System for improving cold chain monitoring using IoT sensors and LSTM-based predictive analytics

Abstract

A system for intelligent monitoring of the cold chain and predictive environmental regulation; the system includes: a microcontroller configured with integrated Wi-Fi and Bluetooth communication protocols, the microcontroller serving as the central processing unit for acquiring, processing and transmitting sensor data; a plurality of sensors operationally coupled to the microcontroller, wherein the plurality of sensors includes at least one digital temperature sensor, at least one humidity sensor, at least one vibration sensor, at least one GPS module and at least one door access sensor, each of these sensors providing corresponding real-time data streams; a wireless communication subsystem comprising a GSM module operationally connected to the microcontroller, wherein the subsystem transmits encrypted sensor data to a cloud computing platform; a thermoelectric cooling unit comprising a Peltier element mechanically coupled to a heat sink and fan assembly, wherein the cooling unit is electrically coupled to a relay module controlled by the microcontroller to dynamically regulate the internal environment; and a cloud-based analytics engine that runs a Long Short-Term Memory (LSTM) model trained on historical sensor data, with the model predicting environmental anomalies before limit violations and transmitting predictive alerts via a dashboard to the microcontroller and authorized remote users.

Description

Technical field of the invention

[0001] The present invention relates to monitoring systems for cold chain logistics, in particular IoT-enabled predictive monitoring and control systems for the storage and transport of temperature-sensitive goods. The invention integrates environmental sensors, cloud-based data analysis, and artificial intelligence models—especially LSTM (Long Short-Term Memory) networks—for anomaly prediction and real-time correction. The invention also comprises a hardware device structure with sensors, actuators, microcontrollers, and thermoelectric control modules to ensure safe cold chain operation. Background of the invention

[0002] Cold chain logistics is an essential component of the pharmaceutical, food, and biotechnology industries. Product integrity depends heavily on maintaining predefined temperature and humidity conditions during storage and transport. Current industry practices rely heavily on manual inspections, static data loggers, or non-intelligent monitoring systems. These methods often lack real-time alerts, predictive anomaly detection, and automated corrective actions when safety limits are exceeded.

[0003] Conventional systems are largely reactive and often react with a delay, leading to deterioration in the quality of perishable goods, spoilage of vaccines, and non-compliance with strict global regulatory standards such as GDP (Good Distribution Practice) and GMP (Good Manufacturing Practice). Furthermore, existing commercial systems are expensive, proprietary, and not easily scalable for small and medium-sized enterprises.

[0004] With the advent of IoT and artificial intelligence, there is great potential to transform cold chain monitoring into a proactive, intelligent, and autonomous process. By integrating multiple sensor modalities with cloud-based machine learning, it becomes possible not only to monitor conditions in real time but also to predict potential deviations before they occur and to autonomously regulate storage conditions.

[0005] The present invention overcomes the limitations of conventional systems by introducing a cost-effective, modular IoT-based device integrated with LSTM prediction models, thus providing continuous environmental monitoring, anomaly detection, automated thermoelectric control, and secure remote access.

[0006] The cold chain industry is a vital pillar of global healthcare, agriculture, biotechnology, and food distribution. It encompasses the full spectrum of processes required for the storage, transportation, and delivery of temperature-sensitive goods such as vaccines, biologics, perishable foods, and high-value agricultural commodities under controlled environmental conditions. The stability, efficacy, and safety of these products depend heavily on maintaining specific temperature and humidity limits throughout their entire life cycle. Even slight deviations, often referred to as temperature excursions, can irreversibly compromise the chemical or biological integrity of such goods. For example, vaccines exposed to conditions outside their permissible range can lose their effectiveness, thereby jeopardizing public health initiatives.Similarly, perishable foods exposed to excessive heat or humidity can spoil quickly, leading to economic losses and health risks. Therefore, the cold chain is not only a logistical challenge but also a crucial infrastructure for global well-being.

[0007] Existing solutions in the cold chain sector can be broadly categorized into manual monitoring systems, standalone data loggers, wireless sensor networks, and advanced automated platforms with cloud integration. The oldest and still prevalent approach in many regions is manual monitoring using thermometers and logbooks. Here, personnel regularly check the temperature or humidity in storage rooms or transport containers and record the values ​​at fixed intervals. Although this method is cost-effective, it is highly susceptible to human error, negligence, and delays. A sudden deviation between two inspection intervals can go unnoticed and irreversibly endanger the shipment. Furthermore, manual methods do not allow for the real-time transparency that is essential in modern supply chains where shipments travel thousands of kilometers through diverse climate zones.

[0008] To address some of the shortcomings of manual monitoring, data loggers have become a widely used solution. These are compact, battery-powered devices placed in storage units or transport vehicles that continuously record environmental conditions. At the end of transport, the data is downloaded and reviewed to ensure that the goods have remained in safe conditions. While data loggers provide more detailed insights than manual records, they are inherently reactive. By the time anomalies are detected, it is usually too late to intervene, resulting in product losses. Furthermore, most conventional data loggers lack communication capabilities, meaning they cannot transmit real-time alerts to stakeholders. Therefore, their usefulness is limited to post-incident audits rather than preventive or corrective action.

