Cold Storage Remote Monitoring System Development

This is because the engineering task was to extend an existing, mature process-control architecture that has been used successfully in an agricultural environment, in particular crop storage, for well over four decades, into a cloud connected multiple sites mobile monitoring system.

This case study looks at the entire system – the edge controller with its embedded MCU class, the sensor and its electronics, the Netmon network gateway, the logging unit, the telemetry channel, and the ColdStores React Native/Flutter class mobile application.

Range was developed by Farmex and has been used for over 40 years in barns and crop storage buildings with the controller fitted to over 100 farms in UK alone.

The philosophy of its design is modular where the controller handles the sensing, logic processing and alarms, while the drivers take care of the high current switching required for the fans, heaters, dampers and refrigeration compressors. The architecture of the design is important for cold store monitoring where the "brains" of the controller remain safely powered from the protected 24 VAC supply which is battery backed, while cheaper driver boards handle the switching load stress.

The ColdStores app sits at the top of this stack. It does not replace the controller's control loop; it consumes the controller's data and exposes multi-site visibility, historical trends, and daily notifications to operators.

3.1 Controller Unit and MCU Class

Each unit is composed of the processor and display board, switch card and applications board. It supports 8 analogue and digital inputs and 8 optically isolated outputs, including peer-to-peer networking functionality. The exact MCU type is not disclosed on the public data sheets, however, the documented electrical behavior defines a narrow list of possible devices.

Inputs of the controller are defined as 0-5VDC with 10-bit resolution (or about 5mV per step). The sensors (TS4) are calibrated by the controller with accuracy of 0.1°C. Digital inputs are 16-bit roll-over counters and can receive timed or pulse signals from meters and auger run-time up to 300 Hz.

These requirements point to an industrial-grade 8/16/32-bit microcontroller or microprocessor with an integrated multi-channel ADC, UART/RS232, and sufficient non-volatile or battery-backed RAM for configuration and short-term logging.

The microcontroller should perform a real-time supervisory loop, which analyses the state of alarms every two seconds, manages the menu-driven display interface, and calibrates the coefficients of several sensors. The design leans more toward being a supervisory microcontroller with an external memory and display controller rather than a simple one-chip hobbyist microcontroller since reliability, determinism, and availability come first, not computational power.

3.2 Sensor Front End

The sensing layer does not have any specific types. In this case, the key sensing parameter is the temperature, which is provided by thermistors' sensors without field calibration since the control unit already knows the correct correction curve.

Other accepted sensing parameters include humidity, CO2, NH3, static pressure, air velocity, weight, illumination, air quality, wind speed, and rain; and ColdStores app offers such parameters as relative humidity, CO2, and electrical consumption "where available". In this way, for the cold store applications, the controller can be used in crop storage, pharmaceutical cold rooms, and multipurpose buildings without any changes in hardware.

A practical constraint is that the analogue front end operates on a 0–5 V reference, with galvanic isolation recommended where inputs leave the controller enclosure. This is a deliberate robustness choice: cold-store environments combine compressor-induced electrical noise, condensation, and long sensor cable runs.

3.3 Output and Driver Stage

However, the outputs of the eight controllers are not used for switching loads directly. Instead, 0-100% "Level" commands are issued using the software. Then, the commands are translated to switching schemes that are fed to the driver module that operates optically isolated. There are different functions provided by drivers, such as 0-10 V DC control, on/off, interval time, three-phase voltage, and alarm.

In the case of monitoring the cold store, it can be stated that the very same platform that provides the compressor run-time and refrigeration results analysis through the application executes the compressor control algorithm.

3.4 Networking and Netmon Gateway

The unit level entity is the node, while the site level aggregator is the Netmon. The Netmon interfaces with up to 29 nodes through a peer-to-peer network cable, with each node having an address ranging from 1-30 and the Netmon normally on address 30. Each unit updates its status (alarms and otherwise) to the Netmon every 2 seconds. The Netmon does not need to recompute the control logic, but will be able to tell whether there is loss of communication by any means including a cable being cut, when it does not receive the "OK" status from the unit.

Netmon serves as the network's real-time clock/calendar, providing time and date to all units in the network, and supports RS232 Modem connection allowing for remote communications. This is the historical link from the network to the outside world; in ColdStores design, this is where the site data gets into the cloud pipeline for the mobile application.

This feature relies on the Netmon's core function. Logging can take place in three categories – sensor inputs, output levels and ASMs.

