The original training data is constructed based on the historical environmental data and the continuous monitoring environmental data. The original training data is subjected to missing value imputation and outlier removal to obtain preprocessed training data; The preprocessed training data is normalized to obtain time series samples; The temperature and humidity prediction model is trained based on the time series samples to obtain the trained temperature and humidity prediction model.

[0014] In some embodiments, the method further includes: The tangerine peel image data is preprocessed to obtain preprocessed image data, wherein the data enhancement processing includes noise reduction processing and color space conversion processing; The YOLOv8-based target detection method extracts key features from the target moldy area in the preprocessed image data to obtain detection results, wherein the detection results include moldy area location information and mold category information; The detection results are returned to the visualization interface of the digital twin visualization module for annotation and display.

[0015] In some embodiments, the method further includes: The power of the photovoltaic energy storage device is periodically monitored, and the remaining power and power supply mode are displayed in the visualization interface of the digital twin visualization module. When the remaining power of the photovoltaic energy storage device is detected to be lower than the safety threshold, power consumption control is applied to non-critical equipment in the target tangerine peel storage, and a low power warning is triggered.

[0016] According to embodiments of the present invention, the intelligent tangerine peel storage and management system and method include a digital twin visualization module for constructing a high-fidelity virtual mapping scene of the target tangerine peel storage and synchronizing the real-time status data of the target tangerine peel storage in real time; and an environmental perception module for collecting real-time status data of the target tangerine peel storage, including environmental data and tangerine peel image data. The environmental data includes environmental temperature data, environmental humidity data, and gas concentration data. The environmental perception module includes a temperature sensor, a humidity sensor, a gas sensor, and a camera device. The temperature sensor is used to collect environmental temperature data, the humidity sensor is used to collect environmental humidity data, and the gas sensor is used to collect gas concentration data. The system comprises several modules: a concentration data acquisition module (camera equipment for collecting images of dried tangerine peel); an inference module (including a tangerine peel mold risk perception model and a temperature and humidity prediction model; the temperature and humidity prediction model for outputting predicted temperature and humidity trends over future periods; and an inference module for inputting real-time status data and predicted data into the tangerine peel mold risk perception model to obtain a mold risk index, determining the mold risk level based on the mold risk index and a preset threshold range, and generating control commands based on the mold risk level); a control module (including ventilation and dehumidification equipment for adjusting their operating status according to the control commands); and a power supply module (including a photovoltaic energy storage device for providing power to the system). Based on this, the intelligent tangerine peel storage and management system and method provided in this embodiment of the invention achieves deep scene adaptation through a digital twin visualization module, intelligent environmental monitoring through an environmental perception module, early warning of mold hazards through an inference module, automatic equipment control through a control module, and green energy supply through a power supply module. This enables precise, intelligent, visualized, and low-carbon management and control of the entire tangerine peel storage process, significantly reducing labor input and energy consumption costs, and improving operational efficiency and the stability of tangerine peel aging quality. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall structure of the intelligent tangerine peel storage and management system provided in one embodiment of the present invention; Figure 2 This is a flowchart of an intelligent warehousing and management method for dried tangerine peel provided in one embodiment of the present invention; Figure 3 This is a flowchart of a training method for a temperature and humidity prediction model provided in one embodiment of the present invention; Figure 4 This is a flowchart of the method steps S301 to S303 provided in one embodiment of the present invention; Figure 5 This is a flowchart of steps S401 to S402 provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the following drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0020] In this embodiment of the invention, the terms "furthermore," "exemplarily," or "optionally" are used as examples, illustrations, or descriptions and should not be construed as being more preferred or advantageous than other embodiments or designs. The use of the terms "furthermore," "exemplarily," or "optionally" is intended to present the relevant concepts in a specific manner.

[0021] To facilitate a more convenient description of the working principle of the embodiments of the present invention, the following introduction of relevant technical scenarios is given first.

[0022] Xinhui tangerine peel, a national geographical indication product and a national intangible cultural heritage, is a pillar industry for characteristic agriculture and rural revitalization in Guangdong Province. Warehousing quality control and intelligent upgrading are core elements for ensuring the authenticity of geographical indication agricultural products, enhancing industrial added value, and supporting rural revitalization and high-quality agricultural development.