[0009] The development of wireless sensor networks enabled real-time communication in cold chain monitoring. By using wireless sensors that transmitted data via Bluetooth, ZigBee, or proprietary radio protocols, stakeholders could remotely monitor temperature and humidity. This advancement allowed for better visibility and faster response times. However, these systems often faced challenges such as limited communication range, susceptibility to interference, and high infrastructure costs for setting up gateways, repeaters, and base stations. Many of these systems also ran on closed, proprietary platforms, resulting in vendor lock-in and high costs for expansion or customization. Furthermore, these networks were not optimized for long-distance mobility and were therefore better suited to warehouses than dynamic logistics operations.

[0010] With the advent of the Internet of Things (IoT), cold chain monitoring underwent another significant transformation. IoT solutions leveraged microcontrollers, wireless modules, and cloud platforms to enable seamless connectivity. Devices equipped with temperature, humidity, and location sensors could continuously upload data to cloud servers via Wi-Fi, GSM, or LTE networks. Cloud dashboards allowed stakeholders to access real-time conditions from any location, while alert systems could trigger SMS or email notifications in the event of anomalies. Despite its immense potential, many commercial IoT solutions encountered practical limitations. First, the cost of proprietary IoT systems remained high, limiting adoption among small and medium-sized enterprises. Second, these systems were primarily reactive, issuing alerts only after a deviation had occurred.Third, these platforms were often limited to a single environmental parameter such as temperature, ignoring other equally important variables like humidity, vibration, or unauthorized access. Finally, connectivity in remote regions with weak mobile network coverage remained a challenge, resulting in blind spots in surveillance.

[0011] Another important advancement was the integration of automatic temperature control into monitoring systems. Some solutions included thermoelectric coolers or cooling units that could be triggered if ambient conditions deviated from safe limits. While this provided a corrective mechanism, activation was typically based on static limit rules. Such rule-based systems lack the ability to anticipate changes or consider contextual factors such as shipping location, handling shocks, or external climate fluctuations. For example, a system might fail to proactively activate cooling if the environment heats up rapidly but the limit has not yet been exceeded. This leads to delayed response and product spoilage.Furthermore, static limit value systems can be energy inefficient, as they may unnecessarily trigger cooling when conditions would have stabilized naturally.

[0012] In recent years, machine learning and artificial intelligence have been explored as ways to improve predictive capabilities in cold chain monitoring. Predictive analytics can identify patterns in historical sensor data and forecast potential future anomalies, enabling stakeholders to take preventative action before breaches occur. However, most existing AI implementations in cold chain logistics are either in an early stage or limited to centralized, expensive platforms used by large enterprises. Small and medium-sized enterprises (SMEs) often cannot afford these solutions due to the high costs of hardware, licensing fees, and cloud integration. Furthermore, many of the predictive models used in these systems rely on simplified regression-based methods that fail to capture the complex temporal dependencies of environmental data.As a result, prediction accuracy remains low, which limits trust in and the usability of such systems in high-risk industries such as the pharmaceutical industry.

[0013] Another drawback of many current solutions lies in their limited integration capabilities. Cold chain security is affected not only by temperature and humidity, but also by external shocks, vibrations, route deviations, and unauthorized access. However, most available systems only monitor a subset of these variables and therefore offer limited transparency. For example, a system that only monitors temperature cannot detect improper handling during transport that could damage the packaging and indirectly compromise product integrity. Similarly, systems without GPS tracking cannot guarantee adherence to routes or provide accurate location data in emergencies.

[0014] Another critical drawback of current market solutions is their lack of openness and scalability. Many systems are tied to proprietary hardware and software ecosystems, making it difficult for users to customize the platform or add new sensors. This vendor lock-in not only increases costs but also stifles innovation. Small businesses and research institutions, on the other hand, often prefer open-source platforms that offer flexibility, transparency, and cost-efficiency. However, such open-source customizations are rarely available in commercial products, creating a gap between expensive proprietary solutions and low-cost systems with limited functionality.

[0015] Furthermore, global regulatory standards such as GDP and GMP mandate strict traceability, data integrity, and auditability in cold chain processes. While some high-end systems offer detailed data logging and compliance reporting, many low- and mid-tier solutions fail to meet these requirements. This creates compliance gaps for internationally operating companies, exposing them to potential legal violations, fines, and reputational damage. Even when logging is available, systems often lack tamper detection and security mechanisms to prevent unauthorized data manipulation, raising concerns about data authenticity.

