Optimization in Smart Grids: Challenges and opportunities
Transcript of Optimization in Smart Grids: Challenges and opportunities
Optimization in Smart Grids: Challenges and opportunities
Prof. El-Ghazali TALBI
Polytech’Lille - University Lille & INRIA
2© El-Ghazali TALBI, INRIA Lille – Nord Europe 2
Mastertitelformat bearbeiten
Power Grids
Traditional grids
Smart grids
Microgrids
Optimization principles
Generation optimization
Design & Placement
Unit commitment
Transmission optimization
Network design
Energy market
Pricing
Distribution optimization
Power flow
Consumer optimization
Demand Management
Outline
3© El-Ghazali TALBI, INRIA Lille – Nord Europe 3
Mastertitelformat bearbeiten
Designed 130 years ago
First coal plant: Thomas Edison,1882
Generation
Electrical power is centrally generated
at large power plants
Transmission
Grid: Large transmission network
Distribution
Consumption is distributed over a large
geographical area
Energy demand will triple by 2050
Deregulation/liberalization market
Power loss : 6% in US, worse for
other countries
USA: estimated $25 billion per year
Environment impact
Greenhouse Gas (GHG) emissions:
contribution of 34%
Traditional grid
4© El-Ghazali TALBI, INRIA Lille – Nord Europe 4
Mastertitelformat bearbeiten
Electricity difficult to store
Advances in battery design
Electrical vehicles (EVs)
Flexibility: Energy production is
very hard to change quickly
Most of the flexibility is provided by
fossil fuel power stations
Energy demand fluctuates widely
during the day/seasons/weather
Electricity generation must match
consumer demand every minute
(power flow equations)
Peak load versus off-peak load
Low utilization of the grid during
off-peak times
Volatility in prices
Challenges and constraints
gasification
fired
5© El-Ghazali TALBI, INRIA Lille – Nord Europe 5
Mastertitelformat bearbeitenPower demand over a day
6© El-Ghazali TALBI, INRIA Lille – Nord Europe 6
Mastertitelformat bearbeitenFrom Traditional Grid to Smart grids
Distributed heterogenous
efficient reliable generation
Plants are distributed almost the
same way the consumers are
Minimal transmission of power to
distant consumers
Two-way information flow
Real-time demand, …
Two-way power flow
Smart meters: usage data
Flexible controllable load &
generation
Renewable energy
Energy storage
Plug in electric vehicles
Smart appliances
Customers can respond to price
signals sent from the utility
7© El-Ghazali TALBI, INRIA Lille – Nord Europe 7
Mastertitelformat bearbeiten
“A Microgrid is a group of interconnected loads and distributed energy
resources within clearly defined electrical boundaries that acts as a single
controllable entity with respect to the grid. A microgrid can connect and
disconnect from the grid to enable it to operate in both grid-connected or island
mode.” Microgrid Exchange Group, October 2010
Microgrids are low voltage intelligent distribution networks comprising various
distributed generators, storage devices and controllable loads which can be
operated as interconnected or as islanded system
Microgrid
8© El-Ghazali TALBI, INRIA Lille – Nord Europe 8
Mastertitelformat bearbeitenOptimization issues
9© El-Ghazali TALBI, INRIA Lille – Nord Europe 9
Mastertitelformat bearbeiten
Exact versus Approximation Algo.
Mathematical programming:
Linear programming,
Mixed integer programming
Relaxation (Lagrangian, SDP,…)
Dynamic programming, ADP, Monte
Carlo search
Artificial intelligence:
Constraint programming
Metaheuristics
Single-solution based algorithms:
local search, tabu search, …
Population based algorithms:
evolutionary algorithms, particle
swarm, …
Optimization methods
10© El-Ghazali TALBI, INRIA Lille – Nord Europe 10
Mastertitelformat bearbeiten
Multi-period optimization
Multiple time periods (time-
horizon)
Uncertainty management
Robust optimization, stochastic
optimization
Ex: wind and solar production
(weather), demand, prices, …
Optimization + Simulation
Meta-modeling, Surrogates
High performance computing
Optimization + Machine learning
Forecasting (demand, renewable
generation)
Neural networks, deep learning,
ensemble methods
Optimization principles
11© El-Ghazali TALBI, INRIA Lille – Nord Europe 11
Mastertitelformat bearbeiten
Economics
Keeping downward prices on electricity prices,
reducing the amount paid by consumers
Efficiency Reducing the cost to produce, deliver, and
consume electricity
Reliability
Reducing the cost of interruptions and power
quality disturbances
Reducing the probability and consequences of
widespread blackouts
Security
Reducing dependence on imported energy as
well as the probability and consequences of
manmade attacks and natural disasters.