[0023] However, the current Xinhui tangerine peel storage industry is still mainly based on the traditional manual management model. Existing tangerine peel storage technologies have technical problems such as difficulty in controlling the storage environment, high risk of mold growth, low labor efficiency, and high energy consumption costs.

[0024] Based on this, the present invention provides an intelligent warehousing and management system and method for dried tangerine peel. The intelligent warehousing and management system for dried tangerine peel includes a digital twin visualization module for constructing a high-fidelity virtual mapping scene of the target dried tangerine peel warehouse and synchronizing the real-time status data of the target dried tangerine peel warehouse in real time; and an environmental perception module for collecting real-time status data of the target dried tangerine peel warehouse, including environmental data and dried tangerine peel image data. The environmental data includes ambient temperature data, ambient humidity data, and gas concentration data. The environmental perception module includes a temperature sensor, a humidity sensor, a gas sensor, and a camera device. The temperature sensor is used to collect ambient temperature data, the humidity sensor is used to collect ambient humidity data, the gas sensor is used to collect gas concentration data, and the camera device is used for... The system collects image data of dried tangerine peel; an inference module includes a tangerine peel mold risk perception model and a temperature and humidity prediction model. The temperature and humidity prediction model outputs predicted data on future temperature and humidity trends. The inference module inputs real-time status data and predicted data into the tangerine peel mold risk perception model to obtain a mold risk index. Based on the mold risk index and a preset index threshold range, a mold risk level is determined, and control instructions are generated based on the mold risk level. A control module includes ventilation equipment and dehumidification equipment, which adjust the operating status of the ventilation equipment and dehumidification equipment according to the control instructions. A power supply module includes a photovoltaic energy storage device, which provides power to the system. Based on this, the tangerine peel intelligent warehousing management system and method provided by this invention achieves deep scene adaptation through a digital twin visualization module, intelligent environmental monitoring through an environmental perception module, early warning of mold risks through an inference module, automatic equipment control through a control module, and green energy supply through a power supply module. This enables precise, intelligent, visualized, and low-carbon management of the entire tangerine peel warehousing process, significantly reducing labor input and energy costs, and improving operational efficiency and the stability of tangerine peel aging quality.

[0025] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0026] like Figure 1 As shown, Figure 1This is a structural block diagram of an intelligent tangerine peel storage and management system provided in one embodiment of the present invention. The intelligent tangerine peel storage and management system includes: a digital twin visualization module, an environmental perception module, an inference module, a control module, and a power supply module. The digital twin visualization module is used to construct a high-fidelity virtual mapping scene of the target tangerine peel storage and synchronize the real-time status data of the target tangerine peel storage in real time. The environmental perception module is used to collect real-time status data of the target tangerine peel storage, including environmental data and tangerine peel image data. The environmental data includes environmental temperature data, environmental humidity data, and gas concentration data. The environmental perception module includes a temperature sensor, a humidity sensor, a gas sensor, and a camera device. The temperature sensor is used to collect environmental temperature data, the humidity sensor is used to collect environmental humidity data, the gas sensor is used to collect gas concentration data, and the camera device is used to collect tangerine peel image data. The system comprises: an inference module, including a tangerine peel mold risk perception model and a temperature and humidity prediction model; a temperature and humidity prediction model that outputs predicted data on future temperature and humidity trends; and an inference module that inputs real-time status data and predicted data into the tangerine peel mold risk perception model to obtain a mold risk index, determines the mold risk level based on the mold risk index and a preset index threshold range, and then generates control commands based on the mold risk level. The control module includes ventilation and dehumidification equipment, which adjusts the operating status of the ventilation and dehumidification equipment according to the control commands. The power supply module includes a photovoltaic energy storage device, which provides power to the system.