[0016] Although significant progress has been made in the evolution of cold chain monitoring—from manual inspections to IoT-enabled platforms—existing solutions still have numerous drawbacks. Manual and logger-based systems are reactive and prone to errors. Wireless networks and IoT systems offer real-time visibility but are expensive, limited in scope, and more reactive than predictive. Automated refrigeration systems reduce risks but are often inefficient due to static thresholds. AI-based approaches show promise but remain either too expensive, too simplistic, or inaccessible to small businesses. Furthermore, most systems fail to integrate all the environmental and safety parameters required for comprehensive monitoring.These disadvantages highlight the urgent need for a cost-effective, modular, open-source, and predictive cold chain monitoring system that combines multi-sensor integration, cloud connectivity, AI-driven forecasting, real-time alerts, and automated control in a single scalable platform. Summary of the invention

[0017] The invention comprises an IoT-based cold chain monitoring system with a device structure containing multiple sensors, including temperature, humidity, vibration, GPS, and door access sensors, all connected to an ESP32 microcontroller. The microcontroller transmits sensor data via GSM / WLAN modules to a cloud platform such as ThingSpeak, where both real-time monitoring and AI-powered predictive analytics are performed. An LSTM model trained on historical sensor data predicts potential deviations and enables the system to send preventative alerts to the relevant parties.

[0018] The device also features a thermoelectric cooling module (Peltier-based) controlled via relay interfaces, which autonomously regulates the internal storage environment. A Django-based dashboard provides live visualization, historical logging, and manual override functions. Thanks to GSM fallback, the system can also be used in environments with limited connectivity, thus ensuring reliability and scalability.

[0019] This invention transforms cold chain management into an intelligent, predictive, and autonomous process that reduces product spoilage, improves regulatory compliance, and ensures energy-efficient operation.

[0020] The main objective of the present invention is to provide an intelligent, reliable, and cost-effective cold chain monitoring system that overcomes the limitations of existing technologies and ensures continuous, real-time visibility and predictive control of environmental conditions. Unlike conventional systems that operate reactively and rely on post-processing analysis, the invention aims to proactively prevent spoilage and loss of temperature-sensitive goods by integrating predictive analytics, anomaly detection, and automated thermal regulation into a single, unified platform. A further objective of the invention is to develop a modular and scalable device architecture using cost-effective, commercially available sensors and microcontrollers. This makes the system accessible to small and medium-sized enterprises that would otherwise be unable to afford expensive proprietary solutions.The invention also aims to provide a multi-parameter monitoring system that not only records temperature and humidity but also integrates vibration, GPS and door access sensors to provide a comprehensive overview of cold chain integrity, including handling quality, route adherence and cargo safety.

[0021] A further objective of the invention is predictive anomaly detection through the use of advanced machine learning models, in particular Long Short-Term Memory (LSTM) networks. These capture complex temporal dependencies in environmental data and predict limit exceedances before they occur. This predictive functionality ensures that corrective measures, such as the activation of thermoelectric cooling modules, can be initiated early, thereby safeguarding product quality and reducing the risk of spoilage. Another objective is the establishment of a secure, cloud-based data infrastructure that enables remote monitoring, real-time dashboards, and historical data logging in accordance with international standards such as GDP and GMP.By ensuring traceability, transparency and verifiability of environmental conditions throughout the entire logistics lifecycle, the invention improves compliance with legal regulations and strengthens trust in global supply chains.

[0022] The invention also aims to integrate intelligent energy optimization mechanisms by activating thermoelectric cooling only when needed. This reduces power consumption and makes the system suitable for mobile, battery-powered applications in the transportation sector. Simultaneously, the integration of GSM-based fallback communication ensures reliability in environments with limited connectivity, guaranteeing uninterrupted monitoring even in remote or rural areas. A further objective is to improve the security of cold chain operations through immediate alerts in the event of unauthorized access, vibrations, or route deviations. This allows those involved to intervene quickly and minimize risks.By utilizing open-source platforms and software frameworks, the invention also promotes adaptability, customization, and research-driven improvements, providing a future-proof platform for academic, industrial, and commercial stakeholders. The overarching goal of the invention is to create a holistic, intelligent, and accessible cold chain monitoring system that combines the strengths of IoT hardware, cloud-based analytics, and artificial intelligence in a seamless device and methodology, thereby ensuring product safety, compliance, and operational efficiency in an increasingly complex and globalized logistics environment. BRIEF DESCRIPTION OF THE FIGURE

[0023] These and other features, aspects, and advantages of the present invention will be better understood if the following detailed description is read with reference to the accompanying drawing, in which the same symbols consistently represent the same parts. The following applies: Fig. Figure 1 shows a block diagram of a system and device for improving cold chain monitoring using IoT sensors and LSTM-based predictive analytics.

[0024] Experts will also recognize that the elements in the drawing are shown for the sake of simplicity and are not necessarily to scale. For example, the flowcharts illustrate the process by highlighting the main steps to enhance understanding of the aspects of this disclosure. Furthermore, with regard to the design of the device, one or more components of the device may be represented in the drawing by conventional symbols, and the drawing may show only the specific details relevant to understanding the embodiments of this disclosure, so as not to clutter the drawing with details that are readily apparent to those skilled in the art after reading this description. Detailed description of the invention

[0025] For a better understanding of the inventive principles, reference is made below to the embodiment shown in the drawing, which is described in specific terminology. However, this does not limit the scope of the invention. Changes and further modifications of the illustrated system, as well as further applications of the inventive principles, are possible, as would normally occur to a person skilled in the art in the field of invention.

