Progress 10/01/19 to 09/30/20
Outputs
Target Audience:Computer and communications industry with a presentation to technical staff and managers at Microsoft: "Smart Farm Irrigation Field Deployment". Virtual meeting. November 2020. Plant biology research community with a presentation to School of Integrative Plant Sciences: "Learning to communicate with plants - The case of water status". Virtual meeting, Cornell U. September 2020. Process systems engineering community with a presentation and conference abstract: Chen, W.H., Shang, C., Zhu, S., Haldeman, K., Santiago, M., Stroock, A.D. and You, F., 2020, July. Theoretical Exploration of Irrigation Control for Stem Water Potential through Model Predictive Control. In 2020 American Control Conference (ACC) (pp. 1992-1997). IEEE. Changes/Problems:Restrictions associated with COVID-19 led to a loss of half of the growing season for our experiments. What opportunities for training and professional development has the project provided?The project provided a rich context for training a cross-disciplinary group of students and researchers in 2020, with participation from chemical engineers, electrical engineers, and plant biologists. Based in part on their experience with embedded systems (internet-of-things), two students involved in the project were hired into industry (General Electric and Sony) to pursue these technical specialties professionally. How have the results been disseminated to communities of interest?One peer reviewed conference abstract: Chen, W.H., Shang, C., Zhu, S., Haldeman, K., Santiago, M., Stroock, A.D. and You, F., 2020, July. Theoretical Exploration of Irrigation Control for Stem Water Potential through Model Predictive Control. In 2020 American Control Conference (ACC) (pp. 1992-1997). IEEE. What do you plan to do during the next reporting period to accomplish the goals?In the coming period, the team will extend these studies in two ways: first, to the impact of water stress dynamics and irrigation management on fruit growth in a collaboration with Prof. Terence Robinson. This work will add dendrometer-based sensors of fruit growth to our sensing framework at two sites: in the Cornell Orchard with potted apple and at Cornell Argitech in field trees with commercial irrigation systems. Second, we will use our matured models and sensing and control system to perform our first live test of various control and optimization strategies.
Impacts
What was accomplished under these goals?
Obj. 1: The team matured its hardware and software for the collection of data and control of irrigation based on a cloud interface. We also optimized the installation of soil sensors to provide new additional constraints on our Soil-Plant-Atmosphere-Continuum models. Obj. 2: In a restricted period in the field due to constraints imposed by COVID, the team repeated a series of irrigation treatments on four potted apple trees in the Cornell Orchard. Importantly, these measurements included continuous measurements of water potential in the soil alongside those acquired in the stems. This additional data provided critical new information to inform our SPAC models: first, it allowed us to show that our first-generation models, while predictive over the range of conditions studied, did not represent the physics of the soil and soil-root interface in an accurate manner. In other words, our first models were, "right for the wrong reasons". This discovery has guided us to a new hypothesis about the dominant resistance to water uptake under water-limited conditions: rather than the limiting resistance being from bulk soil to the root (as conventionally assumed in the literature), our data suggests that it is in the root tissue itself. In pursuit of this hypothesis, we have reformulated our model to test forms of the root "vulnerability curve" (resistance as a function of water potential). This analysis is in progress and may provide important new perspectives on the water stress physiology of apple roots, a basis for a more accurate and robust model of water dynamics in apple, and a basis for interpreting and modeling water relations in a variety of woody fruit and nut perennials. Obj. 3: In this period, we finalized our numerical study of the use of various control strategies in the context of irrigation management based on continuous measurements of stem water potential (SWP) and SPAC models of SWP dynamics. We developed a complete mathematical framework for the use of SPAC-type models within two types of advanced control: data-driven robust model predictive control (DDRMPC) and certainty equivalence MPC (CEMPC). We used pre-existing data sets and preliminary SPAC models for almond to show that DDRMPC can provide water savings relative to CEMPC or on-off control in a case study based on weather data for a typical growing season in the Central Valley of CA. This work was presented and published as an abstract as a part of the IEEE 2020 American Control Conference. Obj. 4: The maturation of our hardware and software puts us in a position to implement our approaches in on different sites in 2021. We are planning for an implementation at in an Orchard at Cornell Argitech in Geneva, NY in 2021.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2020
Citation:
Chen, W.H., Shang, C., Zhu, S., Haldeman, K., Santiago, M., Stroock, A.D. and You, F., 2020, July. Theoretical Exploration of Irrigation Control for Stem Water Potential through Model Predictive Control. In 2020 American Control Conference (ACC) (pp. 1992-1997). IEEE.