[0027] Understandably, the digital twin visualization module uses Unity3D as its core development engine. Based on the actual measured dimensions of the tangerine peel warehouse, architectural layout drawings, and equipment installation coordinates, it constructs a 1:1 high-fidelity virtual mapping scene, achieving complete equivalence between the physical warehouse and the digital space in terms of structure, size, location, and equipment relationships. The modeling process unfolds layer by layer from the whole to the parts. First, it completes the accurate modeling of the basic structure of the warehouse building, such as the main structure, walls, columns, doors and windows, ventilation ducts, and entrances and exits, strictly maintaining the spatial proportions consistent with reality. Then, it meticulously recreates the internal scene, including the stacking method of the tangerine peel crates, the distribution of storage locations, AGV driving channels, temperature, humidity and gas sensor nodes, ventilation and dehumidification equipment, photovoltaic panels and energy storage devices, etc., fully replicating the appearance, installation location, connection relationships and spatial arrangement of the equipment, ensuring a high degree of consistency between the virtual scene and the real warehouse in terms of visual effects and spatial logic.

[0028] In terms of rendering, the system employs Unity3D's real-time lighting pipeline, combined with global illumination, ambient occlusion, and dynamic shadow technology to realistically reproduce the effects of natural and artificial lighting inside the warehouse, simulating light and shadow changes in different areas and at different times, thus enhancing the realism and immersion of the virtual scene. Simultaneously, key objects such as sensor locations, control equipment, warning areas, AGV operating trajectories, and insect and mold monitoring points are rendered and dynamically labeled in a differentiated manner, making the core monitoring targets clearly identifiable and easily understood in the virtual scene. To achieve synchronous mapping between the virtual and real worlds, the system connects in real-time to a Zigbee sensor network, edge computing nodes, and a MySQL database through standardized data interfaces. This allows for the real-time rendering and updating of multi-source dynamic data from the physical warehouse, including temperature and humidity, gas concentration, equipment operating status, storage location occupancy information, mold risk level, insect infestation detection results, and photovoltaic energy storage power, presented in the form of numerical panels, status indicator lights, color warnings, and dynamic effects. This enables the virtual scene to dynamically, accurately, and in real-time reflect the true state of the physical warehouse.

[0029] In addition, the system supports multi-mode interactive visualization rendering, providing multiple observation perspectives such as free scene roaming, first-person immersive inspection, and AGV follow-up tracking. Managers can freely switch perspectives, zoom in and out, and view details in the virtual scene to gain a comprehensive and seamless understanding of the warehouse's overall operational status, providing high-fidelity and highly interactive visualization support for remote supervision, status monitoring, anomaly warning, and scheduling decisions.

[0030] Understandably, the environmental sensing module is used to collect real-time status data of the target tangerine peel storage. The real-time status data includes environmental data and tangerine peel image data. The environmental data includes environmental temperature data, environmental humidity data, and gas concentration data. The environmental sensing module includes a temperature sensor, a humidity sensor, a gas sensor, and a camera device. The temperature sensor is used to collect environmental temperature data, the humidity sensor is used to collect environmental humidity data, the gas sensor is used to collect gas concentration data, and the camera device is used to collect tangerine peel image data.

[0031] Understandably, the inference module includes a tangerine peel mold risk perception model and a temperature and humidity prediction model. The temperature and humidity prediction model is used to output predicted data on the future temperature and humidity change trends. The inference module is used to input real-time status data and predicted data into the tangerine peel mold risk perception model to obtain a mold risk index. Based on the mold risk index and the preset index threshold range, the mold risk level is determined, and then a control instruction is generated based on the mold risk level.

[0032] Understandably, the RNN-based model for perceiving the risk of mold growth in dried tangerine peel uses the general mechanism formula of the Mould Risk Index (MRI) as its underlying physical basis. The formula is as follows:

[0033] In the formula: For real-time ambient temperature in the warehouse, This is the optimal temperature for mold growth. For ambient relative humidity, The minimum humidity level is critical for mold germination. The duration of abnormal environments was considered. A recurrent neural network (RNN) was selected to construct a mold perception network. Using the storage time-series temperature and humidity sequence and real-time calculated MRI mold index as model input features, the study explored the intrinsic mapping relationship between environmental parameter evolution, the accumulation of abnormal duration, and the degree of mold growth in dried tangerine peel. A mold perception model integrating mechanistic formulas and deep learning was established.

