Energy Systems Operational Optimisation. Emmanouil (Manolis) Loukarakis Pierluigi Mancarella

Transcription

1 Energy Systems Operational Optimisation Emmanouil (Manolis) Loukarakis Pierluigi Mancarella Workshop on Mathematics of Energy Management University of Leeds, 14 June 2016

2 Overview What s this presentation about? 1: Perspective 2: Electrical Distribution Networks Management 3: Other Problems 2/18 (Overview)

3 What s the Problem? Towards increased energy efficiency & reduced emissions Large-Scale Storage Distribution medium/low voltage radial networks Inflexible Demand Distributed Generation Large-Scale Renewables Conventional Generation CHP high voltage meshed networks Heating Flexible Demand Distributed Storage Resources Coordination Over Time Gas Network Heat Network Electrical Distribution Networks Management Optimising Gas (or other fuel) Usage Optimising Heat Networks Operation Hot Water / Other Processes 3/18 (Perspective)

4 Current State of Play Unit Commitment (every 24h to <1h) Network Simplified models Operating status Large Generators Detailed model / bids Optimise! Forecasts Distribution System Aggregate model/bids significant uncertainty integer variables typically coupled with reliability requirements Is this the right time to optimise devices at the end-user level? other energy vectors? Not really! Probably not in detail! Economic Dispatch (every 15min) Network Large Generators Distribution System limited number of discrete controls contingency considerations limited look-ahead Detailed models Detailed models Optimise! Local Device Controls (instant) Aggregate measurements Operating state / Control mode / Power set-points Is this the right time to optimise devices at the end-user level? other energy vectors? If not now when?! Network Large Generators real-time 4/18 (Perspective)

5 Distribution Extending Dispatch energy power Distribution Microgrid Large-scale generation (conventional & renewable) system (multiple areas) Bus Aggregate Demand IGs TSOs Users Area 1 Area Area Very large scale! Uncertainty! Peculiarities of individual devices. Need for one more optimisation step! Large-scale generation (conventional & renewable) system (multiple areas) infeasible infeasible Distribution system (high/medium voltage feeders) curtailment time-step curtailment time-step Distribution system (medium/low voltage feeders) Individual Users (inflexible & flexible demand / small scale renewables) 5/18 (Perspective)

6 A Step Further Unit Commitment (every 24h to <1h) Network Large Generators Forecasts Distribution System Simplified models Operating status Detailed model / bids Optimise! Aggregate model/bids Disaggregating the network operators schedule Economic Dispatch (every 15min) Large Generators Network Forecasts Distribution System Microgrids Microgrid (Local) Dispatch (every 1min) Microgrids Users Detailed model / bids OPF Aggregating function / OPF Aggregate models Network constraints Flexible & inflexible energy offers / requests Optimise! Operating state / Control mode / Power set-points Operating state / Control mode / Power set-points Optimise! Local Device Controls (instant) Network Large Generators Users 6/18 (Perspective)

7 Microgrid Dispatch or in other words: close-to-real-time distribution network management IEEE-123 the good old days IEEE-123 in a test case with lots of EVs if left uncontrolled Objectives follow a given power output (market signal) serve customers! alleviate constraints violations Requirements solution time up to a few minutes Controls many discrete: tap changers, capacitor banks, loads some continuous: smallscale generation, storage, some EVs 7/18 (Distribution Networks Management)

8 Modelling Considerations (part 1) Point 1 Return currents not of interest Kron s reduction! Point 2 Symmetrical components no advantage in 1p/2p loads Point 3 Constant power models not good enough go ZIP + VI formulation Non-linear! Non-convex! 8/18 (Distribution Networks Management)

9 Modelling Considerations (part 2) Point 4 If V in polar coordinates the energy balance (right part) is non-linear use rectangular coordinates! Point 5 Voltage constraints non-convex Still non-linear! 9/18 (Distribution Networks Management)

10 Modelling Considerations (part 3) imag{i} (p.u. I max ) real{i} (p.u. c P ) linear approximation feasibility region non-linear exact curve outer approximation Approximation 2 Approximate P-part, as a ZI-part voltage (p.u.) real{i} (p.u. I max ) Approximation 3 Imbalance / capacity bounds linearize! Linear (assuming Z part is fixed)! 10/18 (Distribution Networks Management)

11 Modelling Considerations (part 4) Formulation Multi-time-step? stochastic? A 632B 632C 632D Formulation Single-time-step? deterministic? Approximation 4 Modified utility function to prioritise demand Follow the market power reference 632E Point 6 Do we really need tight voltage bounds? 11/18 (Distribution Networks Management)

12 Does It Work? Algorithm A 632B 632C 632D 632E Collect info from smart meters Approximate problem at a given voltage reference frame YES Needs adjustment? NO Send energy schedules to devices IEEE IEEE IEEE IEEE Time (sec) 0.17 (0.25) 0.24 (0.50) 0.23 (0.43) 0.33 (1.80) 12/18 (Distribution Networks Management)

13 Tap-Changers Algorithm Approximation 5 Taps are continuous Collect info from smart meters Approximate demand & taps at a given voltage reference frame Solution time (sec) Iterations Max. tap rounding errors Power change IEEE % IEEE % IEEE % IEEE % Solve for state and taps (approximate) YES Needs adjustment? NO Send energy schedules to devices Update trust-region 13/18 (Distribution Networks Management)

14 Discrete Controls Algorithm 3 Mixed integer programming Approximation 5 Solve continuous relaxation restricting deviations from nearest integral solution Tap controls number Added EVs number Solution time (sec) IEEE IEEE IEEE IEEE An feasible integral solution was recovered Due to high number of small controls no significant difference between the continuous relaxation objective value Collect info from smart meters Approximate demand & taps at a given voltage reference frame Solve for state and taps (approx. continuous relaxation) YES Needs adjustment? NO YES Is integral? Send energy schedules to devices NO Update trust-region Adjust penalty 14/18 (Distribution Networks Management)

15 Summing Up months/ years ahead minutes / hours ahead min. ahead sec. ahead real-time Model detail Model detail Model detail Model detail Uncertainty Uncertainty Uncertainty Not now! There are more problems out there! How should we solve them? Important! Problem characteristics! Solver characteristics! 15/18 (Distribution Networks Management)

16 Another Problem : energy district management (1) The problem optimising over time subject to network constraints and detailed device and building models boiler CHP other gas demand HEAT / POWER GENERATION INSTALLATION heat exchanger pump BUILDING from gas supply network gas network heat network from / to electricity supply network electrical network OTHER BUILDINGS / INSTALLATIONS Computational difficulties thermal network storage capacity thermal network dynamics building heating hot water renewables storage other electrical demand 16/18 (Other Challenges)

17 Another Problem : energy district management (2) The Manchester University test case electricity gas Solution Fast enough? Reliable enough? heat 17/18 (Other Challenges)

18 Distribution marginal price (m.u./mwh) Another Problem : distributed optimisation applications Solving very large scale problems Getting closer to control <10-1 <10-2 <10-3 < iteration Optimization Problem Structure Power System Decomposition Large-scale generation (conventional & renewable) system (multiple areas) Bus Aggregate Demand IGs TSOs Users Area 1 Area Area 3 2 TSO 1 TSO 2 TSO 3 IG IG 1 Users 2 Users 4 Users 1 Users 3 IG IG 1 Users 5 Users 6 Agent/subproblem representing users Large Generator subproblem Network Operator subproblem 18/18 (Other Challenges)

19 Thank you for your attention Questions?