Hydrometeorological forecasts for hydropower generation

Transcription

1 Hydrometeorological forecasts for hydropower generation Serre Poncon lake on Durance river Pierre BERNARD EDF-DTG Water Ressources Departement Hepex Workshop Reading 8-1th March 24

2 Water is for EDF: -raw material for hydraulic power plants => Hydropower is 2% of EDF installed capacity and 15% of yearly average production - cooling system mean for nuclear power plants Water ressources particularities in France: High variability in space and time Sudden flow rate variations => flood Illustrations

3 VARIABILITY IN SPACE mm Ain basin: VOUGLANS+BOLOZON GENEVE LYON For example: Average yearly rainfall (in mm) Durance basin SERRE PONCON 6-1 mm MARSEILLE

4 VARIABILITY IN TIME Year to year Evolution of the annual hydraulicity coefficient (on all EDF hydropowerplants) Evolution of the daily hydraulicity coefficient during FALL FLOOD RHONE and LOIRE RIVER NORMAL SUDDEN SNOW MELT HEAT WAVE AND DROUGHT Hydraulicity coefficient Yearly hydraulicity coefficient (Global France) 1-janv 31-janv 2-mars 1-avr 2-mai 1-juin 1-juil 31-juil 31-août 3-sept 3-oct 29-nov 3-déc Inside a year (as 23 )

5 FLOOD RISK => «Flash food» Upper Loire river during 2th October 21 flood Texte S=15 km2

6 Electricity can t be directly stored: production has to be continuosly ajusted to consumers needs Demand depends partially on weather conditions EDF production depends on many factor, including hydrology 2 kinds of hydropower generation Reservoirs in mountainous area Along the river without storage capacity For reservoir SERRE PONCON reservoir during 23 Management conditionned by snow cover conditions and long term inflow forecasts Water stored in reservoir is used in winter to produce peak power Filled in spring with snow melt Some of them have level constraint in summer for tourism or agriculture Reservoir level (mngf) oct 31-oct 3-nov 3-déc 3-janv 1-mars 31-mars 3-avr 31-mai 3-juin 3-juil 29-août 29-sept

7 So, need for EDF to have hydrological forecasts: -Long term forecast (few weeks) to optimize global production (nuclear and hydro) and decide reservoir management : normal snowy conditions 775 Reservoir level (mngf) : Winter drought Short term (1-4 days): to optimize hydraulic ressource(avoiding spilling) and assure to produce energy sold to costumers -Flood warning and real time survey: civil works and people safety 74 1-oct 31-oct 3-nov 3-déc 3-janv 1-mars 31-mars 3-avr 31-mai 3-juin 3-juil 29-août 29-sept Forescasts made inside EDF by 2 Hydrometeorological centers : Toulouse and Grenoble

8 TOOLS: Observations Real time monitoring survey, based on hydrological network operation Precipitation Air temperature Snow cover River discharge rainfall-runoff DPFT and MORDORdischarge routing PREDICE) Hydrological modelling Meteorological forecasts Satellite telemetry Meteosat «Arpege- Aladin» modelling outputs (Meteotel and maps) + CEP Radar Survey

9 Hydrological modelling : MORDOR Lumped, continuous model Daily time step Watershed area from 4 to 4 sq.mil Adapted to mountainous areas (snowpack model) Data input : rainfall (MAP), air temperature Automatic calibration (genetic algorithm on historical data) More than 3 in operational forecast use

10 Débit (en m3/s) 95,46 45,46 245,46 195,46 145,46 Retard (en h),45,4,35,3,25,2,15,1,5 MORDOR model t C TEMPERATURE mm Evapotranspiration IRRIGATION mm RAINFALL + Observed data t Atmosphere S 1 Snowpack Surface Subsurface Z U 2 L Direct Runoff Surface Runoff Deep Groundwater N 5 + Base Flow h 1h 2 h 3 h 4 h 5 h 6 h 7 h 8 h 9 h 1 h River discharge m3/s t

11 METEOROLOGICAL FORECASTS Products used: -from METEO FRANCE -EDF ANALOGUES model

12 METEO FRANCE products: -Daily outputs of meteorological models (french ARPEGE ALADIN + european CEP) : maps and numerical grids - real time : satellite telemetry METEOSAT precipitation radar survey