[0034] Based on the growth characteristics of the specific fungal strains used in the storage of Xinhui tangerine peel, the underlying parameters of the model were localized and calibrated: the optimal growth temperature for mold and the minimum critical humidity for mold germination RHmin were determined to be 70%. The model achieves a 75% accuracy rate in matching the physiological growth threshold of endophytic mold in dried tangerine peel, distinguishing it from general mold models used for grains and common Chinese medicinal materials. Based on the calibrated MRI index threshold range, a three-category discrimination system for mold status of dried tangerine peel is constructed. The model can output three categories of stored material status: intact, slightly moldy, and moderately moldy, enabling quantitative and graded perception of mold risk. Simultaneously, it achieves upstream and downstream model linkage, incorporating the future temperature and humidity prediction sequence output by the upstream stacked LSTM to complete advanced mold risk extrapolation, enabling pre-emptive early warning and overcoming the limitations of traditional post-event detection and passive handling of mold in storage. This provides a quantitative decision-making basis for the pre-emptive closed-loop control of platform ventilation and dehumidification equipment.

[0035] For determining mold growth status, the system constructs an RNN model based on multi-source environmental data. The model inputs include parameters such as temperature, humidity, and gas concentration, and the output is different mold growth levels. By training on historical data, the model can reflect the relationship between environmental changes and mold growth risk to a certain extent. During deployment, model training is primarily completed on the server side, and the model is made available to the system via a data interface, thereby enabling real-time assessment of warehouse conditions.

[0036] The model's data primarily comes from multi-source environmental data including temperature and humidity, and training samples are constructed using manually labeled mold level information. During data preparation, data from different time periods are aligned to ensure the integrity of the input sequence.

[0037] During model training, an RNN structure was used to model the time series data, and the model parameters were adjusted through multiple experiments. In the initial training, the model did not perform well in recognizing slight mold growth. Subsequently, by increasing the number of samples and optimizing the labeling method, the model's ability to distinguish between different stages of mold growth was improved.

[0038] After the model is deployed, it interacts with the system through an interface to achieve real-time assessment of the risk of mold growth in the storage environment.

[0039] It is understandable that temperature and relative humidity in the storage environment exhibit typical nonlinear temporal evolution characteristics. The data is affected by multiple factors, including diurnal weather changes, heat storage and release within the storage structure, and the start and stop of ventilation equipment, resulting in significant long-term temporal dependencies. Traditional recurrent neural networks are prone to gradient vanishing problems during the fitting of long-sequence data, failing to effectively capture the long-term variation patterns of environmental parameters. Therefore, this invention employs a temperature and humidity prediction model based on an LSTM structure. Using a Long Short-Term Memory (LSTM) network as the basic network architecture, and leveraging its unique forget gate, input gate, output gate, and cell state unit structure, it effectively addresses the gradient propagation defects of conventional temporal networks. This model accurately mines the temporal correlation features within historical temperature and humidity data, enabling advanced temporal prediction of storage microenvironment parameters for future periods and achieving predictive perception of environmental conditions.

[0040] To meet the specific aging environment requirements of Xinhui tangerine peel, the basic LSTM network underwent end-to-end adaptation and optimization. First, time-series data preprocessing optimization was performed at the input end. For the raw data collected by the low-power sensor network in the warehouse, outlier removal, missing value interpolation and imputation, and maximum-minimum normalization were performed to eliminate sensor noise and environmental outlier interference, improving input data quality and model convergence stability. Second, the network structure was deeply optimized by adopting a multi-layer LSTM stacked architecture to deepen the feature extraction dimension and adding a fully connected mapping layer to enhance the temporal feature fitting capability. Compared to a single-layer basic network, this effectively improves long-term prediction accuracy and numerical stability, adapting to the long-term environmental prediction requirements of the slow aging cycle of tangerine peel over several years. Finally, a boundary constraint specific to the tangerine peel environment was added to the model output end. The output threshold was limited by the optimal aging temperature and humidity range for Xinhui tangerine peel, filtering out invalid prediction results that deviate from the physical reality of the warehouse and ensuring that the model output parameters closely match the actual operating conditions of tangerine peel storage.