[0026] It is clear to the person skilled in the art that the preceding general description and the following detailed description are exemplary and explanatory of the invention and are not intended as a limitation of it.

[0027] References in this specification to “an aspect”, “another aspect”, or similar expressions mean that a particular feature, structure, or property described in connection with the embodiment is included in at least one embodiment of the present disclosure. Therefore, occurrences of the expressions “in one embodiment”, “in another embodiment”, and similar expressions in this specification may all refer to the same embodiment, but need not.

[0028] The terms "includes," "include," or other variations thereof are intended to cover non-exclusive inclusion, so that a process or method that includes a list of steps may not only contain those steps but may also include other steps not expressly listed or inherent in such process or method. Likewise, the statement "includes..." in the case of one or more devices, subsystems, elements, structures, or components does not, without further limitations, preclude the existence of other devices, subsystems, elements, structures, components, or additional devices, subsystems, elements, structures, or components.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by a person skilled in the art in the field of the invention. The systems, methods, and examples provided herein serve only for illustration and are not to be construed as limitations.

[0030] Embodiments of the present disclosure are described in detail below with reference to the attached drawing.

[0031] In Fig.Figure 1 is a block diagram of a system and device for improving cold chain monitoring using IoT sensors and LSTM-based predictive analytics. The system 100 comprises: an ESP32 microcontroller (102) configured with integrated Wi-Fi and Bluetooth communication protocols, the microcontroller serving as the central processing unit for acquiring, processing, and transmitting sensor data; several sensors (104) operationally connected to the microcontroller, including at least one DS18B20 digital temperature sensor, at least one DHT11 humidity sensor, at least one SW420 vibration sensor (104a), at least one NEO-6M GPS module, and at least one door access sensor (104b), each of these sensors providing corresponding real-time data streams;a wireless communication subsystem (106) comprising a GSM module SIM800A operationally connected to the microcontroller, wherein the subsystem transmits encrypted sensor data to a cloud computing platform; a thermoelectric cooling unit (108) comprising a Peltier element mechanically coupled to a heat sink and fan assembly, wherein the cooling unit is electrically coupled to a relay module controlled by the microcontroller to dynamically regulate the internal environment;and a cloud-based analysis engine (110) that executes a Long Short-Term Memory (LSTM) model trained on historical sensor data, the model predicting environmental anomalies prior to limit exceedances and transmitting predictive alerts to the microcontroller and authorized remote users via a Django-based dashboard (110a), thus enabling real-time monitoring, preventive intervention, and compliance with GDP and GMP standards.

[0032] In one embodiment, the ESP32 microcontroller (102) is programmed with firmware written in C / C++ in the Arduino IDE environment. This firmware includes libraries such as DHT.h and OneWire.h for interfacing with the humidity and temperature sensors, WiFiClientSecure for encrypted communication, and logical control routines for evaluating the sensor values ​​based on predefined security thresholds. The microcontroller autonomously outputs control signals to the relay module to activate or deactivate the thermoelectric cooling unit based on real-time sensor inputs or predictive warnings generated by the LSTM model.

[0033] In one embodiment, the cloud computing platform includes the ThingSpeak IoT platform, which connects to the ESP32 microcontroller via REST APIs. The platform is configured to receive, store, and visualize sensor data across multiple fields. A Django-based dashboard, hosted on a secure server, queries the ThingSpeak API to display live environmental data, GPS-based geolocation maps, vibration events, door access status, and predictive analytics results to authenticated users with login credentials.

[0034] In one embodiment, the thermoelectric cooling unit (108) comprises a Peltier element thermally connected to an aluminum heat sink and a forced convection fan. The unit is mounted in an insulated enclosure, with the relay module providing a semiconductor circuit controlled by GPIO outputs of the ESP32 microcontroller, and the microcontroller activating the Peltier element when the sensor values ​​exceed an upper threshold of 22 °C or when the LSTM model predicts a violation within a 30-minute time window, thus enabling proactive thermal regulation with minimal latency.

[0035] In one embodiment, the vibration sensor (104a) includes an SW420 module configured to detect abnormal shocks or continuous vibration patterns. The sensor is calibrated to trigger interrupt signals to the ESP32 microcontroller when the acceleration exceeds a defined safety threshold. Upon receiving such signals, the microcontroller immediately generates SMS and email notifications containing the shipment's GPS coordinates to authorized users, enabling a rapid response to potential mishandling incidents during transport.

[0036] In one embodiment, the door sensor (104b) comprises a reed switch mounted on the cargo or container door, configured to change its state upon unauthorized opening during transport. The ESP32 microcontroller logs the access event locally, transmits an alert to the cloud platform, and adds geolocation coordinates from the GPS module, thereby enabling tamper detection, traceability, and security assurance for transported goods.

