Page 1
ACCURATE & TIMELY INSIGHTS INTO VARIABLE RENEWABLE ENERGY – WIND & SOLAR FORECASTING & SCHEDULING IN
INDIA
Presented by
Mr Abhik Kumar Das Director, ([email protected] )
del2infinity Energy Consulting Private Limited
1www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 2
What is Forecast?
2
Forecast is a Prediction of a variable (value, vector or matrix)
considering other similar or dissimilar variable (s) and/or parameters
Page 3
•What•Why•When•Where•How
3
Page 4
What to Forecast……
– Supply side
• Wind Power generation
• Solar Power generation
–Demand side
• Load Forecast
4
Page 5
Why Solar, Wind ? Clean but Variable
5
20% to 40% Renewable energy is wasted due to variability
Page 6
Variability in Solar
6
Sun does not shine at night, and there are cloudy days
Fig: Variation in solar PV output on two different days in 2011 at Yelesandra
www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 7
Variability in Wind
7
There are days-long lulls in wind power
Fig: Variation in wind power output on four different days in 2011 for Karnataka
www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 8
Why to Forecast? Grid Stability
8Picture: http://www.news.gatech.edu/features/building-power-grid-future
Page 9
Breaking Network Stability?
9
Need some regulation
Smart Power Grid = Complex network
Page 10
10
Wind & Solar Power Forecasting Regulation
The Central Electricity Regulatory Commission (CERC), India hasfinalized the mechanism for Forecasting, Scheduling and DeviationSettlement of wind & solar projects at Inter-State level.
The CERC has issued the Indian Electricity Grid Code (ThirdAmendment) Regulations, 2015 (IEGC) and Deviation SettlementMechanism and related matters (Second Amendment), Regulations 2015(DSMR)respectively.
The mechanism is applicable from November 1, 2015.
www.del2infinity.xyz || [email protected]
Page 11
Wind & Solar Power Forecasting Regulation
Key features of the mechanism : The mechanism shall be applicable to Wind and Solar Generators.
Scheduling of Wind & Solar Generators have been made mandatory.
The maximum number of revisions has been increased from 8 to 16.
A new forecast error computation formula has been formulated, which is: =100*(Scheduled Generation-Actual Generation)/Available Capacity.
The penalties for deviation have been computed as per Power Purchase Agreements and shall be leviedfor deviation beyond +/-15%
11
Page 12
•What•Why•When•Where•How
12
Page 13
When to Forecast…….
13
Uses
Yearly• Resource Planning
• Contingency Analysis
Monthly
Weekly
Day Ahead • Scheduling
• Trading
1-6 hour ahead • Load following
• Commitment for next operating hour
1-2 hour ahead • Real time despatch decision
• Regulation
Page 14
Forecast Time Block
• 1 Data-point = 15 min Time-Block
• 1 Data = Energy Generation (kW-Hr) in 15 min
• 1 Data = Average Power X 0.25
14
Page 15
• A is a vector of size N
15
Forecast Horizon N
1 Hour 4
Day Ahead 4 X 24 = 96
Weekly 96 X 7 = 672
Monthly 96X30 = 2880
Page 16
Forecast Revision
T1 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0 R0
T2 A A A A R0 R0 R0 R0 R0 R0 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1
T3 A A A A A A A A A A A A R1 R1 R1 R1 R1 R1 R2 R2 R2 R2 R2 R2 R2 R2 R2 R2 R2 R2 R2
16
Page 17
Where to Forecast…….
• Turbine level or PV module/array level
• Plant level (same IPP)
• Plant(s) level ( different IPPs)
• Aggregate level
17
An approximate Tree structure. Forecast is possible at any Node
Page 18
18
Aggregation of forecast and Aggregate level forecast are different
Aggregated Forecast
Aggregation of Forecast
If Forecast:
Page 19
19
• Forecast Function is Non-Linear
• Above relation is Stable to Linear Transformation of Error
• Propagation of Uncertainty can create false precision
Aggregated Forecast is better than Aggregation of Forecast i.e.
