BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20250309T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20251102T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20250407T145620Z UID:38E12532-82E3-4C94-8CED-C916ADADCC64 DTSTART;TZID=America/New_York:20250404T110000 DTEND;TZID=America/New_York:20250404T120000 DESCRIPTION:In this talk\, I will share our recent progress on developing l earning algorithms for real-world energy system control\, with stability a nd computational tractability guarantees. The first part is on reinforceme nt learning for power grid control. I will introduce a novel neural netwo rk architecture – monotone neural network (MNN) that ensure the network output is a monotone function of the input. MNN is achieved by first desig ning neural networks that are convex (with universal approximation guarant ee) and using gradients of convex functions to ensure monotonicity. We sho w that MNN is a powerful structure for voltage control – with stability and optimality guarantees compared to standard neural networks. The second part is about operator learning for building control. There is an emergen t need to model indoor air quality to improve occupant health and building energy efficiency. A fundamental challenge is that building airflow dynam ics are governed by nonlinear partial differential equations (PDEs) with u nknown parameters\, which are computationally prohibitive from a real‑ti me control perspective. I will introduce our work on PDE‑constrained opt imization for building model identification and designing neural operator learning for efficient PDE system control.\n\nSpeaker(s): \, Yuanyuan Shi\ n\n371 Fairfield Way\, Ite 336\, Storrs\, Connecticut\, United States\, 06 269-0001\, Virtual: https://events.vtools.ieee.org/m/478817 LOCATION:371 Fairfield Way\, Ite 336\, Storrs\, Connecticut\, United States \, 06269-0001\, Virtual: https://events.vtools.ieee.org/m/478817 ORGANIZER:junbo@uconn.edu SEQUENCE:13 SUMMARY:Learning for Power Grid and Building Control URL;VALUE=URI:https://events.vtools.ieee.org/m/478817 X-ALT-DESC:Description: <br /><p class="MsoNormal" style="mso-layout-grid-a lign: none\; text-autospace: none\;"><span style="font-family: 'Calibri'\, sans-serif\; color: black\; mso-themecolor: text1\;">In this talk\, I will share our recent progress on developing learning algorithms for real-worl d energy system control\, with stability and computational tractability gu arantees.<span style="mso-spacerun: yes\;">&nbsp\; </span></span><span sty le="font-family: 'Calibri'\,sans-serif\; color: black\; mso-themecolor: te xt1\;">The first part is on reinforcement learning for power grid control. <span style="mso-spacerun: yes\;">&nbsp\;</span>I will introduce a novel neural network architecture &ndash\; monotone neural network (MNN) that en sure the network output is a monotone function of the input. MNN is achiev ed by first designing neural networks that are convex (with universal appr oximation guarantee) and using gradients of convex functions to ensure mon otonicity. We show that MNN is a powerful structure for voltage control &n dash\; with stability and optimality guarantees compared to standard neura l networks. </span><span style="font-size: 12.0pt\; font-family: 'Calibri' \,sans-serif\; mso-fareast-font-family: DengXian\; mso-fareast-theme-font: minor-fareast\; color: black\; mso-themecolor: text1\; mso-font-kerning: 0pt\; mso-ligatures: none\; mso-ansi-language: EN-US\; mso-fareast-languag e: ZH-CN\; mso-bidi-language: AR-SA\;">The second part is about operator l earning for building control. There is an emergent need to model indoor ai r quality to improve occupant health and building energy efficiency. A fun damental challenge is that building airflow dynamics are governed by nonli near partial differential equations (PDEs) with unknown parameters\, which are computationally prohibitive from a real</span><span style="font-size: 12.0pt\; font-family: 'Cambria Math'\,serif\; mso-fareast-font-family: De ngXian\; mso-fareast-theme-font: minor-fareast\; mso-bidi-font-family: 'Ca mbria Math'\; color: black\; mso-themecolor: text1\; mso-font-kerning: 0pt \; mso-ligatures: none\; mso-ansi-language: EN-US\; mso-fareast-language: ZH-CN\; mso-bidi-language: AR-SA\;">‑</span><span style="font-size: 12.0 pt\; font-family: 'Calibri'\,sans-serif\; mso-fareast-font-family: DengXia n\; mso-fareast-theme-font: minor-fareast\; color: black\; mso-themecolor: text1\; mso-font-kerning: 0pt\; mso-ligatures: none\; mso-ansi-language: EN-US\; mso-fareast-language: ZH-CN\; mso-bidi-language: AR-SA\;">time con trol perspective. I will introduce our work on PDE</span><span style="font -size: 12.0pt\; font-family: 'Cambria Math'\,serif\; mso-fareast-font-fami ly: DengXian\; mso-fareast-theme-font: minor-fareast\; mso-bidi-font-famil y: 'Cambria Math'\; color: black\; mso-themecolor: text1\; mso-font-kernin g: 0pt\; mso-ligatures: none\; mso-ansi-language: EN-US\; mso-fareast-lang uage: ZH-CN\; mso-bidi-language: AR-SA\;">‑</span><span style="font-size : 12.0pt\; font-family: 'Calibri'\,sans-serif\; mso-fareast-font-family: D engXian\; mso-fareast-theme-font: minor-fareast\; color: black\; mso-theme color: text1\; mso-font-kerning: 0pt\; mso-ligatures: none\; mso-ansi-lang uage: EN-US\; mso-fareast-language: ZH-CN\; mso-bidi-language: AR-SA\;">co nstrained optimization for building model identification and designing neu ral operator learning for efficient PDE system control. </span></p> END:VEVENT END:VCALENDAR