[0037] In one embodiment, the LSTM-based predictive analysis engine is trained using historical temperature and humidity data sets recorded over multiple shipments. These data sets are preprocessed with normalization and time-series encoding. The LSTM network comprises several hidden layers configured with dropout regularization to prevent overfitting. The engine outputs predicted temperature and humidity trends, the performance of which is measured by the Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) metrics. These outputs are transmitted to the microcontroller to initiate corrective actions early.

[0038] In one embodiment, the power supply subsystem comprises a regulated 5 V DC power source derived from a rechargeable lithium-ion battery pack, wherein the battery pack is coupled with a voltage regulator circuit and optional integration of an uninterruptible power supply (UPS), the system being optimized for low power consumption operation by performing sensor data acquisition during the duty cycle and activating the thermoelectric cooler only upon threshold exceedance or predictive triggers, thereby extending battery life during long-distance transport.

[0039] In one embodiment, the Django-based dashboard (110a) comprises a secure web application implemented in Python, HTML, and JavaScript. The dashboard is configured with authentication protocols to restrict access to authorized users. The dashboard displays real-time graphs of environmental conditions, GPS route visualization with geofencing, graphs of predicted and actual conditions, and a manual override interface for remotely enabling or disabling the thermoelectric cooling unit.

[0040] In one embodiment, multi-sensor integration enables compliance with international GDP and GMP standards through continuous data logging of temperature, humidity, vibration, and access events. The data is time-stamped, encrypted, and stored both locally and in the cloud. This ensures auditability, traceability, and regulatory compliance in various cold chain logistics environments.

[0041] The present invention is described in more detail with particular reference to the system architecture and the technical framework that supports the intelligent operation of the proposed cold chain monitoring device. The system integrates a modular arrangement of IoT sensors, communication subsystems, control logic, cloud-based analytics, and predictive machine learning models to enable proactive monitoring, anomaly prediction, and autonomous regulation of environmental conditions in cold chain logistics.

[0042] The core of the system is the ESP32 microcontroller, which serves as the central interface for data acquisition, control, and communication. The microcontroller is equipped with firmware written in C / C++, developed using the Arduino IDE, and communicates directly with several environmental sensors via digital and analog input channels. The DS18B20 digital temperature sensor provides highly accurate temperature measurements, while the DHT11 humidity sensor measures relative humidity. These environmental measurements are complemented by information on the physical condition of an SW420 vibration sensor, which outputs real-time motion data indicating impacts or improper handling, and a reed-based door sensor, which signals unauthorized access during transport. Additionally, a NEO-6M GPS module continuously transmits geolocation coordinates, enabling tracking and geofencing functions.All sensor values ​​are synchronized with precise timestamps and forwarded to the microcontroller, where preliminary filter and threshold checks are performed.

[0043] The firmware implemented on the ESP32 is based on a technical workflow that controls the system's real-time decisions. The first phase of this workflow is data acquisition, where sensor values ​​are read at fixed intervals defined by the sampling rate. A preprocessing routine normalizes the raw sensor values ​​and applies logical constraints to identify erroneous or out-of-range readings caused by noise or sensor malfunctions. These cleaned data streams are then transmitted to a cloud platform such as ThingSpeak using the microcontroller's Wi-Fi capability. In situations where Wi-Fi connectivity is unavailable, the SIM800A GSM module provides a redundant communication channel, ensuring uninterrupted data flow to the cloud. The transmission is protected against unauthorized interception by WiFiClientSecure encryption.

[0044] Once the sensor data reaches the cloud platform, it is processed and visualized via APIs in near real-time. This data is fed into a Django-based dashboard. This dashboard displays temperature, humidity, vibration, door status, and geolocation data graphically and in tabular form, enabling authorized users to remotely monitor shipments. It also offers route visualization, anomaly alerts, and a manual override function that allows operators to directly activate or deactivate the thermoelectric cooling unit as needed. However, the novelty of the invention lies not only in its monitoring and alerting capabilities but also in its predictive intelligence, achieved through the integration of a long short-term memory (LSTM) neural network.

[0045] The LSTM model embedded in the cloud analytics engine is trained using historical temperature and humidity datasets from multiple shipping cycles. During training, data preprocessing includes normalizing sensor values ​​and encoding temporal dependencies across time-series sequences. The LSTM architecture comprises several hidden layers, each designed to capture long-term and short-term dependencies in the environmental data. Dropout regularization is applied to prevent overfitting, and the model is optimized using backpropagation through time (BPTT). Once deployed, the trained LSTM model continuously receives new sensor data streams and forecasts environmental trends over a defined time horizon, typically 30 to 60 minutes in advance.

