Abstract
India's agricultural system has been facing a severe problem of crop wastage. A key contributing factor to this problem is that many small farmers lack access to reliable cold storage that extends crop shelf-life. To avoid having leftover crops that spoil, these farmers often sell their crops at unfavorable low prices. Inevitably, not all crops are sold before spoilage. Even if the farmers have access to cold storage, the farmers may not know how long to hold different crops in cold storage for, which hinges on strategizing over when and where to sell their harvest. In this note, we present progress toward a simultaneous hardware and software solution that aims to help farmers reduce crop spoilage and increase their profits. The hardware is a cost-effective solar-powered refrigerator and control unit. The software refers to a produce price forecasting system, for which we have tested a number of machine learning methods. Note that unlike standard price forecasting tasks such as for stock market data, the produce price data from predominantly rural Indian markets have a large amount of missing values. In developing our two-pronged solution, we are actively working with farmers at two pilot sites in Karnataka and Odisha.
Introduction
In India, small and marginal farmers—those managing less than one hectare of land—represent approximately 81% of agricultural holdings. Despite their significant contribution to the food supply, these farmers face a $15 billion annual crop wastage problem. Perishable fruits and vegetables often spoil before they can be sold, primarily due to a lack of reliable cold storage infrastructure. This deficiency forces farmers into "crash selling," where they must accept whatever price middlemen offer to avoid total loss.
This paper proposes a two-pronged approach to address this crisis: a low-cost, solar-powered cold storage unit to bridge the infrastructure gap, and a machine learning-based price forecasting system to bridge the information gap. By providing both the means to preserve crops and the data to strategize sales, the solution aims to optimize market participation for small-scale farmers.
Hardware: The CoolCrop System
The hardware component centers on a custom-designed solar-powered cold storage unit. Traditional commercial off-the-shelf (COTS) units are often too energy-intensive and expensive for rural applications. To overcome this, the researchers developed the "CoolCrop Controller," a flexible management unit built around a Raspberry Pi 3.
The controller interfaces with a custom data acquisition board to manage cooling units, monitor temperature and humidity via Ethernet-connected sensors, and optimize power consumption from solar arrays. A single 5-metric-ton unit is designed to serve a cooperative of 40–50 small farmers, making the technology economically viable through shared use.
Software: Price Forecasting for Strategic Selling
Physical storage is only half the battle; farmers also need to know when and where to sell their preserved crops. The authors developed a price forecasting system using data from Agmarknet, an Indian government portal.
Predicting exact prices in rural markets is challenging due to extreme data sparsity—some markets are missing up to 81% of daily pricing information. To address this, the problem was framed as a classification task: predicting whether prices will move "up," "down," or "stay" the same over a specific forecast horizon. The models utilize "masks" to handle missing data and include seasonality features like the "day of the year."
Results and Real-World Impact
Benchmarking various models, including Support Vector Machines and LSTMs, the researchers found that Gradient Tree Boosting (GBoost) and AdaBoost offered the best performance. GBoost, for example, achieved a raw accuracy of 63.7% while successfully identifying 20.4% of "up" movements and 15.7% of "down" movements.
The real-world impact of this integrated solution has been significant. At a pilot site in Dandeli, Karnataka, the introduction of cold storage led to a 100% increase in vegetable production. Farmers, no longer capped by the fear of immediate spoilage, were able to expand their output and improve their livelihoods.
Conclusion and Future Directions
While the technology shows promise, significant hurdles remain. The capital expense of these units is high relative to the average farmer's annual income (~$600 USD), necessitating micro-financing or NGO-led deployment models. Future work will focus on improving forecast precision by incorporating variables like rainfall and market demand, and refining communication strategies (e.g., SMS alerts) to ensure that data-driven insights are accessible and actionable for rural farmers.
Summary updated