|
Progress 10/01/18 to 09/30/19
Outputs
Target Audience:Process and Systems Engineering research community with a publication on Model Predictive Control for the optimization of crop irrigation: Shang, C., Chen, W.H., Stroock, A.D. and You, F., 2019. Robust model predictive control of irrigation systems with active uncertainty learning and data analytics. IEEE Transactions on Control Systems Technology, 28(4), pp.1493-1504. General public and scientific audience with a presentation at the annual meeting of the American Association for the Advancement of Science (AAAS): "Sustainably Feeding 10B People: Transforming Agriculture and Food Systems", Washington DC, April 2019 Agricultural scientists, growers, and industries with a presentation at the Annual Workshop of the Cornell Initiative for Digital Agriculture. "A new foundation for digital irrigation scheduling in apple orchards". Ithaca, NY. October 31, 2019. Agricultural scientists with a presentation at a meeting with Zhejiang Agricultural University. "Sensing for Ag Management and Discovery". Ithaca, NY. July 14, 2019. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The project has continued to provide a rich training ground for a diverse group of students from engineering and physical sciences device engineering (custom sensors), the electrical engineering and computer science of an "Internet of Things" system, and the plant physiology and agronomy of water management. How have the results been disseminated to communities of interest?One peer reviewed publication: Shang, C., Chen, W.H., Stroock, A.D. and You, F., 2019. Robust model predictive control of irrigation systems with active uncertainty learning and data analytics.IEEE Transactions on Control Systems Technology,28(4), pp.1493-1504. Stroock made four presentations on this project to audiences of engineers, agricultural scientists, and the broader public. These presentations were made at the following venues: Annual meeting of the American Association for the Advancement of Science (AAAS) Annual Workshop of the Cornell Initiative for Digital Agriculture. Department of Chemical Engineering, City College of New York Department of Chemical Engineering, Carnegie Mellon University Empire State Producers Expo Gordon Research Conference on the Physics and Chemistry of Microfluidics Department of Civil Engineering, Massachusetts Institute of Technology Department of Chemical Engineering, New York University What do you plan to do during the next reporting period to accomplish the goals?In the comping period, we aim to refine our hardware and software and collect additional data across drought stress scenarios in order to refine our models. A particular priority will be the collection of high quality soil data to constrain our SPAC model further with the aim of distinguishing between various hypotheses about the origins of variable properties (resistances and capacitances) in defining the responses in the water-limited regime.
Impacts
What was accomplished under these goals?
Obj. 1: The team developed an interface for the sensors and electronic irrigation valves with connectivity to the cloud via a cellular modem. This the cloud captured and displayed real-time data streams from multiple trees as well as micro-meteorological parameters and allow for remote actuation of the irrigation values. Obj. 2: Using the cloud interface, the team ran a series of irrigation scenarios aimed at elucidating the dynamics of stress in well-watered and deficit irrigated cases. Under these conditions, we also performed an initial exploration of the impacts of different timing of irrigation. These data sets provided new insights and the creation of a predictive model covering both well-watered and deficit regimes. An important insight relates to the distinctions between the dynamics and appropriate process models for well-watered and water-stressed scenarios: in the well-water case, the soil-plant-atmosphere-continuum (SPAC) behaves as a single compartment with fixed capacitance (C) and resistance (R) and a single characteristic response time (R*C). We associate these properties with the conductive tissue of the tree; the properties of the soil are not important as the soil water potential never deviates significantly from its saturated value (i.e., zero). In contrast, under water-deficit conditions, the dynamics cannot be captured by a single compartment. Rather, we find that we must add at least one additional compartment with variable resistance and capacitance. By using conventional models (e.g., van Genuchten) of water retention and conductance to parameterize the resistance and capacitance of the soil-root compartment, we can accurately predict the observed dynamics of stem water potential. We note though, that the parameters required for the soil to achieve this agreement are distinct from those expected for the sand-rich mix used in the pots. This observation leaves open questions: if the model is appropriate, what processes have lead to the changes in the effective soil properties relative to those expected; alternatively, might the model be predictive but physically wrong. Future experiments will aim at distinguishing these cases and further refining and testing our models. Obj. 3: Our first-generation model of irrigation optimization based on MPC use a rudimentary water balance model on the soil to develop methods focused on the particularities of this application of process control strategies (e.g., the precipitation and irrigation can only be additive). This work was published (Shang et al., IEEE Transactions on Control Systems Technology, 2019). In this period, we have adapted this approach to work with stem water (SWP) potential and using data collected with a microtensiometer before the grant period in almond and an initial SPAC model of SWP dynamics. This effort developed appropriate linearized forms of the more complex governing equations (coupled ordinary differential equations) involved in SPAC models and allowed for an initial comparison of several variants of MPC relative to simple "on-off" control of irrigation. Obj. 4: The hardware and code developed for the on-site sensing network, communication to the cloud, and cloud-based processing of the data streams and remote actuation of irrigation all provide a basis for implementation of our approaches in different sites at modest cost and time. We have not yet engaged a commercial partner for a demonstration at this time.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Shang, C., Chen, W.H., Stroock, A.D. and You, F., 2019. Robust model predictive control of irrigation systems with active uncertainty learning and data analytics. IEEE Transactions on Control Systems Technology, 28(4), pp.1493-1504.