[0046] The LSTM's forecast results are compared with predefined safety thresholds. If the model predicts a temperature increase above 22°C within the forecast window, the system sends a preventative control signal to the ESP32 microcontroller. This activates the thermoelectric cooling unit via the relay interface before the threshold is exceeded in real time. The thermoelectric cooling unit consists of a Peltier element thermally coupled to an aluminum heat sink and a forced-convection fan, which quickly lowers the internal temperature. Conversely, if the temperature forecast falls below 19°C, the cooling unit is deactivated to conserve energy. In this way, the technology uses predictive intelligence to maintain conditions within the safety range while simultaneously optimizing energy consumption.

[0047] In addition to predicting temperature and humidity, the system also detects anomalies in vibrations and door states. The vibration sensor outputs analog values ​​corresponding to acceleration events. The firmware uses a threshold-based interrupt routine that immediately reports shocks exceeding calibrated limits. Upon detection, the system packages the vibration intensity, timestamp, and GPS coordinates into an alert message, which is sent via SMS and email to selected stakeholders. A similar approach is used for door sensor events: If the reed switch detects unauthorized opening, the microcontroller immediately logs the event, records the geolocation, and sends an alert. These anomaly detections for security and handling operate independently of predictive forecasts and ensure that mechanical violations or access breaches are reported in real time.

[0048] The system's control logic therefore operates in two complementary layers. The first is the reactive layer, where exceedances of real-time sensor readings trigger immediate control actions and warnings. The second is the predictive layer, where the LSTM model forecasts potential anomalies and initiates corrective measures in advance. The integration of these two layers ensures that the system does not only react after a damaging event, but actively prevents deviations. This two-layer design significantly increases reliability and robustness compared to conventional cold chain monitoring solutions.

[0049] Another important aspect of the technical framework is energy optimization. The ESP32 firmware operates in energy-saving duty-cycling mode, in which sensors are polled at adaptive intervals depending on environmental stability. For example, if conditions remain stable within the safety zone, the polling interval is extended to conserve battery power. If instability is detected, the polling frequency is increased to improve responsiveness. Furthermore, the thermoelectric cooling unit is only activated when absolutely necessary—either due to real-time violations or predictive triggers. This ensures that no battery capacity is wasted on unnecessary cooling.

[0050] All captured data, including real-time values, predicted trends, and system actions, are logged both locally in the microcontroller's non-volatile memory and in the cloud. Local logging ensures continuity during network outages, while cloud logging supports traceability, compliance, and auditability. Data is encrypted before transmission to guarantee data integrity and prevent tampering—a crucial requirement for compliance with Good Distribution Practice (GDP) and Good Manufacturing Practice (GMP) regulations.

[0051] In operation, the system enables seamless monitoring of the cold chain throughout storage and transport phases. For example, during vaccine transport, the system continuously monitors temperature, humidity, and handling conditions. If the LSTM model predicts a potential temperature deviation, the thermoelectric unit is activated early, thus preventing vaccine degradation. If the container is tampered with during transport, the door sensor immediately sends a geolocated alert, allowing intervention before theft or contamination occurs. The integrated technology therefore ensures that all critical risks—thermal, environmental, mechanical, and safety-related—are addressed proactively and in real time.

[0052] Through the detailed interplay of IoT hardware, secure communication, cloud-based visualization, predictive machine learning, and intelligent control logic, the invention enables groundbreaking advancements in cold chain monitoring. The described technology not only ensures compliance and security but also minimizes energy consumption, reduces human dependence, and offers a scalable, cost-effective solution suitable for use in a wide range of industries—from pharmaceuticals and agriculture to food logistics.

[0053] The present invention is described in more detail with particular reference to the system architecture and the technical framework that supports the intelligent operation of the proposed cold chain monitoring device. The system integrates a modular arrangement of IoT sensors, communication subsystems, control logic, cloud-based analytics, and predictive machine learning models to enable proactive monitoring, anomaly prediction, and autonomous regulation of environmental conditions in cold chain logistics.

[0054] The core of the system is the ESP32 microcontroller, which serves as the central interface for data acquisition, control, and communication. The microcontroller is equipped with firmware written in C / C++, developed using the Arduino IDE, and communicates directly with several environmental sensors via digital and analog input channels. The DS18B20 digital temperature sensor provides highly accurate temperature measurements, while the DHT11 humidity sensor measures relative humidity. These environmental measurements are complemented by information on the physical condition of an SW420 vibration sensor, which outputs real-time motion data indicating impacts or improper handling, and a reed-based door sensor, which signals unauthorized access during transport. Additionally, a NEO-6M GPS module continuously transmits geolocation coordinates, enabling tracking and geofencing functions.All sensor values ​​are synchronized with precise timestamps and forwarded to the microcontroller, where preliminary filter and threshold checks are performed.

