REAL TIME CONTROL OF ELECTRICAL GRIDS WITH EXPLICIT POWER SETPOINTS - PowerPoint PPT Presentation

REAL ‐ TIME CONTROL OF ELECTRICAL GRIDS WITH EXPLICIT POWER SETPOINTS Dagstuhl Seminar Feb 23 ‐ 27, 2015 Jean‐Yves Le Boudec Mario Paolone joint work with Dr. Andrey Bernstein and Lorenzo Reyes, EPFL Laboratory for Communications and Applications and Distributed Electrical Systems Laboratory

References [Commelec] Andrey Bernstein, Lorenzo Reyes‐Chamorro , Jean‐Yves Le Boudec , Mario Paolone, “A Composable Method for Real‐Time Control of Active Distribution Networks with Explicit Power Setpoints”, arXiv:1403.2407 (http://arxiv.org/abs/1403.2407) http://smartgrid.epfl.ch [Campus smart grid] M. Pignati et al ,“Real‐Time State Estimation of the EPFL‐ Campus Medium‐Voltage Grid by Using PMUs”, to appear at Innovative Smart Grid Technologies (ISGT2015) 2

1. Motivation: Real Time Control of Electrical Grids Electrical grids are controlled in real‐time to ensure Energy balance + Quality of Service Generators react to frequency variations (droop control) Issue: inertia‐less systems (DC/AC converters : wind mills, PVs) 3

The problem of inertia ‐ less systems at distribution level 2003 blackout in Italy frequency trend Source: UCTE Interim Report of the Investigation Committee on the 28 September 2003 Blackout in Italy 2 s 2009 blackout during the islanding maneuver of an active distribution network Source: A. Borghetti, C. A. Nucci, M. Paolone, G. Ciappi, A. Solari, “Synchronized Phasors Monitoring During the Islanding Maneuver of an Active Distribution Network”, IEEE Trans. On Smart Grid, vol. 2 , issue: 1, march, 2011, pp: 70 – 79. 4

Short ‐ term volatility of Solar PV Measured on EPFL Roof 5

Current methods for real time control of electrical grids do not work well with a high penetration of intermittent distributed generation (e.g. solar photovoltaics, combined heat and power) 6

The same PV peak does not always have the same effect… 10:05: consumption is small 14:40 10:05 No voltage surge 14:40 load Voltage surge PV power =180 kW PV power =160 kW absorbs the peak 7

Controling a Distribution Network or a Microgrid Traditional Explicit control + storage Storage can be real or Upgrade lines and virtual: storage, intelligent transformers buildings, e‐cars Fast ramping fuel generators Chandolon 8

The EPFL Commelec Project Grid Real Time control Intelligent Consume Consume Consume I would like to building 50kW 50kW 10kW consume [0 – 50 kw] Building Agent Consume Consume Produce 20 kW 20 kW 423 kW I can either Grid consume or Battery Agent Grid produce produce Battery Agent 9

Problems with Explicit Control inexpensive platforms (embedded controllers) scalability do not build a monster of complexity ‐ bug‐free Such a system must be scalable and composable (i.e. built with identical small elements) 10

2. COMMELEC’s Architecture Software Agents associated with devices load, generators, storage grids Grid agent sends explicit power setpoints to devices’ agents Leader and follower resource agent is follower or grid agent e.g. LV grid agent is follower of MV agent 11

The Commelec Protocol Every agent advertises its state (every ms) as PQt profile, virtual cost and belief function Grid agent computes optimal setpoints and sends setpoint requests to agents Communication is over D‐TLS and IPRP – details not discussed today 12

A Uniform, Simple Model Every resource agent exports ‐ constraints on active and reactive power setpoints (PQt profile) ‐ virtual cost ‐ belief function I can do �, � It costs you (virtually) ���, �� BA GA 13

Examples of PQt profiles 14

Examples of PQt profiles: the case of a battery  requirement: compute the internal limits that the battery must respect for the next time step.  HP: the state of charge SoC is fixed between two setpoints implementations (correct if Δt is small enough)  battery model is extremely simple dc  V t  t R t   V dc dc dc I t  I dc  V t dc  I t  t R t dc I t dc E t V t dc  R t I t E t  V t dc 15