Sensors may be logged as analogue values (from 0 to 1023 ADC counts in engineering units) or as digital time/pulse values.

Outputs may be logged as either on/off state or percentage level, however the technical documentation highlights the fact that the logging is based on open loop output and records the commands given to the driver rather than the load response.

So, for the ColdStore the useful set of telemetry consists of store temperature, crops temperature (as well as the channels of individual sensors), the set-point, outside temperature, compressor hours, relative humidity, CO2 and energy consumption. Daily statistics such as minimum, maximum and average are calculated and presented in the app together with the 24-hour and 7-day interactive graphs. Those customers who use the Renewable Energy Optimisation (REO) package receive information on import, export and kWh produced and the ColdStore becomes an energy management tool as well.

5.1 Application Architecture and Features

ColdStores is a cross-platform mobile application that gives users a multi-site view using the most recent data from each connected site as well as the latest data for the particular site in a site-view page that includes ambient data, refrigeration performance, and energy data.

The charts provided cover the past 24 hours or seven days. There is an option to display or hide the data series in the chart controls. Notifications are received every day whenever there is new data available. From the latest updates, it can be seen that the emphasis has been put on making the application accessible.

From an architectural perspective, the application itself is a read-mostly monitoring client. It sends no commands to the controller, it is for visibility, analysis of trends, and notifications only.

This segregation is critical, since the control loop stays in place on the tested hardware, and the mobile application uses a telemetry stream generated from the Netmon logging flow.

5.2 Security and Privacy Posture

For an industrial monitoring app, this is a conservative and appropriate posture: the operational telemetry belongs to the cold-store operator, and the app's role is to render it, not to monetise it. The deletion mechanism is also consistent with GDPR-style requirements for a UK-based developer serving international customers.

The firmware of the controller is a real-time supervisory program and not an operating-system type task. The timing constraint for this application is the 2-second cycle required for alarm reporting among up to 30 nodes.

All of the following functions must be performed in a single cycle: the analogue and digital input scan, calibration of the sensors, alarm threshold checking, display driver, menus, and communication between peers via the network using a 10-bit ADC microcontroller.

The longevity of this system in use is proof in itself of firmware optimized for deterministic behavior and maintenance rather than for features.

On the connectivity side, the RS232 modem interface on the Netmon is the legacy telemetry path. The ColdStores cloud service is the modern replacement: a gateway or modem translates the data stream into an IP-based telemetry feed, and the mobile app consumes that feed through a cloud API. This layered design means the app can evolve independently of the controller firmware, and the controller firmware does not need to understand HTTP, MQTT, or mobile authentication.

The toughest engineering challenge in a system like this isn't the mobile user interface, it is maintaining data consistency in a system consisting of a series of independently clocked and independently powered components.

While the Netmon central clock takes care of providing time for all components, thus eliminating one source of timestamp inconsistency, the open loop approach to logging makes it impossible for the cloud to know whether the "compressor on" log entry indicates the actual operation of the compressor.

And, therefore, a production monitoring solution needs to correlate temperature change with compressor operations, which is what ColdStores' "fridge performance charts" are meant for.

Another difficulty is the variety in the topology of the network. There might be only one unit at a site, or there might be thirty, the site might use wired peer-to-peer networking, wireless bridging, or modem backhaul.

The ability of the application to give an overview of multi-site statistics using a standard format but retaining the details of each individual sensor is important. The daily-statistics approach is thus an answer.

ColdStores shows how an established industrial control framework can evolve into a modern mobile monitoring application without sacrificing its real-time control capabilities.

The controller serves as the deterministic layer 8 inputs, 8 outputs, 10-bit analog sensing, thermistor calibration, and 2-second alarm network—while the Netmon ensures clocking, aggregation at the site level, and the data route to the cloud.

The ColdStores app brings the additional layer that people interact with multi-site visibility, 24-hour and 7-day trends, daily statistics, refrigeration performance charts, and daily alerts, all within a zero third-party sharing and data encryption in transit paradigm.

The MCU used in the particular case is not publicly known however, considering the electrical and timing constraints, it can be said that it belongs to the group of industrial MCUs which include multi-channel ADC, UART, and deterministic supervision firmware, a robust and long-life device choice for equipment intended to be in use for many decades.

In general, this is a lesson on architecture: separation of the tested control loop from the cloud and mobile layers lets each of them grow independently, and it is due to such architecture that a 40-year-old control platform becomes a good base for a remote monitoring system.

Live App: Google Play Store