Cow-centric data made available in real-time for model development Matthijs Vonder (TNO) Gerben Donker (Rovecom) [email protected] [email protected] 1 Contents 1. Introduction Smart Dairy Farming (SDF) 2. Large Scale Sensor Data Logistics for model development 3. Impression of SDF-Demonstrator 4. Wrap up 5. Next steps 6. Questions and discussion 2 2. Introduction Smart Dairy Farming • Collaboration project 3 Cooperations 7 SME’s 5 Research institutes 7 Real farmers • Timeline: 2011 – 2014 • Nothern part of the Netherlands • Website (in Dutch): http://www.smartdairyfarming.nl/nl/ aantallen NL noemen • Goal of SDF: to support dairy farmers in the care of individual animals. with the specific goal of a longer productive stay at the farm due to improvement of individual health. • Challenge: when successful: make it possible for the whole sector Numbers for the Dutch situation: • 18000+ farmers • in total more then 1.5 million milk cows • 20 to 200+ datafields per cow • many different stakeholders in the chain 3 Support dairy farmers in the care of individual animals • SDF provides support on 3 specific theme’s Young cattle Transition period Fertility -> see previous presentation higher growth as calf results in • Support is given with work instructions (aka SOP’s) higher production later based on sensor data 1e lact • e.g. weight, multiple times per day • e.g. within 5 min. after measuring for a specific cow • use unique life numbers 2000 effect on production [kg] in (near) real time 1e-3e lact 1000 0 700 800 900 1000 1100 -1000 Soberon et al., 2012 -2000 a specific instruction is given growth per day [gram/day] • e.g. “give this calf more milk, because it does not grow according to ideal curve” 4 SDF - starting points • The farmer is in control (in Dutch: “boer aan het roer”) he is owner of the cow related (sensor) data and decides if he carries out the work instruction (or not) • The cow is key: “cow centric data” model the cow and not the measuring device, e.g. • ask the specific cow C for the weight W at time T • versus: ask all (!) scales S1-Sn if they know the weight W for cow C at time T (sensor) data stays connected to the cow • also when the cow changes farm • Result is an open platform / solution essential for easy development of real-time models with open and generic interfaces where all farmers, all suppliers and all service providers could connect 5 Challenge: upscaling • Project 7 farmers with 100 – 350 cows each 20+ different sensor systems (with 200+ different datafields) 3 models some standards (on static data); almost no standards on sensor data upscaling • The Netherlands Requires good IT architecture and ITinfrastructure 18000+ farmers totally > 1.5 million milkcows xxx different sensor systems (yyy different fields) lots of potential new models (existing or new companies) from no standards to a lot of standards upscaling Our motto: “Think big, start small” • Other places in the world (milk)cows pigs/sheeps/horses/... 6 Current / traditional way (example) M1 M2 6 different connections sensor1 farmer 1 sensor2 sensor3 M3 4 different connections sensor4 sensorA sensorB each new model needs to make a lot of connections sensorC sensorD farmer 2 Note: Next to the physical connections also the permissions have to be arranged Per farmer, per sensor system for each model (nowadays : manually) 7 2. Large Scale Sensor Data Logistics for model development: InfoBroker M1 M2 M3 each new model connects to the InfoBroker InfoBroker sensor1 farmer 1 sensor2 sensor3 sensor4 sensorA sensorB sensorC sensorD farmer 2 8 InfoBroker concept • InfoBroker functionalities SOP Lore ipsum Open interfaces for data exchange (API) Authentication Permissions cow specific SOPs Model cow specific data which data may be used by whom to be set by the farmers Namingservice location where the data can be found – – InfoBroker cow specific data Cow centric Sensor data Cow centric static data cow-centric sensor data Integration Static data (e.g. feed) Static data (e.g. date of birth) who are you (are you allowed to login) combining