Environmental friendliness
Reducing emissions by enabling a larger
penetration of renewables and improving
efficiency of generation, delivery,
consumption.
Multi-Objective optimization
12© El-Ghazali TALBI, INRIA Lille – Nord Europe 12
Mastertitelformat bearbeiten
Generation optimization
Unit commitment
Economic dispatch
Transmission optimization
Distribution optimization
Pricing and markets
Dynamic pricing
Customer management
Demand response
management
Optimization challenges in Smart Grids
NIST Model
13© El-Ghazali TALBI, INRIA Lille – Nord Europe 13
Mastertitelformat bearbeitenGeneration optimization
14© El-Ghazali TALBI, INRIA Lille – Nord Europe 14
Mastertitelformat bearbeiten
Nuclear
Shut down only for refueling
Ex: Scheduling maintenance of nuclear pants
«Zero» cost renewable sources
Hydroelectric
Solar (photovoltaic, combustion)
Wind
Biomass
Geothermal
Gas < 200 Mw, some hours (8-24) to start
Coal Long start time (days), +250 MW
Generation sources
15© El-Ghazali TALBI, INRIA Lille – Nord Europe 15
Mastertitelformat bearbeiten
Lifecycle greenhouse gas emissions by electricity source
Technology Description50th percentile
(g CO2/kWhe)
Hydroelectric reservoir 4
Wind onshore 12
Nuclearvarious generation II
reactor types16
Biomass various 18
Solar thermal parabolic trough 22
Geothermal hot dry rock 45
Solar PV Polycrystaline silicon 46
Natural gasvarious combined cycle
turbines without scrubbing469
Coalvarious generator types
without scrubbing1001
Grenhouse gas emissions / sources
p.s: sometimes eletrical vehicles are not environment friendly (ex. Singapour & Tesla)
16© El-Ghazali TALBI, INRIA Lille – Nord Europe 16
Mastertitelformat bearbeitenCapital versus operating cost – Ramping (flexibility)Technology Capital Cost ($/kW) Operating Cost ($/kWh)
Coal-fired combustion turbine $500 — $1,000 0.20 — 0.04
Natural gas combustion turbine $400 — $800 0.04 — 0.10
Coal gasification combined-cycle
(IGCC)
$1,000 — $1,500 0.04 — 0.08
Natural gas combined-cycle $600 — $1,200 0.04 — 0.10
Wnd turbine (includes offshore
wind)
$1,200 — $5,000 Less than 0.01
Nuclear $1,200 — $5,000 0.02 — 0.05
Photovoltaic Solar $4,500 and up Less than 0.01
Hydroelectric $1,200 — $5,000 Less than 0.01
Technology Ramping Time Min. Run Time
Simple-cycle cumbustion turbine minutes to hours minutes
Combined-cycle cumbustion turbine hours hours to days
Nuclear days weeks to months
Wind Turbine (includes offshore
wind)
minutes none
Hydroeletric (includes pumped
storage)
minutes none
17© El-Ghazali TALBI, INRIA Lille – Nord Europe 17
Mastertitelformat bearbeitenCentralized / distributed generation
Distributed generation
More independance
More flexibility on the source
Less electricity loss
Local generation
More suitable local sources: wind, solar, biomass, geothermal
More reliability
All your eggs aren’t in one basket
Brings a lot of problems (balancing, intermittent, uncertainty) !!
18© El-Ghazali TALBI, INRIA Lille – Nord Europe 18
Mastertitelformat bearbeiten
Engineering design
problems
Wind turbines
Solar systems, …
Placement problems
Placement of generators
Placement of storing
devices ...
Maintenance
Predictive maintenance
scheduling
Nuclear sites [Dupin,
Talbi 2016]
Generation optimization
19© El-Ghazali TALBI, INRIA Lille – Nord Europe 19
Mastertitelformat bearbeiten
Given
A set of geographical areas
A set of units
Find
A set of assignment of units to geographical areas
Objective
Cost
Power loss
Capacity (production)
Constraints
Unit specific
Examples
Placement of wind turbines
Optimal placement of units
20© El-Ghazali TALBI, INRIA Lille – Nord Europe 20
Mastertitelformat bearbeiten
Given
Land topology and leases
Wind resource variation
Operating and maintenance costs
Road and cabling costs
Find
Placement of wind turbines
Objective
Max global revenue, Max production
Constraints
Noise and environmental factors
Ensuring safety area (size turbines blades)
Inter-turbine interferences
Literature solutions: Metaheuristics CMA-ES [Wagner 2011]
Evolutionary algorithms [Xu 2011]
Particle Swarm Optimization [Chowdhury2012]
Placement of wind turbines
21© El-Ghazali TALBI, INRIA Lille – Nord Europe 21
Mastertitelformat bearbeiten
Given
Time horizon: short, term, long term
Generation sources
Demand prediction
Determine
For each t : power to produce
From which source
Objective
Min Cost, Min Risk, Max Sustainability, …
Constraints
Demand: Meet the demand
Ramp constraints (no fast change)
Power limits
Min Up/Down time constraints
Reserve, …
Literature solutions
Lagrangian relaxation, Mixed Integer integer
programming, Stochastic and robust opt.
