The SmartAppi project seeks to provide two of Ireland’s largest milk processors, Glanbia plc and Dairygold Ltd., with a reliable means of predicting fluctuation and trends in their raw material supply chains. It is an 18 month project, 65% funded by EI (IP/2014/0474) with the two milk processors providing the balance of the funding.
SmartAppi brings together basic research on regression modelling by Cork Institute of Technology (CIT), applied research on parameters effecting milk production in the national herd by Teagasc at Moorepark and advanced data mining and data visualization techniques currently being researched and developed by TSSG.
Data for the project is provided by Glanbia, Dairygold, ICBF, Met Eireann and Teagasc. By applying advanced regression modelling techniques, sophisticated new data mining algorithms and state of the art data visualization techniques the three research partners, CIt, Teagasc and TSSG will provide both commercial partners with an innovative decision support system to predict supply chain fluctuations and thus provide help with strategic business decisions such as capital expenditure timing and operational decisions such as the size of the raw material collection fleet.
SmartAppi system concept evolved from discussions between TSSG and Teagasc following the publication of the food strategy document, “Food Harvest 2020” by the Department of Agriculture, Food & the Marine. This strategy paper gave the concept immediate context as part of an overall plan to promote “smart agri” solutions for Irish industries based on Irish agriculture
The proposition to the dairy processors was timely as they formulated plans to address the government’s ambitious target of a 50% increase in milk output by 2020, combined with the cessation of the milk quota systems in 2015. They recognised that they needed to accurately forecast and manage the fluctuations in milk production in order to optimise their processing infrastructures to meet these targets.
With the cessation of the milk quota system in 2015, producers and processors in the Irish dairy supply chain are challenged with a new business paradigm based on the global milk price and a free market approach to production and pricing. As the producers strive to maximise output, a key issue for processors is how to reliably plan transport, refrigeration, processing and finished product storage capacity. This is a particularly challenging (and costly) problem for the processors as currently throughout Ireland there is a +/- 10% variance year on year of milk supply into the system with short term (weekly) fluctuations throughout the year.
A further challenge arises for the processors in seeking to increase profitability from the supply chain by directing milk with differing qualitative properties to the optimum processing facility. If production of milk with these qualitative characteristics could be accurately predicted in a reliable mathematical model, then milk with the relevant macro nutrients could be directed efficiently to the most suitable processing facility and thus optimise its profitability.
The problem for the processors and the opportunity for the proposed platform is the prediction of quantitative and qualitative milk production levels so as to enhance the dairy processors profitability in the face of global competition, while at the same time ensuring that there is adequate processing capabilities for the increased milk supply as well as ensuring that the investments made in additional processing capacity are appropriate and timely. Where this global competition is becoming most challenging is in the securing of large scale, long term contracts by Irish dairy processors in emerging markets, such as the Chinese market for powdered infant formula and powdered milk. Planning for the increased supply will ensure that the maximum returns can be achieved through the development of markets will ensure the maximum returns can be achieved for the dairy industry in Ireland.
The SMART APPI platform envisioned in this project will provide a means of increasing the processors ability to predict, and therefore manage, supply chain fluctuations in the short (10 days), medium (3 months) and long term (3 years). Figure 01 below gives a high level conceptual overview of the platform providing an insight into its overall purpose.
This platform will provide the processors with the means to reliably foresee variations in milk supply over the production season, both in relation to quantitative production levels, volume of milk, and the qualitative nature of the milk, fat protein and solids content. To achieve these objectives the platform will encompass a backend system hosted in a suitable data centre with high speed access provided to client interfaces. The client interfaces will serve two distinct user types, managers of the system and users of the system. While the final design of these interfaces is a task to be undertaken in the project, a browser based interface could provide some hardware and operating system independence.
Two types of user will interact with the system and to cater efficiently for both, two distinct client interfaces will be developed. One will provide users of the system with requested results, i.e. predictions regarding supply variability and “what if” scenario development, while the other will provide managers of the system with the necessary tools to control, optimise and maintain the system.
The system users interface will provide users with a dual functionality interface, firstly enabling them to view predictions for short, medium and long term qualitative and quantitative milk production and secondly providing a means to run “what if” scenarios to test the impact of possible parameter changes on production levels. It will, for instance allow a user to evaluate the impact of a 15% fall in grass growth rates on milk production and the subsequent upstream fluctuation in the supply chain.
The system managers’ interface will provide system control facilities. With these the appointed system managers will have the ability to control the system and carry out the day to day maintenance tasks such as adding new users, monitoring system performance and maintaining system integrity. The management client will also provide the means to evaluate the accuracy of the system and modify the underlying assumptions encapsulated in the regression models utilised.
Using new models developed by CIT and Teagasc, a software platform will be developed by TSSG utilising a state of the art software architecture to data mine multiple data sources and predict fluctuations in the milk supply chain so as to provide milk processors Glanbia and Dairygold with a competitive advantage in the altered Irish 2015 dairy sector.
The use of mathematical models to predict variations in milk supply is well established in the dairy sector. These models are usually deployed on spreadsheet based systems and require semi-manual capture of inputs from various data sources. These approaches have given some useful results, but have proven to be labour intensive, prone to error and sometimes too unresponsive to changing conditions to enable timely forecasting. With the imminent change in circumstances for Ireland’s dairy sector with the disbanding of the milk quota system, these shortcomings in the models must be overcome.
The SmartAppi platform architecture will provide the following novel improvements to dairy supply chain prediction:
- Support for new regression models unconstrained in their development by the potential of existing computer software processing capabilities.
- Software architecture capable of accepting inputs from multiple data sources in both batch mode and real-time.
- Software architecture capable of accepting data from multiple schema types and of processing this data without the need to convert it to a unified schema.
- An architecture capable of handling new data sources with the minimum of adjustment. Specifically it is anticipated that when European Space Agency completes the Earth Observation Satellite constellation and it becomes fully operational with the data available regularly from the Irish Centre for High End Computing (ICHEC), this data stream can be incorporated into the architecture without any difficulty.
- With the ability to handle multiple data sources in real-time and to accept new data sources when these become available, the system is extendable and adaptable to a greater extent than any other currently in use.
- A user friendly interface that will enable the systems full capabilities to be utilised by non-specialist personnel with both processors.