Summary Presentation: Systems Engineering Research in Energy Ding, Gautam, Gibson, Huang, Johnson, Moreno-Centeno April 17, 2013
Summary Presentation: Systems Engineering Research in Energy
Ding, Gautam, Gibson, Huang, Johnson, Moreno-Centeno
April 17, 2013
About usYu Ding,Professor of Industrial & Systems Eng. and of Electrical & Computer Engineering
Natarajan Gautam,Associate Professor of Industrial & Systems Eng. and of Electrical & Computer Engineering
Rick Gibson,Associate Professor of Geology & Geophysics
Jianhua Huang,Professor of Statistics
Andy Johnson,Associate Professor of Industrial & Systems Engineering
Erick Moreno-Centeno,Assistant Professor of Industrial & Systems Engineering
Systems Engineering Focus
Statistics &Data Analytics
Optimi-zationMethodologies
& Tool Sets
Stochasticprocesses
Economics & Game Theory
Geophysics Modeling
ProblemsAddressed
• System-level models and performance metrics;• Predictive models and uncertainty quantification;• Production, economics, social assessment;• Large-scale optimization for decision making.
Systems Engineering Impact
Wind turbine reliability& maintenance
Robust adaptivetopology control
Data center information & resource management
Data acquisition design for hydrocarbon exploration
Regulation for distribution costs and SO2 and NOx emissions
Yu Ding & Jianhua Huang• Expertise
– Machine learning and data analytics;– Big Data applications to energy systems;– Quality, reliability, and maintenance engineering.
[email protected] [email protected]
Wind speed (m/s)
Low production efficiency
High production efficiency
Pow
er o
utpu
t (K
W)
• Estimate the endogenous power curve:– A system-level performance metric for turbine performance assessment;– Enhance wind power prediction.
19612
10144
10128
10125
9614
9613
9612
9611
240
239
238
237236
235234233
232
231
230229
228
227
226
225
224
223
222
221
220219
218
217216215
214
213
212
211210
209
208207206205
204203202
201200199
198
197
196
195194193
192
191
190189188
187
186185
184
183
182181
180
179
178
177
176
175174
173
172
171170169168
167
166165
164 163
162
161160
159158
157156
155
154153
152151
150
149148
147
146145144
143142
141
140139
138
137136
135
134
133132
131130
129
128127126
125
124123
122
121
120
119118
117116
115
114
113
112
111110109
108107106
105
104103102
101
1009998
9796
95
9493
9291
908988
8786
8584
83828180
79
7877
76
75 7473
72
71 7069
68
67
6665
64
6362
61
605958575655
5453
52
5150
4948
47464544
434241
4039
38
3736
35
3433
32
31 30
2928
2726
25
24232221
20191817
1615
1413
12111098
765
4
321
Yu Ding & Jianhua Huang
• Multi-fidelity modeling of localized wind field:– Turbine anemometers: in-situ but not calibrated; more than 20% error.– Mast anemometers: calibrated but lack spatial resolution.– Impact: characterization of the wake effect, modeling of turbine
interactions, and enhancement in wind power prediction.
Met mastTurbine
16.5 miles
5.7
mile
s
Natarajan Gautam• Expertise
– Stochastic systems: analysis and control;– Queuing models;– Networks and optimization.
Cryogenic tunnel freezerDynamically adjust cryogen level based on sensed loadEnergy savings: 20-60%
• Applications
Data CentersWith L. NtaimoVirtualization, cluster sizing, voltage/freq. scalingIT energy savings: 25-50%
Underwater SensorsWith R. GibsonRouting and battery changeCost savings: 60%
Natarajan Gautam• Future work
sensors
actuators
Building
Building’s state
Control SystemOptimal
lighting level
Historical weather data + prediction
Controlling energy consumption in buildings
Stochastic models for hour-ahead solar power predictions
Tracking customer behavior in smart-metered systems
Ener
gy U
sage
Low
High
Normalized Time
SmartMeterinstalled
Applied for rebate
Ideal Customer
Oblivious Customer
Inconsistent Customer
Energy-conscious Customer
Richard Gibson• Expertise
– Modeling of seismic wave propagation;– Seismic reservoir characterization;– Seismic data acquisition design and optimization.
Seismic survey designQuantitative, model-basedmeasures of image quality;assess effects of knownsubsurface structure
• Applications
Numerical modelingWith Y. EfendievMultiscale finite elementmethods; simulations ofhighly heterogeneous media
VOI-Seismic AcquisitionWith E. Bickel
Quantify value of competingseismic acquisition tech-nologies
Richard Gibson• Cableless acquisition of seismic data is growing because of recent
technological advances:− Facilitates acquisition in areas with environmental concerns, difficult
terrain, or human populations − Challenges exist in developing systems with real-time access to data
Impact: high quality data in difficult environments
Real-time data access• drive-by or fly-by• rib and backbone architecture
(radio/fiber optic)• potential optimizations of
retrievalGeophone, power, storage
May bury instrumentsin case of wildlife!
Andrew Johnson• Expertise
– Regulation of • distribution prices;• SO2 and NOx generation in coal power plants
– Productivity and efficiency analysis.
Distribution Cost RegulationQuantitative, model-based measures of efficient cost; Control for operating environment
• Applications
Estimating marginal abatement costsConsider multiple pollutant and abatement processes to quantify the cost associated with lower pollution levels
Andrew JohnsonOur research group developed the model currently being used to regulate 86 regional distributors in Finland for the 2012-2015 regulation period
An efficient isocost surface for electricity distribution in Finland based on 2010 data
Map of distribution regions in Finland
Erick Moreno-Centeno• Expertise
– Computational optimization and combinatorial optimization;– Network flows;– Integer programming.
• Applications:− Optimization of the smart grid;− Efficient algorithms for (Big) Data mining.
• Corrective Topology controlNew
Paradigms
• Practical• Effective• Efficient
OptimizationAlgorithms
Load Shed
Demandfully met
Erick Moreno-Centeno• Practical, effective and efficient algorithm for corrective topology
control:− Given an N-2 contingency prevents significantly more load shed than
traditional control,− Finds near-optimal topology two orders of magnitude faster than best
known tool.− Impact: FASTER load recovery and MORE load shed prevented.
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
80.00
90.00
100.00
1 2 3 4 5 6
MIP_HMFDP (k)MDFP (Best)
% o
f Loa
d Sh
ed P
reve
nted
10 20 30 40 50 60Minutes from contingency
with Topology control (1 switch / 10 min)no Topology control (pure re-dispatch)Absolute best with no Topology Control
SE in Energy: Summary
− Modeling strengths: • System level; • Stochasticity and uncertainty; • Physical-natural-societal interactions;
− Solution techniques: • Large-scale optimization; • Machine learning and data mining;• Queuing networks; • Computational methods;
− Impacts: • Reliability of energy systems;• Efficiency and cost of energy production;• Timeliness and quality of decision making.