Dynamic programming, metaheuristics, Contraint
programming, matheuristics [Talbi et al. 2015]
Unit Committment Problem (UCP) / Economic dispatch
22© El-Ghazali TALBI, INRIA Lille – Nord Europe 22
Mastertitelformat bearbeitenTransmission Optimization
23© El-Ghazali TALBI, INRIA Lille – Nord Europe 23
Mastertitelformat bearbeiten
Transmission and distribution losses
Losses in transmission between sources of supply and points of
distribution and in the distribution to consumers
6% in USA, 6% in France, 17% in Brazil, 22% in India, http://data.worldbank.org/indicator/EG.ELC.LOSS.ZS
Transmission optimization
24© El-Ghazali TALBI, INRIA Lille – Nord Europe 24
Mastertitelformat bearbeiten
Network design problems
Connection of new end users (factory, …)
Interconnection with other countries
Planning network evolution
Transmission and communication networks expansion
planning
Lines, substations, …
Presence of uncertainty
Load, …
Monte-Carlo simulations, …
Objectives: investment cost, reliability, power losses
Design problems
25© El-Ghazali TALBI, INRIA Lille – Nord Europe 25
Mastertitelformat bearbeitenEnergy markets
Creates new markets attracting consumers and innovations
26© El-Ghazali TALBI, INRIA Lille – Nord Europe 26
Mastertitelformat bearbeiten
Traditional retail view
Customer choice and control
Customer decide when and how
much to consume
Fully hedged service
Flat, uniform price
Tiered (inclining or declining rates)
Time of use (TOU) rates
(peak load pricing)
Pricing : Competitive markets
Wholesale view (load as resource)
Customer bid to supply
Day ahead market
4 hour ahead
Real-time market (5mn ahead)
Reliability products
Capacity
Emergency
Ancillary services
27© El-Ghazali TALBI, INRIA Lille – Nord Europe 27
Mastertitelformat bearbeiten
Flat pricing
Fixed price for all periods
Time-of-use (TOU) rates - Peak-load pricing
The intended cycle is divided into several periods
Distinct price value for each period is announced at the beginning of the
operation
Offer very low rates to customers who can shift high-demand operations
away from the times of day when the utility receives its peak demand for
energy.
The utility benefits from a more consistent daily load pattern, the customer
pays less
Pricing tiers (i.e. off-peak, peak) are established to correspond to specific
time intervals. Utilities can publish these rates to consumers to provide
financial incentives to shift demand to off-peak hours and reduce overall
consumption.
Time-shifting consumption - Real-time pricing (RTP)
Some utilities now offer their major customers real-time pricing
The exact price value for each period is calculated in real-time and is
announced only at the beginning of each operation period
Pricing strategies in whole sale market
28© El-Ghazali TALBI, INRIA Lille – Nord Europe 28
Mastertitelformat bearbeiten
Interaction between smart meters and
energy providers
Given
Subscribers preferences (utility functions)
Energy consumption patterns
Energy provider(s) single, multiple
Find
Real-time prices, day-ahead prices, …
Objectives
Min imposed energy cost,
Max the agregate utility to all subscribers,…
Constraints
Power generation capacity
Literature solutions
Bi-level optimization/Game theory [Sezin
2013]
Interior point methods [Samadi 2012]
Dynamic programming [Joe-Wang 2012]
Greedy heuristic algorithm [Wang 2012]
Multi-agent systems [Ramachandran 2010]
Optimal real-time pricing for revenue management
29© El-Ghazali TALBI, INRIA Lille – Nord Europe 29
Mastertitelformat bearbeitenOptimal real-time pricing for revenue management
Bi-level model [Brotcorne 2017]
Decision process involving two
decision-makers with a
hierarchical structure
Two decision levels : a leader
and a follower, controlling their
decision variables, seeking to
optimize their objective function.
The leader sets his decision
variables first. Then the follower
reacts based on the choices of
the leader.
Bilevel programming is strongly
NP-hard.