|
Progress 12/07/17 to 09/30/18
Outputs
Target Audience:4 technical presentations were made to academic colleagues in engineering and horticulture. 2 presentations to mixed audiences of academics and stakeholders from a broad array of agricultural industries. Personal interactions and communications occurred with stakeholders from the specialty crop industry, including apple growers from New York State and Washington State. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The project's focus to date have focused on engineering questions and involved engineering students and one post-doc. This effort has provided these chemical engineers with unusual training in plant physiology, crop modeling, and environmental dynamics. The specialization of the participants involved has been microengineering, process engineering, and process control. The post-doc started a faculty position at Tsinghua University at the end of 2018. How have the results been disseminated to communities of interest?Stroock made four presentations to physical science and engineering audiences on this and related topics. These presentations were made at the following venues: University of Washington - Department of Bioengineering seminar American Chemical Society Annual Meeting American Institute of Chemical Engineering Annual Meeting NSF Workshop on autonomous nanosystems - University of Chicago Alan Lakso presented an invited talk "Water Use, Water Stress and Water Management in Fruit Trees" to the International Fruit Tree Association meeting in Rochester, NY February 2019. Alan Lakso was awarded the International Fruit Tree Association Applied Research Award for 2019 at the International Fruit Tree Association meeting in Rochester, NY February 2019. Alan Lakso has been invited to present a keynote talk "Plant-based sensing for irrigation management in the field" at the 9th International Symposium on Irrigation of Horticultural Crops in Matera Italy, June 2019. Alan Lakso will present a contributed talk "A microtensiometer sensor to continuously monitor stem water status in woody plants - design and field testing" at the 9th International Symposium on Irrigation of Horticultural Crops in Matera Italy, June 2019 with co-authors Siyu Zhu, Michael Santiago, Ken Shackel Vadim Volkov and Abraham D. Stroock. What do you plan to do during the next reporting period to accomplish the goals?In the coming period, the team will develop the hardware and software to implement the MPC control strategy on individual potted apple trees. This prototype will provide an opportunity to refine our models and methods and prepare for larger scale, in-field demonstrations in the final year of the proposal.
Impacts
What was accomplished under these goals?
Objective 1:The team has made important improvements in the embedding strategy of the microtensiometer within the conductive tissue of the plant. Most significantly, the use of a nanoparticulate paste at the interface between the sensor and the exposed xylem tissue has proven effective at maintaining fluid and thermal contact with the sensor. This method has minimized artifacts caused due to diurnal variations in temperature. Objective 2:Working in potted apple, the team has used the readings of the microtensiometer to develop a robust model of the soil-plant-atmosphere continuum (SPAC). This SPAC model provides a basis for the implementation of Model-Predictive Control (MPC) within our irrigation strategy. Objective 3:The team has developed the theoretical foundation for the implementation of MPC within the irrigation context. This component of the project involved new mathematical developments to accommodate the character of the weather data (in particular precipitation) as a robust strategy that accounts for uncertainty. A first study based on these methods has been accepted for publication (preprint: https://arxiv.org/abs/1810.05947). Objective 4:No progress to date.
Publications
- Type:
Journal Articles
Status:
Awaiting Publication
Year Published:
2018
Citation:
Shang, C., Chen, W.H., Stroock, A.D. and You, F., 2018. Robust Model Predictive Control of Irrigation Systems with Active Uncertainty Learning and Data Analytics. arXiv preprint arXiv:1810.05947.
|
|