Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.

Stochastic modelling of intermittent renewables in TIMES models.

66 vues

Publié le

Stochastic modelling of intermittent renewables in TIMES models.

Publié dans : Environnement
  • Soyez le premier à commenter

  • Soyez le premier à aimer ceci

Stochastic modelling of intermittent renewables in TIMES models.

  1. 1. Stochastic modelling of intermittent renewables in TIMES models Lisa Kvalbein, Pernille Seljom Renewables energy systems, IFE 02.07.2020 ZOOM 77th Semi-annual ETSAP workshop
  2. 2. • Assessment of the Value of Flexibility Services from the Norwegian Energy System (ASSETS) • Important to model intermittent renewables in a proper manner to provide good model insights on the low carbon transition • This work investigates how to model wind and PV generation that provides reasonable results • Demonstrated in IFE-TIMES-Norway Motivation 2
  3. 3. 3 Wind PV (real production) Intermittent renewable has a high short- term uncertainty. In Norway the weather changes rapidly throughout the days. For a good representation of wind and PV in the model, uncertainty is needed to be taken into account.
  4. 4. • Stochastic programming (SP) is a mathematical framework to consider uncertainty & value flexibility in optimisation models • Ignoring uncertainty can give misleading decision support from TIMES models • Depending on analysis • SP can be used to explicit model uncertainty and value flexibility Stochastic programming of short-term uncertainty 4
  5. 5. • Two-stage stochastic model • Scenarios represent realisation of uncertain parameters • Scenario independent investment decisions & scenario dependent operational decisions • Minimise: Investment costs + expected operational costs Short-term uncertainty 5 2015 2020 2030 2040 s1 s2 s3 p1 p2 p3 Investments Operation
  6. 6. • Scenario is a discrete representation of uncertainty with a given probability to occur • Number of scenarios effect the computational effort • Poor scenarios can give misleading results and insights • Important to evaluate the quality of the stochastic scenarios and model solution Stochastic scenario generation • Example: 21 stochastic PV production scenarios Objectives: - How many scenarios is needed to provide reasonable model insights? - What is a good method to generate stochastic scenarios? 6
  7. 7. • Stochastic scenarios considerations • Temporal and regional correlations • Statistical properties/moments • Scenario generation methods • Random sampling (2 seconds per season) • Statistical methods: Moment matching • Repetitive sampling (10 min per season) • Optimization (Time limit of 6h per season) • Solution evaluation methods • In-sample and out-of-sample stability tests Stochastic scenario generation 7
  8. 8. • Input – 19 years of hourly resolution capacity factors for solar and wind in Norway • Total of 65 sets of scenarios created • M = 3, 9, 15, 21, 30 • Optimizing – 1 set for each M • Random Sampling – 10 set for each M • Repetitive Sampling – 10 set for each M • Total 21 set of scenarios created for each M • Also created deterministic scenario of median values Stochastic scenarios 8
  9. 9. • Is the scenario generator creating scenarios resulting in similar optimal values? Model result: In-sample stability 9 • The energy system cost depends on how PV and wind is modelled • More scenarios gives less variations among sampling methods
  10. 10. • 101 = 10 min • 102 = 100 min ~ 1,5 hours • 103 = 1000 min ~ 16,5 hours • 104 = 10 000 min ~ 167 hours ~7 days and nights Computational time 10 Scenario creation time not included The solution time depends highly on number of stochastic scenarios
  11. 11. Model results: Wind and PV in 2040 11 PV capacity depends highly on model representation of intermittent generation
  12. 12. 12 Conclusion and further work • Stochastic modelling of short-term uncertainty in TIMES models gives more realistic representation than a deterministic approach • Stochastic scenarios, number of scenarios and methods used, affect the solution quality • PV is influenced the most in our analysis • Further analysis and quality tests on in-sample and out-of-sample stability will give insights on • Required number of scenarios • Suitable methods for scenario generation
  13. 13. Questions? Lisa Kvalbein Master of Science IFE