[0041] Understandably, in addition to environmental parameter analysis, the system also utilizes a tangerine peel mold risk perception model for image recognition to detect mold growth on the surface of the tangerine peel. In the implementation process, image data is first acquired using a camera device and preprocessed as needed, such as denoising and color space conversion, to reduce interference from environmental factors. In the model part, the YOLOv8 object detection algorithm is used to analyze the images. Through training, the model is able to identify moldy areas and output their location and category information.

[0042] To improve detection performance, an attention mechanism was introduced during model optimization to enhance the extraction of features from key regions. Simultaneously, the loss function was adjusted to mitigate the impact of uneven sample distribution. During system operation, the model can perform real-time inference on images and annotate the recognition results in a visual interface, thus intuitively reflecting the distribution of mold.

[0043] Regarding image data, the system primarily uses cameras to collect images of dried tangerine peel samples in the warehouse environment, acquiring multiple sets of image data under different lighting conditions. To improve the model's generalization ability, data augmentation processing was also performed on some images, including brightness adjustment, rotation, and noise superposition.

[0044] During the training phase, transfer learning was performed based on the YOLOv8 model, with fine-tuning done on the pre-trained weights. Manual annotation tools were used to label moldy areas, generating the dataset required for training. In the initial training, the model performed poorly in recognizing small areas of mold. Subsequent training improved detection accuracy by adjusting the input resolution and introducing an attention mechanism to enhance features in key areas.

[0045] In actual deployment, the model runs on edge devices or host computers, analyzes real-time images by calling the inference interface, and returns the detection results to the visualization interface for annotation and display.

[0046] Understandably, the control module includes ventilation and dehumidification equipment. It adjusts the operating status of these equipment based on control commands. Based on this, the system forms a complete management and control chain encompassing environmental time-series prediction, mold growth mechanism deduction, risk classification and early warning, and closed-loop equipment control. This deeply adapts to the industry characteristics of Xinhui tangerine peel: "the older the better, susceptible to moisture and mold, requiring long-term storage." Deep learning-based time-series prediction addresses the lag in environmental control, while a mold growth physical mechanism model constrains the AI ​​output boundary, balancing algorithm accuracy with industry scenario adaptability. This suppresses mold growth losses at the source, achieving refined, intelligent, and proactive control of the tangerine peel storage microenvironment.

[0047] Understandably, the tangerine peel intelligent warehousing and management system requires a continuous and stable power supply to support the operation of various modules, sensors, and control equipment. Considering the characteristics of the tangerine peel storage environment (typically located in a well-ventilated and dry environment with good sunlight exposure), the power supply module of this invention employs a photovoltaic energy storage device, which provides power to the entire system. The photovoltaic energy storage device includes solar panels, a lithium battery charging module, and a power management module.

[0048] The solar panels are installed at a fixed tilt angle on the warehouse roof. A solar panel is a photovoltaic semiconductor wafer that directly generates electricity using sunlight. It's a device that directly converts solar energy into electrical energy through the photovoltaic effect. When a solar panel is exposed to sunlight, the charge distribution within it changes, generating an electromotive force and current, thus directly converting solar energy into electrical energy. The solar panels will use solar epoxy resin panels, which are made by laser-cutting solar cells into small pieces to generate the required voltage and current, and then encapsulating them. Due to their small size, they generally do not use the same encapsulation method as solar photovoltaic modules. Instead, they use epoxy resin to cover the solar cells and bond them to a PCB circuit board, which features fast production speed, pressure resistance, corrosion resistance, and a beautiful crystal appearance. Multiple cells are connected in series or parallel to form a high-power battery array to meet the system's input voltage requirements. Solar panels can convert sunlight into electrical energy, facilitating the use, transportation, and storage of solar energy.

[0049] The lithium battery charging module is used to store photovoltaic power. As the core energy storage unit, it buffers and regulates photovoltaic power generation, enabling the system to operate continuously.

[0050] The power management module is configured to: when there is sufficient sunlight, the solar panel supplies power and charges the lithium battery charging module; when there is insufficient sunlight, it switches to supply power from the lithium battery charging module, thus ensuring that the system can work normally without the need for an external power source.