Page 20
•What•Why•When•Where•How
20
Page 21
Forecast Error and Accuracy
21
Page 22
Forecast Error• Average Error
– MAE & Normalized MAE
– RMSE & Normalized RMSE
• Point Error (Time block wise error)
– Based on forecast value
– Based on available capacity
22www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 23
Some simple relation
23www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 24
Forecast Accuracy =
• Using MAE
• Using RMSE
• Using point error
24www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 25
Forecast Accuracy vs Penalty for New Regulation
25
Average Penalty per kw-Hr of Installed Capacity
Page 26
An example for Wind Forecast
26
For CERC Regulation considering PPA = INR 5.00/kW-Hr, the slab wise and total penalties are as follows for 27.4 MW Wind Plant
Page 27
An example for Solar Forecast
27
For CERC Regulation considering PPA = INR 5.00/kW-Hr, the slab wise and total penalties are as follows for 40.2 MW Solar Plant
Page 28
Energy Accuracy
28
Page 29
What is the acceptable value of m
29
Forecast Process:
Forecast Error:
No Deviation Charge if:
Page 31
31
Regulation m
CERC 0.15
FOR 0.10
Wind Power Forecast with revision in Wind
Page 32
What effects the accuracy?
• Model limitations
• Chaos
• Data & Data analysis uncertainties
32
Page 33
Uncertainty of Forecast
• Requirement of more variables or less variables
• Is uncertainty grows with data complexity or data complexity
reduces uncertainty
• Uncertainty of data availability and uncertainty of forecast
33
Page 34
Who is responsible for Low Accuracy & False Precision?
• Power generators if they do not share the value of Available Capacity
• Power generators if they do not share their correct Schedule
• Forecast & Scheduling Service providers if their accuracy is low and produce fake precision
Forecasting is computationally expensive, but if the Energy Accuracyis below a certain level (say 85%-90%), Power Generators may chargeF&S Service providers for Low Accuracy and False Precision
34
Page 35
•What•Why•When•Where•How
35
Page 36
How to Forecast…..
• Persistence method : “What you see is what you get”
• Using Numerical Weather Prediction to predict meteorological variables
• Physical approach
• Statistical approach
• Del2infinity’s Mixed Approach
36www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 37
37
del2infinity works in the domain of Energy Analytics
del2infinity serves AAAS (Analytics As A Service) to its different clients
An IT integrated and solution oriented approach for every energy analytics problems
del2infinity’s Wind & Solar Power Forecasting product is capable of doing 24 hours day ahead wind
power forecast with maximum 16 revisions
del2infinity’s forecast Integrator is capable of integrating maximum 7 parallel power forecast
About del2infinity
Page 38
del2infinity Solution for Wind & Solar Energy
Automatically delivers the wind and solar power forecasts via a customizable web-based/ FTP-based / Email-based platform
Proprietary algorithm based on statistical machine learning and pattern recognition
Parallel architecture to integrate other forecast solutions to reduce computation time &delay effects
Secured data storage & data transmission protocols (SSL encrypted)
38www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 39
39
• del2infinty believes : “Essentially, all models are wrong, but some are useful”
• del2infinity uses its proprietary useful F&S model(s) to forecast which
– Maximize the Energy Accuracy
– Minimize the Deviation Penalty
- Accepts fair percentage of Financial Responsibility (Client may charge Penalty on del2infinity’s
Service cost if forecast accuracy is not adequate)
Page 40
Forecasting Performance Analysis
40
Fig.: a) Average wind speed vs forecast wind speed b)Error margin of Wind speed vs Probability of error in forecast without revision (R0)
Page 41