info from different sources Pay-per-use fixed costs (connections) variable costs (used data) Sensor data • So: S S no central datastore for (sensor)data! but indeed a broker which re-uses available (standard) interfaces and reduces/prevents duplication 9 3. Impression of SDF-demonstrator Farmer: Dairy Campus Dashboard SOP-generation by models Speenschema aanpassen Krachtvoer aanpassen Ruwvoer aanpassen Pink insemineren Kalf/pink behandelen Voeradvies inwinnen Selecteren voor afvoer • Young cattle – SOP’s • Transition – SOP’s • Fertility – SOP’s Model Transition SOP-application SOPs Model Young cattle Sensor data logistics Datacollection on the farmer N.A. Model Fertility 40x InfoBroker Milk intake RFID 2x Weight 4x RFID 4x Water intake 4x Dairy Campus laptop q loggers 6x q Forster.csv pass trough q Realtime cow centric sensor data Match & Merge q milkintake.csv weight.csv waterintake.csv RFID -> lifenr sftp AnySense Static data CRV_Animalregistr. CRV_Diagnosys CRV_Treatment 3. Impression of SDF-demonstrator Farmer: Dairy Campus 2000 Soberon et al., 2012 1000 0 Dashboard SOP-generation by models Speenschema aanpassen Krachtvoer aanpassen Ruwvoer aanpassen Pink insemineren Kalf/pink behandelen Voeradvies inwinnen Selecteren voor afvoer • Young cattle – SOP’s • Transition – SOP’s • Fertility – SOP’s 700 800 900 1000 1100 -2000 Model Fertility Model Transition SOP-application SOPs Model Young cattle Sensor data logistics Datacollection on the farmer N.A. -1000 40x InfoBroker life_number,sensor,date_time,wcorr,wstable,wzero,wavg,wavgmin,wavgmax,werrors,wnousecounter,wnousetime,wusetime NL 916075572,dc_roostervloerhok1_weegschaal1,2013-11-04 04:18:35 UTC,129.0,129.0,0.0,102.3,30.8,129.0,0,0,470,18590 NL 916075572,dc_roostervloerhok1_weegschaal1,2013-11-04 04:18:40 UTC,129.5,129.5,0.0,129.6,129.5,130.0,0,0,475,18590 NL 916075572,dc_roostervloerhok1_weegschaal1,2013-11-04 04:18:45 UTC,130.0,130.0,0.0,129.5,129.0,130.0,0,0,480,18590 NL 916075572,dc_roostervloerhok1_weegschaal1,2013-11-04 04:18:50 UTC,130.0,130.0,0.0,129.8,129.5,130.0,0,0,485,18590 Milk intake RFID 2x Weight 4x RFID 4x Water intake 4x Dairy Campus laptop q Forster.csv Realtime cow Now a temporary route and storage in SDF 1 project centric sensor data • to make it possible with existing sensor systems milkintake.csv loggers 6x q pass trough q Match & Merge q weight.csv waterintake.csv In near future direct coupling with InfoBroker • • Static data storage at the sensor system/provider RFID -> lifenr and registration at the InfoBroker sftp AnySense CRV_Animalregistr. CRV_Diagnosys CRV_Treatment 4. Wrap up Cow-centric data made available in real-time for model development • Model development made easier: one–stop-place for all required data • model does not need to know the storage location(s) – multiple sources per farm (typically 10+) – multiple farms (up to 18000+ in the Netherlands) • for all kinds of data one open, generic and simple interface – no one-2-one connections – only the connection with the InfoBroker • model does not need to store the data sensor supplier agnostic – model using eg. weight and milkintake does not need to know (nor connect to) sensorsupplier • Open and scalable solution: all farmers, all sensor providers, all model-developers 12 5. Next steps • SDF 1.0: Proof of Concept (2011-2014) Finalize and test realtime models/work instructions Test the data retrieval via the InfoBroker Some extensions on the models/work instructions • SDF 2.0: Proof of Practice (project proposal) Improvement of existing models (better and more SOP’s) InfoBroker: new functionalities (e.g. permissions) Upscaling: • more farms • more sensor providers • more data consumers (next slide →) 13 Examples for using the InfoBroker Farmer Consumers / ... Benchmarking Advisor Generic app (graphs, avg) Realtime-Model (expert) Coöperation InfoBroker Static data Real time cow-centric data Large customer (Unilever etc) Your organisation? Contact: [email protected] [email protected] This project is made possible by: Thank you for your attention
© Copyright 2024 ExpyDoc