[0055] The firmware implemented on the ESP32 is based on a technical workflow that controls the system's real-time decisions. The first phase of this workflow is data acquisition, where sensor values ​​are read at fixed intervals defined by the sampling rate. A preprocessing routine normalizes the raw sensor values ​​and applies logical constraints to identify erroneous or out-of-range readings caused by noise or sensor malfunctions. These cleaned data streams are then transmitted to a cloud platform such as ThingSpeak using the microcontroller's Wi-Fi capability. In situations where Wi-Fi connectivity is unavailable, the SIM800A GSM module provides a redundant communication channel, ensuring uninterrupted data flow to the cloud. The transmission is protected against unauthorized interception by WiFiClientSecure encryption.

[0056] Once the sensor data reaches the cloud platform, it is processed and visualized via APIs in near real-time. This data is fed into a Django-based dashboard. This dashboard displays temperature, humidity, vibration, door status, and geolocation data graphically and in tabular form, enabling authorized users to remotely monitor shipments. It also offers route visualization, anomaly alerts, and a manual override function that allows operators to directly activate or deactivate the thermoelectric cooling unit as needed. However, the novelty of the invention lies not only in its monitoring and alerting capabilities but also in its predictive intelligence, achieved through the integration of a long short-term memory (LSTM) neural network.

[0057] The LSTM model embedded in the cloud analytics engine is trained using historical temperature and humidity datasets from multiple shipping cycles. During training, data preprocessing includes normalizing sensor values ​​and encoding temporal dependencies across time-series sequences. The LSTM architecture comprises several hidden layers, each designed to capture long-term and short-term dependencies in the environmental data. Dropout regularization is applied to prevent overfitting, and the model is optimized using backpropagation through time (BPTT). Once deployed, the trained LSTM model continuously receives new sensor data streams and forecasts environmental trends over a defined time horizon, typically 30 to 60 minutes in advance.

[0058] The LSTM's forecast results are compared with predefined safety thresholds. If the model predicts a temperature increase above 22°C within the forecast window, the system sends a preventative control signal to the ESP32 microcontroller. This activates the thermoelectric cooling unit via the relay interface before the threshold is exceeded in real time. The thermoelectric cooling unit consists of a Peltier element thermally coupled to an aluminum heat sink and a forced-convection fan, which quickly lowers the internal temperature. Conversely, if the temperature forecast falls below 19°C, the cooling unit is deactivated to conserve energy. In this way, the technology uses predictive intelligence to maintain conditions within the safety range while simultaneously optimizing energy consumption.

[0059] In addition to predicting temperature and humidity, the system also detects anomalies in vibrations and door states. The vibration sensor outputs analog values ​​corresponding to acceleration events. The firmware uses a threshold-based interrupt routine that immediately reports shocks exceeding calibrated limits. Upon detection, the system packages the vibration intensity, timestamp, and GPS coordinates into an alert message, which is sent via SMS and email to selected stakeholders. A similar approach is used for door sensor events: If the reed switch detects unauthorized opening, the microcontroller immediately logs the event, records the geolocation, and sends an alert. These anomaly detections for security and handling operate independently of predictive forecasts and ensure that mechanical violations or access breaches are reported in real time.

[0060] The system's control logic therefore operates in two complementary layers. The first is the reactive layer, where exceedances of real-time sensor readings trigger immediate control actions and warnings. The second is the predictive layer, where the LSTM model forecasts potential anomalies and initiates corrective measures in advance. The integration of these two layers ensures that the system does not only react after a damaging event, but actively prevents deviations. This two-layer design significantly increases reliability and robustness compared to conventional cold chain monitoring solutions.

[0061] Another important aspect of the technical framework is energy optimization. The ESP32 firmware operates in energy-saving duty-cycling mode, in which sensors are polled at adaptive intervals depending on environmental stability. For example, if conditions remain stable within the safety zone, the polling interval is extended to conserve battery power. If instability is detected, the polling frequency is increased to improve responsiveness. Furthermore, the thermoelectric cooling unit is only activated when absolutely necessary—either due to real-time violations or predictive triggers. This ensures that no battery capacity is wasted on unnecessary cooling.

[0062] All captured data, including real-time values, predicted trends, and system actions, are logged both locally in the microcontroller's non-volatile memory and in the cloud. Local logging ensures continuity during network outages, while cloud logging supports traceability, compliance, and auditability. Data is encrypted before transmission to guarantee data integrity and prevent tampering—a crucial requirement for compliance with Good Distribution Practice (GDP) and Good Manufacturing Practice (GMP) regulations.

[0063] In operation, the system enables seamless monitoring of the cold chain throughout storage and transport phases. For example, during vaccine transport, the system continuously monitors temperature, humidity, and handling conditions. If the LSTM model predicts a potential temperature deviation, the thermoelectric unit is activated early, thus preventing vaccine degradation. If the container is tampered with during transport, the door sensor immediately sends a geolocated alert, allowing intervention before theft or contamination occurs. The integrated technology therefore ensures that all critical risks—thermal, environmental, mechanical, and safety-related—are addressed proactively and in real time.