13 ANALOGUES model Aim: Daily rainfall probabilistic forecast ( Day to Day+6) based on analogy of meteorological situation Idea: «Same meteorological situations (pressure field) will lead to same consequences on rainfall phenomenon». Principle: 1- For each term (from D to D+6), research of the closest historical meteorological situations similar to the forecasted one ( pressure field) => K situations

14 ANALOGUES model : principle La recherche d analogues analogues situation search Current Jour situation courant J + A K voisins closest plus situations proches Historical Archive data météorologique A A Ensemble Set of analog de dates days ANALOGUES A Analogy Critère criterion d analogie 4 Actuellement, Geopotential 4géopotentiels: Fields used Teweless Score Teweless-Wobus Score 2 1

15 2-For the K analog days, retrieval of observed rainfall on every monitored watershed, in historical EDF data base. Les distributions conditionnelles Set of analog Ensemble days de journées ANALOGUES Bassin A 1953 Bassin B 1953 % 1993 % 1993 Historical Archive de précipitations Daily Rainfall journalières Database par bassin Empirical Distributions distributiionc empiriques 1% % 1% A % P (mm/24h) B P (mm/24h) % % Fonctions Fitted de continuous répartition ajustées distribution 9% 6% 2% A 9% 6% 2% B Q6 Q9 P (mm/24h) Q2 Q6 Q9 P (mm/24h)

16 3-For each monitored watershed and for each day (D to D+6), quantitative and probabilized rainfall forecasts

17 OPERATIONAL ORGANIZATION -Analysis of meteorological and hydrological situation continously made each morning to adapt hydrometeorological information to watershed scale and to EDF end users -Bulletins and warning message, phone assistance -Human investment is important : -Meteorological uncertainties -Monitoring différences between forecasts and field reality is useful for hydrologist: appreciates uncertainties, good recativity - 2 aspects: Real time and short time ( < 4 days) Long term

18 J J+1 J+2 J+3 J+4 J+5 J+6 6% 2% 9% Ma Mo REAL TIME OBSERVED DATA => RAINFALL => AIR TEMPERATURE => DISCHARGE =>(SNOW COVER) Short term (1 hour to 4 days ) Satellite telemetry Meteosat Radar Survey QUANTITATIVE FORECASTS Precipitation and Air température = > meteorological model ARPEGEALADIN (from MeteoFrance) 48h => ANALOGUES model: 6 days Hydrologic Model Or expert evaluation Débit horaire (en m3/s) : 13: 15: 17: 19: 21: 23: 1: 3: 5: 7: 9: 11: 13: 15: 17: 19: 21: 23: 1: 3: 5:? 7: 9: 11: 13: 15: 17: 19: 21: 23: 1: 3: 5: 7: 9: 11: 13: 15: 17: 19: Pluie spatiale (en mm/h) BASSIN: 4 AIN-VALSERINE Date de la prévision 19/2/99 2% 6% 9% Max Moy J J J J J J J

19 Particularities and perspectives - 1- Meteorology is determinant - without future rain with perfect future rain Crue du 17/12/97: l'ardèche à Sauze Saint Martin Crue du 17/12/97: l'ardèche à Sauze St Martin In real time, hydrological modelling run automaticly with only two hypothesis on future meteorology: no rain- same rain during 1hour than last observed. Débit (m3/s) 17/12 3h 17/12 5h 17/12 7h 17/12 9h 17/12 11h 17/12 13h 17/12 15h 17/12 17h 17/12 19h 17/12 21h 17/12 23h 18/12 1h 18/12 3h 18/12 5h 18/12 7h 18/12 9h 18/12 11h 18/12 13h 18/12 15h 18/12 17h 18/12 19h 18/12 21h 18/12 23h 19/12 1h 19/12 3h 19/12 5h 19/12 7h 19/12 9h 19/12 11h 19/12 13h 19/12 15h 19/12 17h 19/12 19h 19/12 21h 19/12 23h 2/12 1h 2/12 3h 2/12 5h Pluie (mm/h) /12 19h 17/12 3h 17/12 5h 17/12 7h 17/12 9h 17/12 11h 17/12 13h 17/12 15h 17/12 17h 17/12 19h 17/12 21h 17/12 23h 18/12 1h 18/12 3h 18/12 5h 18/12 7h 18/12 9h 18/12 11h 18/12 13h 18/12 15h 18/12 17h 18/12 19h 18/12 21h 18/12 23h 19/12 1h 19/12 3h 19/12 5h 19/12 7h 19/12 9h 19/12 11h 19/12 13h 19/12 15h 19/12 17h Débit (m3/s) 19/12 21h 19/12 23h 2/12 1h 2/12 3h 2/12 5h Pluie (mm/h)