[0051] In addition, such as Figure 2 As shown, an embodiment of the present invention also discloses a method for intelligent storage and management of dried tangerine peel, which is applied to the above-mentioned intelligent storage and management system for dried tangerine peel. The method includes, but is not limited to, the following steps.

[0052] Step S101: Collect real-time status data of the target tangerine peel storage through the environmental perception module; Step S102: Input the real-time status data and predicted data into the tangerine peel mold risk perception model through the inference module to obtain the mold risk index; Step S103: Determine the mold risk level through the reasoning module based on the mold risk index and the preset index threshold range; Step S104: Generate control instructions based on the mold risk level through the reasoning module; Step S105: Adjust the operating status of the ventilation and dehumidification equipment according to the control instructions through the control module; Step S106: Visualize the warehouse status, mold risk level, and equipment operation status using the digital twin visualization module.

[0053] Based on this, the intelligent warehousing and management method for dried tangerine peel of the present invention achieves deep scene adaptation through a digital twin visualization module, intelligent environmental monitoring through an environmental perception module, early warning of mold hazards through an inference module, automatic equipment control through a control module, and green energy supply through a power supply module. This enables precise, intelligent, visualized, and low-carbon management and control of the entire warehousing process for dried tangerine peel, significantly reducing labor input and energy consumption costs, and improving operational efficiency and the stability of the aging quality of dried tangerine peel.

[0054] Understandably, temperature and humidity prediction models employ long short-term memory network structures, such as... Figure 3 As shown, the training methods for the temperature and humidity prediction model include: Step S201: Acquire historical environmental data and continuous monitoring environmental data for a preset time period; Step S202: Construct the original training data based on historical environmental data and continuous monitoring environmental data; Step S203: Perform missing value imputation and outlier removal on the original training data to obtain preprocessed training data; Step S204: Normalize the preprocessed training data to obtain time series samples; Step S205: Train the temperature and humidity prediction model based on time series samples to obtain the trained temperature and humidity prediction model.

[0055] To address the continuous temporal variation of temperature and humidity in warehouse environments, the system employs a temperature and humidity prediction model based on a Long Short-Term Memory (LSTM) network structure. In implementation, historical data is first normalized, and time-series samples are constructed. Subsequently, the model is trained in Python, and the training results are exported for system use. During actual operation, the model can predict temperature and humidity trends over a future period. When the predicted results approach a set threshold, the system can issue an early warning, thus providing a basis for environmental control.

[0056] In the implementation process, the model training data mainly comes from historical environmental data collected in the early stages of the system, and is combined with continuous monitoring data over a certain period of time to construct the dataset. To ensure data quality, missing value imputation and outlier removal were performed on the original data before training, and normalization methods were used for data preprocessing.

[0057] During the model training phase, an LSTM network structure was built using a Python environment (TensorFlow / PyTorch). The model was trained and validated multiple times by continuously adjusting parameters such as the time step and the number of hidden layer nodes. In the initial experiments, the model exhibited some prediction lag. Subsequent adjustments to the window length and learning rate parameters resulted in smoother predictions that more closely reflected actual trends.

[0058] After training, the model is exported in a deployable format and can be used to make real-time predictions of temperature and humidity changes during system operation via interface calls.

[0059] It is understandable that, such as Figure 4 As shown, the intelligent warehousing and management method for dried tangerine peel also includes the following steps: Step S301: Preprocess the tangerine peel image data to obtain preprocessed image data, wherein the data enhancement processing includes noise reduction processing and color space conversion processing; Step S302: Based on the YOLOv8 target detection method, key features are extracted from the target moldy area in the preprocessed image data to obtain the detection results, which include the location information of the moldy area and the moldy category information; Step S303: Return the detection results to the visualization interface of the digital twin visualization module for annotation and display.

[0060] In addition to environmental parameter analysis, the system also utilizes a tangerine peel mold risk perception model for image recognition to detect mold growth on the surface of the tangerine peel. In the implementation process, image data is first acquired using a camera device and preprocessed as needed, such as denoising and color space conversion, to reduce interference from environmental factors. In the model part, the YOLOv8 object detection algorithm is used to analyze the images. Through training, the model is able to identify moldy areas and output their location and category information.

