Arun K u m ar

06 .09 .2019

Technologies

Indian Forecasting E xperience & Issues 2nd International Conference 4th – 6th September

Agenda Steps involved in creating a high accuracy forecasting

Our forecasting Model

Key challenges we face while forecasting

States with high RE face huge challenge in balancing

Solar and Wind Project Profitability

Day ahead forecast - Wind, Solar & Demand curve (Utility)

Cross Border Power Trade

1. Data collection, sanitization and processing - The inputs used for forecasting are :

• Historical Demand/Generation Data

• Historical and Forecasted Weather data i.e. Temperature, Humidity, Wind speed & Weather Conditions

• Real time Demand/Generation data (in 15 mins block wise manner)

2. Trend Analysis using the data –Historical data used to study the trend of the Demand/Generation with respect to :

• Seasonality

• Weather impact

• Festival behavior and special days (if any)

V a riou s S te p s in volve d in c re a tin g a h ig h a c c u ra c y fore c a stin g -

3. Weather Analysis & its Procurement - Weather is procured through multiple weather sites for particular locations

(usually 2- 3 locations per state) according to following observation :

• High energy demand cities of particular state’s

• Nearest weather station’s of the plant

4. Algorithm Deployment - Artificial Intelligence (AI)/Machine learning based algorithm are used for Demand & RE

Forecast .

5. Result Analysis - The results are analyzed as per MAPE & RMSE to test the accuracy of all deployed models, by

machine learning and through manual observation .

S te p s to fore c a stin g ……………-

F ore c a stin g M od e l d e s ig n e d to d e live r h ig h a c c u ra c y

Saas bas e d m ode l A c hie ving a high le ve l of fore c as ting ac c urac y

Kreate F ore c as ting M od e l and A rc hite c ture

R E a nd loa d fore c a s ting A rc hite c ture Dual -ensemble method AI, Machine Learning and Statistics based algorithms for forecasting Multiple weather forecast services from leading International and Indian institutions Real -Time Analytic Dashboard for monitoring load/ generation data and grid penalties Current model is being tested for Norther Regional Grid with a load of 50GW and RE of 10GW and a central Indian state with load of 12GW ad RE of 5GW

1. Inaccuracy in weather Forecast : Causing high penalties in few time blocks of the day

• Instantaneous changes in weather : Weather prediction using statistical Numerical Weather Pridiction (NWP) models are

not able to capture instantaneous changes in weather which contributes high deviation in RE forecast during monsoon

and high wind season

• Wide weather variation across same state : Across any state, wide variation in weather forecast are noticed which may

cause high deviation in demand pattern

• Interpolation of weather forecast : As of now weather forecasts are available in hourly/three hourly forecast from

domestic where we get 12 hourly update and international 6- 12 hourly update weather service providers which is

downscaled in 15 mins . granularity to generate RE/Demand forecasts .

K e y C h a lle n g e s w e fa c e …

W e a th e r C h a lle n g e -

1. Unpredicted breakdowns due to

thunderstorms and rain :

• Whenever turbulent weather conditions

occur, the unaccounted breakdowns in

demand pattern occur .

• Lightning and thunderstorms induce

unwanted spikes/dips in the telemetry

(SCADA) recording mechanism . This may be

visualized in the image :

2. Ensuring reliable real time SCADA data is a

problem .

C h a lle n g e s

D a ta C h a lle n g e -

If S E M / m e te r d a ta a ls o re c e ive d on re a l- t im e the n high d e via tions ob s e rve d in fe w tim e - b loc k s c a n b e m inim iz e d

C h a lle n g e s fa c e d b y S ta te s w ith h ig h e r sh a re of R e n e w a b le E n e rg y (R E )

Pre Forecasting • No RE and load forecast

• High uncertainty in supply due to Wind and Solar generation

• Huge deviations in block - wise supply - demand balance

• Large spot purchases from exchange, generators and neighboring grids

• Dispatching (cost of energy) was becoming expensive • Coal plants are not backing down when excess RE • Excess RE generation leads to higher Grid penalty for DISCOM • Curtailment of RE generation

• Higher GHG emissions

• Regulation in place where there are penalties for deviation from schedule

• Several pilots by both generators and grid manager started

• Penalties imposed by regulation likely to force focus on high quality forecasting

• DISCOM pays Grid penalty for poor balancing —excess or under drawl from grid

• DISCOMs with higher RE share in energy are paying significantly higher penalties on demand deviation & so are the RE generators

• Data acquisition and quality of data remain a challenge for forecast accuracy

• Quality weather forecast will remain a challenge

Evolving forecasting regime

S ola r a n d W in d P roje c t P rofita b ility c a n b e Im p a c te d b y p e n a ltie s d u e to p oor fore c a st

0,00093 0 ,0 0 0 8 6

0 ,0 0 0 72

0 ,0 0 0 43

0 ,0 0 10 0

0 ,0 0 110

0 ,0 0 0 9 3

0 ,0 0 0 6 5

0,00000

0,00020

0,00040

0,00060

0,00080

0,00100

0,00120

Jan'19 Feb'19 Mar'19 Apr'19

Impact of Grid Penalty (US Cents / unit)

Solar

Wind

Cost of grid p en a lt ies : 2 to 3% of p ow er off-ta k e ra te

Typical day ahead forecast for W ind, Solar & dem and curve for a utility

0

1000

2000

3000

4000

5000

6000

0,00

2000,00

4000,00

6000,00

8000,00

10000,00

12000,00

00:0

0:00

00:3

0:00

01:0

0:00

01:3

0:00

02:0

0:00

02:3

0:00

03:0

0:00

03:3

0:00

04:0

0:00

04:3

0:00

05:0

0:00

05:3

0:00

06:0

0:00

06:3

0:00

07:0

0:00

07:3

0:00

08:0

0:00

08:3

0:00

09:0

0:00

09:3

0:00

10:0

0:00

10:3

0:00

11:0

0:00

11:3

0:00

12:0

0:00

12:3

0:00

13:0

0:00

13:3

0:00

14:0

0:00

14:3

0:00

15:0

0:00

15:3

0:00

16:0

0:00

16:3

0:00

17:0

0:00

17:3

0:00

18:0

0:00

18:3

0:00

19:0

0:00

19:3

0:00

20:0

0:00

20:3

0:00

21:0

0:00

21:3

0:00

22:0

0:00

22:3

0:00

23:0

0:00

23:3

0:00

MW

MW

State Load Solar Generation Wind Generation

Forecasting can facilitate an optim al Cross B order P ower Trade

Kreate’s predictive analytic tool can facilitate a optimal regional grid integration through better planning through introduction of predictive analytics

At present, limited power is being traded through bilateral arrangements between Bangladesh and India (1160 MW), Bhutan and India (1,416 MW), and Nepal and India (190 MW) .

Effective utilization of surpluses and matching peaking deficit through better forecast .

Precise forecast aids in Competitive Market participation of cross border country

Current India - Nepal CBET ~190 MW

Current India - Bhutan CBET ~1160 MW

Current India - Bangladesh CBET ~1416 MW

Total Maximum CBET Trade in SA 2450 MW

Happy to do pilot projects on D e m and and R E fore c as ting

Contact us at : red foreca st@ k rea tetec h n ologies .com b d _reforeca st in g@ k rea tetec h n ologies .com

T H A N K YO U