[0064] Through the detailed interplay of IoT hardware, secure communication, cloud-based visualization, predictive machine learning, and intelligent control logic, the invention enables groundbreaking advancements in cold chain monitoring. The described technology not only ensures compliance and security but also minimizes energy consumption, reduces human dependence, and offers a scalable, cost-effective solution suitable for use in a wide range of industries—from pharmaceuticals and agriculture to food logistics.

[0065] The drawing and the preceding description show examples of embodiments. Those skilled in the art will recognize that one or more of the described elements can be combined to form a single functional element. Alternatively, certain elements can be divided into several functional elements. Elements of one embodiment can be added to another embodiment. For example, the sequence of the processes described here can be changed and is not limited to the manner described here. Furthermore, the actions of a flowchart need not be implemented in the sequence shown; nor does it necessarily have to be performed by all actions. Actions that are not dependent on other actions can also be performed in parallel with the other actions. The scope of the embodiments is in no way limited by these specific examples.Numerous variations are possible, whether explicitly stated in the specification or not, such as differences in structure, dimensions, and material usage. The range of embodiments is at least as broad as specified in the following claims.

[0066] Advantages, further benefits, and solutions to problems have been described above with regard to specific embodiments. However, the advantages, benefits, solutions to problems, and all components that may lead to a particular advantage or solution occurring or becoming more apparent are not to be construed as critical, necessary, or essential features or components of any or all claims. REFERENCES 100 A system and device for improving cold chain monitoring using IoT sensors and LSTM-based predictive analytics. 102 microcontrollers 104 Variety of Sensors 104a Vibration sensor 104b Door access sensor 106 Wireless Communication Subsystem 108 Thermoelectric cooling unit 110 Cloud-based analytics engine 110a Dashboard

Claims

[1] A system for intelligent monitoring of the cold chain and for predictive environmental regulation, the system includes: a microcontroller configured with integrated Wi-Fi and Bluetooth communication protocols, the microcontroller serving as the central processing unit for acquiring, processing and transmitting sensor data; a plurality of sensors operationally coupled to the microcontroller, wherein the plurality of sensors includes at least one digital temperature sensor, at least one humidity sensor, at least one vibration sensor, at least one GPS module and at least one door access sensor, each of these sensors providing corresponding real-time data streams; a wireless communication subsystem comprising a GSM module operationally connected to the microcontroller, wherein the subsystem transmits encrypted sensor data to a cloud computing platform; a thermoelectric cooling unit comprising a Peltier element mechanically coupled to a heat sink and fan assembly, wherein the cooling unit is electrically coupled to a relay module controlled by the microcontroller to dynamically regulate the internal environment; and a cloud-based analytics engine that runs a Long Short-Term Memory (LSTM) model trained on historical sensor data, with the model predicting environmental anomalies before limit violations and transmitting predictive alerts via a dashboard to the microcontroller and authorized remote users. [2] System according to claim 1, wherein the cloud computing platform comprises an IoT platform connected to the microcontroller, the platform being configured to receive, store and visualize sensor data in multiple fields, and wherein a dashboard provided on a secure server queries the IoT platform to display live environment data, GPS-based geolocation maps, vibration events, door access status and results of predictive analytics to authenticated users with access credentials. [3] System according to claim 1, wherein the thermoelectric cooling unit comprises a Peltier device thermally connected to an aluminum heat sink and a forced convection fan, wherein the unit is mounted in an insulated housing, and wherein the microcontroller activates the Peltier device when the sensor values ​​exceed an upper threshold of 22 °C or when the LSTM model predicts a break within a time window of 30 minutes. [4] System according to claim 1, wherein the vibration sensor is configured to detect abnormal shocks or continuous vibration patterns, wherein the sensor is calibrated to trigger interrupt signals to the microcontroller when the acceleration exceeds a defined safety threshold, wherein upon receiving such signals the microcontroller immediately generates SMS and email notifications with the GPS coordinates of the shipment to authorized users, thus enabling a rapid response to possible incidents of mishandling during transport. [5] System according to claim 1, wherein the door sensor comprises a reed switch mounted on the cargo or container door, configured to change its state upon unauthorized opening during transport, wherein the microcontroller locally logs the access event, transmits a warning to the cloud platform and adds geolocation coordinates from the GPS module, thereby enabling tamper detection, traceability and security assurance for transported goods. [6] System according to claim 1, wherein the LSTM-based predictive analysis engine is trained on historical temperature and humidity data sets recorded over multiple deliveries, wherein the data sets are pre-processed with normalization and time-series encoding, wherein the LSTM network comprises several hidden layers configured with dropout regularization to prevent overfitting. [7] System according to claim 1, wherein the power supply subsystem comprises a regulated 5 V DC power source derived from a rechargeable lithium-ion battery pack, wherein the battery pack is coupled with a voltage regulator circuit and optional integration of an uninterruptible power supply (UPS), wherein the system is optimized for low power consumption operation by controlling the sensor data acquisition in the duty cycle and activating the thermoelectric cooler only when thresholds are exceeded or predictive triggers occur, thereby extending battery life during long-distance transport.

Images