[0061] To improve detection performance, an attention mechanism was introduced during model optimization to enhance the extraction of features from key regions. Simultaneously, the loss function was adjusted to mitigate the impact of uneven sample distribution. During system operation, the model can perform real-time inference on images and annotate the recognition results in a visual interface, thus intuitively reflecting the distribution of mold.

[0062] Regarding image data, the system primarily uses cameras to collect images of dried tangerine peel samples in the warehouse environment, acquiring multiple sets of image data under different lighting conditions. To improve the model's generalization ability, data augmentation processing was also performed on some images, including brightness adjustment, rotation, and noise superposition.

[0063] During the training phase, transfer learning was performed based on the YOLOv8 model, with fine-tuning done on the pre-trained weights. Manual annotation tools were used to label moldy areas, generating the dataset required for training. In the initial training, the model performed poorly in recognizing small areas of mold. Subsequent training improved detection accuracy by adjusting the input resolution and introducing an attention mechanism to enhance features in key areas.

[0064] In actual deployment, the model runs on edge devices or host computers, analyzes real-time images by calling the inference interface, and returns the detection results to the visualization interface for annotation and display.

[0065] It is understandable that, such as Figure 5 As shown, the intelligent warehousing and management method for dried tangerine peel of the present invention further includes the following steps: Step S401: Periodically monitor the power of the photovoltaic energy storage device and display the remaining power and power supply mode in the visualization interface of the digital twin visualization module; Step S402: When the remaining power of the photovoltaic energy storage device is detected to be lower than the safety threshold, power consumption control is performed on non-critical equipment in the target tangerine peel storage, and a low power warning is triggered.

[0066] The system periodically monitors the power of the photovoltaic energy storage device and displays the remaining power and power supply mode in the visualization interface of the digital twin visualization module.

[0067] During the day, when there is sunlight, the photovoltaic energy storage device is powered by solar panels, which simultaneously charge the lithium battery. At night or during cloudy or rainy weather when sunlight is insufficient, the system automatically switches to lithium battery power to maintain the normal operation of sensors and communication modules. When the battery level is low, power consumption of some non-critical modules is controlled to prioritize core monitoring functions and trigger low battery warnings. During actual commissioning, power supply mode switching is achieved by setting voltage thresholds to avoid frequent power switching under critical conditions, thus improving system stability. Simultaneously, battery power is periodically monitored and displayed on a visualization platform, facilitating maintenance personnel's understanding of energy usage.

[0068] The above provides a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of the present invention.

Claims

1. A smart warehousing and management system for dried tangerine peel, characterized in that, include: The digital twin visualization module is used to construct a high-fidelity virtual mapping scene of the target tangerine peel warehouse and synchronize the real-time status data of the target tangerine peel warehouse in real time. An environmental sensing module is used to collect real-time status data of the target tangerine peel storage. The real-time status data includes environmental data and tangerine peel image data. The environmental data includes environmental temperature data, environmental humidity data, and gas concentration data. The environmental sensing module includes a temperature sensor, a humidity sensor, a gas sensor, and a camera device. The temperature sensor is used to collect the environmental temperature data, the humidity sensor is used to collect the environmental humidity data, the gas sensor is used to collect the gas concentration data, and the camera device is used to collect the tangerine peel image data. The reasoning module includes a tangerine peel mold risk perception model and a temperature and humidity prediction model. The temperature and humidity prediction model is used to output predicted data on the temperature and humidity change trends in the future period. The reasoning module is used to input the real-time status data and the predicted data into the tangerine peel mold risk perception model to obtain a mold risk index, and determine the mold risk level based on the mold risk index and a preset index threshold range, and then generate control instructions based on the mold risk level. The control module includes ventilation equipment and dehumidification equipment, and the control module is used to adjust the operating status of the ventilation equipment and dehumidification equipment according to the control command; The power supply module includes a photovoltaic energy storage device, which is used to provide power to the system.