20 2-Meteorological forecast is manually taken into account Deterministic approach (Aladin output modelling) Manual analyse and human expertise Some probabilistic approach with Analogues model (manually made but automatic process would be used) Meteorological ensemble forecasts important for hydrologist because of no linear hydrologic phenomenon 3- Probabilistic approach and ensemble forecast is interesting for short term forecast probability to reach a critical discharge (spill,historical flood,...) improve, for the end users, the understanding of the forecast and its uncertainties provide to the end users graduated informations, that can be translated,with his own other 5 additionnal informations, to cost -5 (or benefices)curves in order to 45 5 take better decision Critical discharge level 4-radar survey: actually only a qualitative way of use need to have a quantitative product (MAP improvement,anticipation

21 Long term (several weeks ) Real time observed data recipitation,air temperature, snow cover, discharge) MORDOR Model Water balance for the present time =>Water stored in snow and ground And the future? No meteorological forecasts Representation of future alea with local climatology and historical data base (P,T )since 1948 Hypothesis of climate stationnarity

22 Ensemble forecasts with différent years of climatology oct 8-oct 15-oct 22-oct 29-oct 5-nov 12-nov 19-nov 26-nov 3-déc 1-déc 17-déc 24-déc 31-déc 7-janv 14-janv 21-janv 28-janv 4-févr 11-févr 18-févr 25-févr 3-mars 1-mars 17-mars 24-mars 31-mars 7-avr 14-avr 21-avr 28-avr 5-mai 12-mai 19-mai 26-mai 2-juin 9-juin 16-juin 23-juin 3-juin 7-juil 14-juil 21-juil 28-juil With 1948 PAST FUTURE Precipitation and air température of year 1948 Observed precipitation Simulated discharge Precipitation year 1948 Observed discharge Q1 Q5 Q9 MORDOR simulated discharge Simulated discharge with 1948 climatology Observed air temperature Ait temperature year 148 Current hydrological situation

23 oct 8-oct 15-oct 22-oct 29-oct 5-nov 12-nov 19-nov 26-nov 3-déc 1-déc 17-déc 24-déc 31-déc 7-janv 14-janv 21-janv 28-janv 4-févr 11-févr 18-févr 25-févr 3-mars 1-mars 17-mars 24-mars 31-mars 7-avr 14-avr 21-avr 28-avr 5-mai 12-mai 19-mai 26-mai 2-juin 9-juin 16-juin 23-juin 3-juin 7-juil 14-juil 21-juil 28-juil With 1949 And so on..

24 oct 8-oct 15-oct 22-oct 29-oct 5-nov 12-nov 19-nov 26-nov 3-déc 1-déc 17-déc 24-déc 31-déc 7-janv 14-janv 21-janv 28-janv 4-févr 11-févr 18-févr 25-févr 3-mars 1-mars 17-mars 24-mars 31-mars 7-avr 14-avr 21-avr 28-avr 5-mai 12-mai 19-mai 26-mai 2-juin 9-juin 16-juin 23-juin 3-juin 7-juil 14-juil 21-juil 28-juil Distribution des apports Prévisions Histo Probabilité Apports (en hm3) Probabilistic ditribution => probabilistic inflow forecast

25 Inflow during the period Without forecasts:historical data Variability (3) Total Period

26 22 forecasts

27 Compared to observed inflows

28 -Questions and Perspectives 1-Observation and hydrological modeling are determinant and reduces variability. Residual variability is due to future precipitation. 2- How to reduce meteo alea using monthly forecasts? seasonal prediction? Downscaling difficulty: study research project to use analogue model to downscale seasonal forecast (ensemble geoptentiel fields=> watershed precipitation and air temperature)

29 Psaison/Pan Is climatology of the past enough representative of future? What we observed in the Alps: signal on precipitation and air temperature distribution inside the year Automne (oct-dec) Hiver (jan-mars) Printemps (avr-juin) Eté (juil-sept) Température ( C) An. hydro. Hiver Printemps Eté Automne Durance at SERRE PONCON reservoir 4 snow melt is earlier during the 1 past year 35 3 Quantile 1% Average Quantile 9% Average Year 23 Daily inflow (m3/s) janv 31-janv 2-mars 1-avr 2-mai 1-juin 1-juil 31-juil 31-août 3-sept 3-oct 29-nov 3-déc