Forecasting Accuracy (aggregated Wind Power)
41
Accuracy: Normalized Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) for different forecast horizons (hours)
www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 42
Forecast (R0) on 12-July in a Wind Plant at KA
42
Page 43
Actual & Forecast Power (R1) (40.2 MW Solar, 24 April, 2016)
43www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 44
Forecast (R0) on 05-July in a Solar Plant at Gujrat
44
0
5
10
15
20
25
1 4 7 1013161922252831343740434649525558616467707376798285889194
Actual
Forecast
0
5
10
15
20
25
30
1 4 7 1013161922252831343740434649525558616467707376798285889194
Error
Error
Page 45
Accuracy (40.2 MW Solar, 15 April 2016 – 30 April 2016)
45
Date Normalized RMSE %PPA No-Penalty %PPA No-Penalty %PPA No-Penalty
TNERC Probability(%) FOR Probability(%) CERC Probability (%)
15-04-2016 4.54 0.45 91.67 0.24 95.83 0.11 95.83
16-04-2016 4.32 0.38 78.13 0.04 93.75 0.00 100.00
17-04-2016 4.47 0.40 80.21 0.10 93.75 0.02 98.96
18-04-2016 3.55 0.21 87.50 0.08 98.96 0.04 98.96
19-04-2016 5.53 0.67 85.42 0.36 94.79 0.19 95.83
20-04-2016 2.22 0.03 95.83 0.00 100.00 0.00 100.00
21-04-2016 1.86 0.03 96.88 0.00 100.00 0.00 100.00
22-04-2016 2.50 0.05 91.67 0.00 100.00 0.00 100.00
23-04-2016 2.63 0.05 90.63 0.00 100.00 0.00 100.00
24-04-2016 1.21 0.00 98.96 0.00 100.00 0.00 100.00
25-04-2016 4.45 0.42 84.38 0.12 93.75 0.01 98.96
26-04-2016 1.08 0.00 98.96 0.00 100.00 0.00 100.00
27-04-2016 1.06 0.00 100.00 0.00 100.00 0.00 100.00
28-04-2016 2.53 0.04 95.83 0.01 98.96 0.00 100.00
29-04-2016 2.08 0.06 91.67 0.00 100.00 0.00 100.00
30-04-2016 2.08 0.06 91.67 0.00 100.00 0.00 100.00
www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 46
46
Forecast Accuracy in Wind (R12) & Solar (R0) Forecast (IPP) New Regulation
Absolute Error Margin
Probability (%)Wind
Probability (%)Solar
< 15% 93.36 +/- 5 98.69 +/- 2.5
15%-25% 4.37 +/- 5 1.27+/-2.5
25%-35% 1.16 +/- 5 0.04+/- 2.5
>35% 1.11 +/- 5 0
www.del2infinity.xyz || [email protected] June 2016
Page 47
Not only Power forecast
Analyse Variability
47
Page 48
1-Ramp Analysis Approach
48
Probability of power ramping up from 2040 MW in the time interval of interest?
Karnataka wind
Abhik Kumar Das et al., “An Empirical Model for Ramp Analysis of Utility-Scale Solar PV Power” Solar Energy, Elsevier, vol. 107, September 2014
•Gather deeper insights into power variability
Similar approach applied for solar PV power (kW-scale variability)
Abhik Kumar Das et al., “A Statistical Model for Wind Power on the Basis of Ramp Analysis,” International Journal of Green Energy, 2013
www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 49
2-Ratio Based Approach
49
Ensuring Grid Reliability:Renewable plant operators have to comply with grid code
µ is related to Ramp Limit
Dimensionless “Ratio Based” Model
AK Das, “An Analytical Model for Ratio Based Analysis of Wind Power Ramp Events,” Sustainable Energy Technology and Assessments, Elsevier vol. 9, pp.49-54, March 2015
Page 50
Variability Representation: Simplified
50
PV output < 60% of maximum power for 80% of plant-operation time
Abhik Kumar Das, “Quantifying Photovoltaic Power Variability using Lorenz Curve,” Journal of Renewable and Sustainable Energy, Journal of Renewable and Sustainable Energy, AIP, vol.6 (3), June 2014
One step ahead for wind:
System operators want simple, yet robust, insights Enables decision in fast paced environment
www.del2infinity.xyz || [email protected] © del2infinity Energy Consulting Private Limited
Page 51
• Massively ambitious targets for renewable power across the globe
• Variability is our enemy
Do Forecast, Analyse Variability
“If you know the enemy and know yourself, you need not fear the result of a hundred battles” – Sun Tzu, The Art of War
51
Page 52
Thank Youdel2infinity Energy Consulting Private Limited
[email protected]
52