Team 1709: Utility Grid Enhancement for Deep Integration of - PowerPoint PPT Presentation

Team 1709: Utility Grid Enhancement for Deep Integration of Distributed Photovoltaic Power Generation Sponsored by: The United Illuminating Company Alsandy Jacot (EE) Derek McCormack (EE) Rahul Vachhani (EE) Joel Velez (EE) Project Advisor: Dr. Peng Zhang

Overview Introduction Background Design Examples Conclusion

Introduction - Sponsor ● We are being sponsored by United Illuminating ● Southwest area of Connecticut ● Territory covers 17 towns ● 9 of which are along the shore

Introduction - Our Project ● Perform big data analytics to develop a PV capacity factor prediction tool Input Specific Algorithm based on Big Output expected Parameters Data Analytics values Time/Season Machine learning Capacity factor Size (kW) Output (kW) Efficiency Inv. Tilt (Deg) Azimuth (Deg)

Background - Photovoltaics and the Distribution System ● Excess energy from photovoltaics flow back into the power distribution system. Due to the intermittent and fast ramping nature of PVs they seldom reach their ● nameplate capacity value in daily operations. ● This can have negative effects on the system. Voltage issues ○ Overloading ○ ○ Protection schemes not working as designed

Background - Photovoltaics and the Distribution System

Background - Importance ● It is important for UI to have a good understanding of how much power is flowing back into the system during specific operating conditions so that they may: ○ Mitigate potentially negative effects of PV. ○ Make changes and upgrades to system without over engineering($$). Perform more accurate analysis of their system. ○

Background - Capacity Factor ● To understand how much power is flowing into the system we will be using capacity factors. Capacity Factor - is the predicted output of a PV system over a given time ● period compared to that of the nameplate rating of the PV system. Nameplate rating is typically much higher than the actual output. ○ Actual output will vary based on several variables. ○

Design - Data ● Perform big data analytics Determine which variables most affect PV output. ○ Determine number of residential sites and locations to analyze. ○ ○ Determine what times to analyze. ○ Determine which algorithm to best model our data. Obtain data from CT Greenbank ○ Use analytics to determine how each variable affects the output so a ○ weighting can be assigned to the variables.

Design - Data - Variables ● Decide which variables most affect PV output. ○ Irradiance ○ Tilt angle ○ Azimuth angle ○ Manufacturer of PV System ○ Time/ Season ○ Specifics of PV System (Size, quantity, etc…)

Design - Data - Sites ● Determine number of residential sites and locations to analyze. Decided on analyzing 75 sites split into 3 areas. ○ Data will be at 5 minutes intervals. ■ ○ Chose 25 sites each from Fairfield, Hamden, and Milford. ■ Chosen due to high concentration of PV sites. Further narrowed locations geographically based on relativity to shore. ■

Design - Data - Sites ● Determine number of residential sites and locations to analyze.

Design - Data - Time ● Determine what times and seasons to analyze. May 17, 2015 → May 23, 2015 10:00AM → 8:00PM ○ May 15, 2016 → May 21, 2016 10:00AM → 8:00PM ○ ■ High PV output with low power consumption. ○ July 26, 2015 → August 1, 2015 10:00AM → 8:00PM July 17, 2016 → July 23, 2016 10:00AM → 8:00PM ○ High PV output with high power consumption ■ ○ February 15, 2015 → February 21, 2015 10:00AM → 8:00PM ○ February 14, 2016 → February 20, 2016 10:00AM → 8:00PM Low PV output with low power consumption ■

Design - Data - Modeling ● Determine which algorithm to best model our data. ○ K-means clustering Popular, efficient, simple and intuitive in nature. ■ Determined to be effective for smaller sample sizes (75 sites) ■ ○ Random forest (decision trees) ■ Good for studying relationships among variables. Applicable to both regression and classification problems ■ SVM (Support vector machine) ○ It works really well with clear margin of separation ■ ■ Applicable to both regression and classification problems

Design - Prediction Tool ● Based on our analysis we intend to assign weights to our variables so that we can derive a tool that with given variables input can output a predicted capacity factor. Once we have a working tool we would like to implement it into a user friendly ● interface that UI can easily utilize. ○ Excel? Java? ○ Etc..? ○

Examples - SVM

Conclusion - Where we are now ● Currently have a request out to CT Greenbank for the required data. While we wait we will be doing more research on which machine learning ● algorithm will be best for our project and we will be doing more sample analysis in Matlab. Keep working on this semester's final report. ●

Conclusion - Where we intend to be ● By the end of this semester we intend on having all the needed data and plan on preparing it for analysis. By the end of the Fall semester we intend on having an easy to use tool for ● capacity factor prediction that UI can use for in regards to their system. If time permits we will also perform basic research and analysis on the ● economical impact of distributed energy resources. PV in particular.

Conclusion - Gantt

References 1. https://cleantechnica.com/solar-power/ 2. http://www.ipsi.net/commercial-power/cogeneration-systems 3. http://www.decodingsustainability.com/blog/2016/2/28/does-combined-heat-and-power-fit-into-the-future-low-carbon-world 4. https://www.civicsolar.com/support/installer/articles/microgrid-regulatory-policy-us 5. https://www.emaze.com/@AQWLRIT/Wind-Turbine 6. http://www.spheralsolar.com/ 7. http://www.nyiso.com/public/webdocs/media_room/publications_presentations/Other_Reports/Other_Reports/A_Review_of_Distributed_Energ y_Resources_September_2014.pdf 8. http://pvwatts.nrel.gov/index.php 9 https://maps.nrel.gov/nsrdb-viewer/#/?aL=8VWYIh%255Bv%255D%3Dt&bL=groad&cE=0&lR=0&mC=29.305561325527698%2C-84.63867 10. http://gizmodo.com/rooftop-solar-panels-are-almost-all-facing-the-wrong-di-1644518413 11. http://costofsolar.com/best-direction-to-face-solar-panels-south-or-west/ 12. http://energy.gov/eere/energybasics/articles/solar-radiation-basics 13. http://brightstarsolar.net/2014/02/common-sizes-of-solar-panels/ 14. http://www.weatherquestions.com/What_causes_the_seasons.htm 15. http://news.energysage.com/best-solar-panel-manufacturers-usa/ 16. https://www.wunderground.com/history

Questions?