Examples of PQt profiles: the case of a battery  limits on DC: V dc and I dc need to respect specific battery constraints V min ≤ V dc ≤ V max and I min ≤ I dc ≤ I max , therefore, the DC limits in power are:     V max E t  V max     I min dc  max dc  min P , E t  R t I min  , P Vdc , P Idc  P min max max max   R t    V min  2 4 R t , if E t 2    I max    E t 2 4 R t , if E t  2 R t   E t Vdc  Idc    , P   V min E t  V min P   I max otherwise max max E t  R t I max  otherwise   R t   16

Examples of PQt profiles: the case of a battery  limits on AC: the battery is interfaced with a power converter of rated power Sr and efficiency η ac   P  dc  P if P dc  0 min min  ac   P dc  P  max max ac  P dc    P if P dc  0 min min  ac  P  dc  P  max max 2  Q t 2  S r     ac ac P t 17

Virtual cost act as proxy for Internal Constraints If state of charge is 0.7, I am willing to inject power If state of charge is 0.3, I am interested in consuming power I can do �, � It costs you (virtually) ���, �� BA GA 18

Examples of Virtual Costs 19

Commelec Protocol: Belief Function Say grid agent requests setpoint ��� from a resource; ��� actual setpoint will, in general, differ. Belief function is exported by resource agent with the semantic: resource implements ��� ��� Essential for safe operation 20

Examples of Belief Function PQt profile = setpoints that this resource is willing to receive Belief function = actual operation points that may result from receiving a setpoint 21

Grid Agent’s job Leader agent (grid agent) computes setpoints for followers based on state estimation advertisements received requested setpoint from leader agent penalty function keeps voltages close to 1 p.u. Grid Agent attempts to minimize and currents within bounds � � � � � virtual cost of cost of power flow resource � at point � of common connection Grid Agent does not see the details of resources a grid is a collection of devices that export PQt profiles, virtual costs and belief functions and has some penalty function problem solved by grid agent is always the same 22

Grid Agent’s algorithm Given estimated (measured) state � computed next � setpoint is where is a vector opposed to gradient of overall objective Proj is the projection on the set of safe electrical states This is a randomized algorithm to minimize 23

Aggregation (Composability) A system, including its grid, can be abstracted as a I can do � � , � � single component It costs you (virtually) � � � � , � � given PQt profiles of S � , S � , � � solve load flow and compute possible � � , � � + overall cost � � � � , � � 24

Aggregation Example non controlled load microhydro battery boiler PV 25

Aggregated PQt profile safe approximation ( subset of true aggregated PQt profile) 26

Aggregated Belief safe approximation ( superset of true aggregated belief) 27

28 Commelec Traditional control

Simulations – Results Local power management Boiler WB2 charges at full power because PV3 produces Boiler WB2 starts because WB1 stops at mid power due to line congestion 29

Separation of Concerns Resource Agents Grid Agents Device dependent Complex and real time Simple: But: all identical translate internal state (soc) into virtual cost Implement setpoint received from a grid agent 30

Reliability and Security Grid Agent development uses Prof Sifakis’s rigorous system development approach and the BIP framework Grid Agent are triplicated, Resource Agents use voting Communication used authentication (D‐TLS) and real time reliability protocols 31

An Operating System for Electrical Grids Resource control uses the Commelec API (publicly available) and does not need to be aware of the grid Intelligent Building Application Commelec API Commelec Grid Agent E-car Commelec API Commelec API PVs 32

EPFL Campus Smart Grid smartgrid.epfl.ch 33

System architecture Sensors and Phasor Measurement Units ‐ Voltage and current sensors ‐ Class 0.1 ‐ Nodal voltages and injected currents ‐ Phasor Measurement Units ‐ GPS synchronization ‐ Synchrophasor estimation based on the enhanced IpDFT algorithm [1] running on a FPGA ‐ Encapsulation and streaming according to IEEE c37.118.2‐2011 34

Conclusion Commelec is a practical method for automatic control of a grid exploits available resources (storage, demand response) to avoid curtailing renewables while maintaining safe operation Method is designed to be robust and scalable separation of concerns between resource agents (simple, device specific) and grid agents (all identical) a simple, unified protocol that hides specifics of resources aggregation for scalability We have started to develop the method on EPFL campus to show grid autopilot 35