Originated from Stackelberg
games, related to principal-agent
problem, mathematical programs
with equilibrium constraints.
prices reaction
30© El-Ghazali TALBI, INRIA Lille – Nord Europe 30
Mastertitelformat bearbeitenDistribution optimization
31© El-Ghazali TALBI, INRIA Lille – Nord Europe 31
Mastertitelformat bearbeiten Grid connected micro-grids: High variability & uncertainty, distributed
generation, storage, pricing
Given
A distribution network, storage, prices
Find
Power flow, voltage, storage (multi-periodic)
Objective
Min Energy Losses, Imported energy, Carbon emissions, Max profit
Constraints
Bounds: powers, pressure, voltage, frequency, generation, load
Power flow equations, generation/load balance, storing capacities/losses
Literature solutions
Traditional gradient based solvers (such as Newton-Raphson), are inadequate in time
domain & cannot be applied to the combined network-storage problem (non-convex)
Semi-definite programming (SDP) relaxation [Dall’Anese 2014]
Dynamic programming [Levron 2017]
Particle swarm optim. [Srtomme 2009]
Optimal distributed power flow problem
32© El-Ghazali TALBI, INRIA Lille – Nord Europe 32
Mastertitelformat bearbeitenConsumer: Demand management optimization
33© El-Ghazali TALBI, INRIA Lille – Nord Europe 33
Mastertitelformat bearbeiten
Reduce electricity consumption in
homes, offices, and factories by
continually monitoring electricity
consumption and actively managing
how appliances consume energy.
It consists of demand-response
programs, smart meters and
variable electricity pricing, smart
buildings with smart appliances,
and energy dashboards.
Combined, these innovations allow
utility companies and consumers to
manage and respond to the
variances in electricity demand
more effectively
Demand management
34© El-Ghazali TALBI, INRIA Lille – Nord Europe 34
Mastertitelformat bearbeiten
Currently Flat price for electricity regardless of
the time of day or actual demand.
Smart meters give customers the ability to
choose variable-rate pricing based on the time of
day.
Consumers can respond accordingly by shifting
their energy consumption from high-price to low-
price periods.
Joint benefit of reducing costs for typical
consumers while lowering demand peaks for
utility companies.
Shifting
Re-scheduling operations so that some activities
take place during off-peak times of the day
Shedding
Shutting off non-essential equipment during the
peak period
Load shifting (shedding) with smart meters
35© El-Ghazali TALBI, INRIA Lille – Nord Europe 35
Mastertitelformat bearbeiten
Given
Local production (wind, solar, …)
Local load
Energy tariff
Find
Home devices scheduling
Energy plan
Buy/sold/store
Objectives
Min Bill, Max Profit
Min Peak demand, Max Confort, …
Constraints
Production
Batterie constraints
Home devices scheduling
Time windows
Literature solutions
ILP (Integer Linear Programming) [A. Barbato 2012]
Mixed Linear Integer Programming [D. Zhang 2011]
Multi-objective evolutionary algorithms [Z. Garroussi,
E-G. Talbi & R. Ellaia, 2016]
House energy demand response
36© El-Ghazali TALBI, INRIA Lille – Nord Europe 36
Mastertitelformat bearbeiten
Large scale problems
Huge number of generators, clients, big data
USA: 12M distributed generators, 3M miles lines, …
Multi-objective problems
reliability, availability, efficiency, sustainability, cost.
Dynamic optimization
Sensing and real-time measurements
Multi-periodic planning and optimization
Distribution networks evolution (different scenarios)
Optimization under uncertainty
Stochastic data
Optimization Challenges & Perspectives
37© El-Ghazali TALBI, INRIA Lille – Nord Europe 37
Mastertitelformat bearbeiten
Bi-level optimization
Dynamic pricing
Optimization & Simulation
Meta-modeling, Surrogates
High performance computing,
Parallel algorithms
GPU, Multi-cores, Cluster,
Heterogeneous computing
Optimization & Machine learning
Forecasting (demand, renewable
generation)
Short-term, medium-term, long-
term
Neural networks, deep learning,
…
Optimization Challenges & Perspectives
GPU
Upper-level / Leader
Lower-level / Follower
GPU
38© El-Ghazali TALBI, INRIA Lille – Nord Europe 38
Mastertitelformat bearbeiten
Smart grids and Logistic/Transportation systems
Electrical vehicles (charging systems)
Electric vehicle routing problem
[J. Serrar, E-G. Talbi, R. Ellaia, 2017]
Electrial buses
Smart grids and Cloud computing systems
Green data centers
Energy-aware Job Scheduling
Cooling
Smart grids and Smart city
Urban configuration
Smart green building
Electrical vehicles as storing devices
More flexibility
Smart Grid Application perspectives
39© El-Ghazali TALBI, INRIA Lille – Nord Europe 39
Mastertitelformat bearbeiten
https://bioma2018.sciencesconf.org
https://meta2018.sciencesconf.org