Preventive Power System Operation During Hurricanes
Mostafa Ardakani and Ge [email protected]
Power Component Failure Estimation
Preventive Power System Operation
Outage w/o PPSO
Outage with PPSO
Hurricane Forecasting Module
Hurricane Impacts on PowerSystems
FERC Software Conference 2018
• Damage level: Low
• Main cause: Flooding
• Wind: Rarely an issue
• Damage level: High
• Main cause: Wind force
• Flooding: May aggravate the situation
• Damage level: High
• Main cause: Wind force
• Flooding: May aggravate the situation
Power Outage Statistics
• Hurricane season of 2017:
FERC Software Conference 2018
Harvey Irma MariaAugust September September
• 300,000 customer outages in Texas
• 6 Million customer outages in FL (59%)• ~1 Million customer outages in GA (22%)
• 100% customer outage in PR
Employment of Weather Data
• Would integration of weather data in power system operation reduce the size and duration of power outages?
• Availability of weather data:– System operators have access to weather forecast services– In some cases, they also have access to meteorologists onsite
• Existing technologies:
FERC Software Conference 2018
Power System Operation
Weather Forecast Data
– Pre-storm outage forecast– Post–storm restoration
planning
– Long-term grid hardening– Emergency operation based on
engineering judgement
Proposed Integrated Platform• Systematic integration of weather forecast data in power system
operation– Translation of weather data into useful information for operation:
component damage probability
FERC Software Conference 2018
Power Component Failure Estimation
Preventive Power System Operation
Outage w/o PPSO
Outage with PPSO
Hurricane Forecasting Module
Weather Forecasting(Atmospheric Sciences)
• High-resolution wind field modeling– 1 Km horizontal
• Hurricane track and movement speed estimation• Ensemble forecasting– Multiple tracks with different probabilities
• Forecast at different time scales– 5-day ahead, 48-hr ahead, day-ahead, hour-ahead
FERC Software Conference 2018
Component Failure Estimation
• Vulnerable components:– Transmission lines– Distribution lines
• We ignore distribution, because:– Distribution network is often radial, which makes preventive operation
ineffective– Distribution-level damage causes local power outage– Transmission-level damage can cause power outage in areas, not
directly affected by the hurricane• Transmission failure is estimated based on:
– The dynamic loading of the wind– Likely important factors that are neglected in our existing model:
• Debris modeling• Flooding and precipitation
FERC Software Conference 2018
Why Focus on Transmission?
• Power outage in the areas, not in the hurricane track, is due to transmission-level damage.
• Such outages may be manageable, through weather-aware preventive operation.
• Transmission line outages in the past:– Harvey: 97 lines (>139 kV)– Sandy: 218 lines (>115 kV)
FERC Software Conference 2018
Transmission Failure Estimation I
FERC Software Conference 2018
Top drift
Structural Drawings Finite Element Modeling
Stability under Dynamic Wind Loading
Tower drift1.5%, 2%, 2.5%, 3%
Transmission Failure Estimation II
FERC Software Conference 2018
• Finite element (FEM) models can be computationally demanding
• FEM will be used to develop fragility curves– Probabilistic description of failure likelihood, based on the
wind speed
200 random wind speed samples
Finite Element Model
Line Outage Scenario Generation
FERC Software Conference 2018
00
0
1
Line 1
Line 3Line 2
...
Line N
t0 tH tH+1 . . .tH+2 t0+TF
---------------------------------------------------------11
1
...
11
1000
0 or 1
0 or 10 or 1
0 or 1
The total number of scenarios can be calculated as!" = (%& + 1)*+,
the probability for each scenario can be calculated as-" =.
/01
*+,-/,34.3035
3461(1 − -/,3) .
Preventive Power System Operation
• Stochastic optimization– Scenarios: unplanned line outages – Load shedding is penalized by a large penalty factor
• Computational needs:– Stochastic optimization can be computationally demanding– Scenario reduction can help reduce the computational
burden:• Elimination of unlikely scenarios, below a threshold• Elimination of inconsequential scenarios
• The solution will change the dispatch to minimize load shedding
FERC Software Conference 2018
Case Study Setup
• Test system: IEEE 118-bus
• Two layouts:– Areas affected by
the hurricane• Two synthesized
hurricanes:– Harvey– Irma
FERC Software Conference 2018
68
84
69
83
75
85
96
82
78
9777
9880
99
65
81
70
116
Bus
Transmission Line
18
17
32
31
91030
826
23
25
22
21
20
113
114115
27 2829
Layout 1 Layout 2
4AM 7AM 10AM 1PM 4PM 7PM 10PM 1AM 4AM40
60
80
100
Win
d Sp
eed
(mile
/h) Hurricane Harvey
5PM 8PM 11PM 2AM 5AM 8AM 11AM 2PM 5PM110
115
120
125
130
Win
d Sp
eed
(mile
/h) Hurricane Irma
Results: Reliability Improvement
FERC Software Conference 2018
0%2%4%6%8%
10%12%14%16%
Expected Lost Load Expected Over-Generation
Expected Lost Load Expected Over-Generation
Business as Usual Preventive Operation
Harvey I Harvey II Irma I Irma II
Results: Reliability versus Cost
FERC Software Conference 2018
0%
20%
40%
60%
80%
100%
Harvey I Harvey II Irma I Irma II
Cost Increase Reliability Improvement
Conclusions
• Predictable weather-related natural hazards are the cause of about half of the blackouts in the US.
• Weather forecast data can be used to estimate component damage likelihood.
• Component damage estimations can be used to guidepreventive operation.
• The simulation results confirms the effectiveness of our integrated platform in substantially reducing power outages.
• Appropriate integration of weather forecast data within power system operation can enhance system reliability.
FERC Software Conference 2018
Discussion and Future Work
• Stochastic optimization was used in this work:– Computationally demanding– Power system operation software by in large use
deterministic models– We are currently working to develop proxy
deterministic rules that:• Capture the majority of stochastic optimization• Do not substantially add to the computational burden
• The framework is general and can be applied to other weather hazards such as ice storms.
FERC Software Conference 2018
Acknowledgement
FERC Software Conference 2018
Atmospheric Sciences
Civil Engineering
Electrical Engineering
Our Research Team
Principal
Investigators
Dr. Zhaoxia Pu Dr. Ge Ou Dr. Mostafa Ardakani
Graduate
Students
Jiayue Xue, Yuanrui Sang,
Farshad Mohammadi
Funding
Agency
References and Further Reading
FERC Software Conference 2018
[email protected] You!
• Y. Sang, J. Xue, M. Sahraei-Ardakani, and G. Ou, “Effective Scenario Selection for Preventive Stochastic Unit Commitment during Hurricanes,” 2018 IEEE International Conference on Probabilistic Methods Applied to Power Systems (PMAPS), Boise, ID, USA.
• M. Sahraei-Ardakani and Ge Ou, “Day-Ahead Preventive Scheduling of Power Systems During Natural Hazards via Stochastic Optimization,” IEEE PES General Meeting 2017, Chicago, IL, USA.
• Y. Sang, J. Xue, M. Sahraei-Ardakani, and G. Ou, “Reducing Hurricane-induced Power Outages through Preventive Operation,” working paper, available at: https://ardakani.ece.utah.edu/wp-content/uploads/sites/75/2018/05/HurricanePaper.pdf