2. The intelligent warehousing and management system for dried tangerine peel according to claim 1, characterized in that, The digital twin visualization module includes: The 3D modeling unit is used to construct a detailed model of the main building, equipment, sensors, and storage locations of the target tangerine peel warehouse based on the Unity3D engine. The virtual-real synchronization unit is used to connect to the MySQL database through a standardized interface and map the real-time status data to dynamic identifiers in the virtual scene; The interactive rendering unit is used to realize interactive visualization rendering of multiple viewing perspective modes, including scene free roaming mode, first-person inspection mode and AGV trajectory tracking mode.

3. The intelligent tangerine peel storage and management system according to claim 1, characterized in that, The expression for the mold risk index is as follows: Among them, MRI is the risk index for mold growth. For real-time ambient temperature in the warehouse, This is the optimal temperature for mold growth. For ambient relative humidity, The minimum humidity level is critical for mold germination. Duration of the abnormal environment.

4. The intelligent tangerine peel storage and management system according to claim 1, characterized in that, The tangerine peel mold risk perception model adopts a mechanism fusion RNN architecture, and the mold risk level output by the tangerine peel mold risk perception model includes intact level, slight mold level and moderate mold level.

5. The intelligent warehousing and management system for dried tangerine peel according to claim 1, characterized in that, The tangerine peel mold risk perception model is deployed on an edge device or a host computer. The tangerine peel mold risk perception model uses a target detection method based on YOLOv8 to identify the tangerine peel image data, and extracts key features of the target moldy area in the tangerine peel image data based on an attention mechanism to obtain the detection result. The detection result is then returned to the visualization interface of the digital twin visualization module for annotation and display.

6. The intelligent warehousing and management system for dried tangerine peel according to claim 1, characterized in that, The photovoltaic energy storage device includes: Solar panels are installed at a fixed tilt angle on the warehouse roof. Lithium-ion battery charging module for storing photovoltaic energy; The power management module is configured to: when there is sufficient sunlight, power is supplied by the solar panel and the lithium battery charging module is charged; when there is insufficient sunlight, the power supply is switched to the lithium battery charging module.

7. A method for intelligent warehouse management and control of dried tangerine peel, characterized in that, The method, applied to the intelligent tangerine peel storage and management system as described in any one of claims 1 to 6, comprises: The environmental perception module collects real-time status data of the target tangerine peel storage. The real-time status data and the predicted data are input into the tangerine peel mold risk perception model through the reasoning module to obtain the mold risk index. The reasoning module determines the mold risk level based on the mold risk index and a preset index threshold range. The reasoning module generates control instructions based on the mold risk level. The control module adjusts the operating status of the ventilation equipment and the dehumidification equipment according to the control instructions. The digital twin visualization module visually displays the warehouse status, mold risk level, and equipment operating status.

8. The intelligent warehousing and management method for dried tangerine peel according to claim 7, characterized in that, The temperature and humidity prediction model employs a long short-term memory network structure; the training method for the temperature and humidity prediction model includes: Acquire historical environmental data and continuous monitoring environmental data for a preset time period; The original training data is constructed based on the historical environmental data and the continuous monitoring environmental data. The original training data is subjected to missing value imputation and outlier removal to obtain preprocessed training data; The preprocessed training data is normalized to obtain time series samples; The temperature and humidity prediction model is trained based on the time series samples to obtain the trained temperature and humidity prediction model.

9. The intelligent warehousing and management method for dried tangerine peel according to claim 7, characterized in that, The method further includes: The tangerine peel image data is preprocessed to obtain preprocessed image data, wherein the data enhancement processing includes noise reduction processing and color space conversion processing; The YOLOv8-based target detection method extracts key features from the target moldy area in the preprocessed image data to obtain detection results, wherein the detection results include moldy area location information and mold category information; The detection results are returned to the visualization interface of the digital twin visualization module for annotation and display.

10. The intelligent warehousing and management method for dried tangerine peel according to claim 7, characterized in that, The method further includes: The power of the photovoltaic energy storage device is periodically monitored, and the remaining power and power supply mode are displayed in the visualization interface of the digital twin visualization module. When the remaining power of the photovoltaic energy storage device is detected to be lower than the safety threshold, power consumption control is applied to non-critical equipment in the target tangerine peel storage, and a low power warning is triggered.

Images