Presentation on theme: "Energy production forecasting based on renewable sources of energy"— Presentation transcript:
1Energy production forecasting based on renewable sources of energy Good afternoon ladies and gentlemen.Thank you for inviting me at this conference.This is a privilege and also a pleasure.I would like to thank Prof. Miguel Tufino and Rolando Jimenetz for this opportunity and their kindness.S. LevaPolitecnico di Milano, Dipartimento di EnergiaVia La Masa 34, Milano, Italy
2Goal and outlineThe goal of this speech is to analyze how, starting from weather forecast, we can predict in term of hourly-curve the energy production by RES for one day – two days, a week ahead.Introduction: the energy production forecasting and the role of RES set up by the international energy agencyThe energy forecasting from RESWeather forecastingThe PV forecasting and error definitions, some examplesThe wind forecasting, some examplesConclusionsThe goal of this speech is analyzing how, starting from meteo forecast, we can predict in term of hourly-curve the energy production by RES for one day – two days, a week ahead.Here you can find the outline of this presentation.First of all I will introduce same data about the energy production forecasting and the role of RES following the international energy agency scenario. Then the role of res in Italy well be presented.This can allow us to understand how the grid is changing and the importance of forecasting the energy produced by unpredictable sourses like PV and Wind.Speacking about forecasting we’ll start on “weather forcasting”.Then, the core of this presentation, PV forecasting. In particolar a brief review of the forcasting models communly used and of same error definitions will be presented.Furthermore some examples coming from real plants in Italy will show underlying the positive results but also the limits of the RES energy forecast.Then few minutes about wind forecasting.Finally some conclusions.
3Introduction: the energy production forecasting and the role of RES in the world and in Italy The energy forecasting from RESWeather forecastingThe PV forecasting and error definition, some examplesThe wind forecasting, some examplesConclusions
4Introduction: the energy production forecasting and the role of RES The IEA forecasts confirm that the demand for energy (not just electricity) will grow especially in non-OECDShare of global energy demandThe International Energy agency confirms that the demand for energy in the next years will grow especially in non-OECD countries.Global energy demand resis will be mainly due to the rise of living standars in China, India and Middle east countries.Global energy demand rises by over one-third in the period to 2035, underpinned by rising living standards in China, India & the Middle East
5Introduction: the energy production forecasting and the role of RES IEA predictions for the future (scenario "reference"): oil, gas, coal continue to dominate the energy (not just electricity) productionMeeting the global demand for energy continue to be the key challenge to sustained industrialization.IEA prediction for the energy production confirms oil, gas and coal continue to dominate.We are speacking about energy production at all not just about electricity.
6Introduction: the energy production forecasting and the role of RES IEA predictions of how will be satisfied the demand of electricity in the world.«KING» COAL!And what about electricity?RES (wind and PV) will continue to have the largest growths in percentage.Nevertheless gas and coal will is still going to grow.It seem that Coal should continue to be the king of electricity production.Considering the IEA prediction of how will satisfied the demand of electricity in the world we can see that renewables (wind, PV) will continue to have the largest growths in percentage but the energy production from gas and coal will is still going to grow.Coal should continue to be the king of energy production.
7Introduction: the role of RES in Italy In five years the electricity generation by RES in Italy has doubled.HydroGeothermalBioenergyWindSolarNow some words about the situation of Italy. Italy can be an interesting example of how the electricity generations is changing and how the electric system is moving to distributed system giving us new challenges.In Italy RES have experienced one of the largest growths: in the last five years the electricity generation by res is doubled. Thanks in particularly to PV and wind pushed by public policy (feed in tariff). Instead the production of energy by hydro has not changed at all.The data are really up to date: august 2013!
8Introduction: the role of RES in Italy Electricity generation in Italy in the first seven monthes of 2013Number of plants producing electricity passes in a decade from 1 thousand to 550,000Centralized system tends towards a mixed system of generation (distributed generation)A growing number of households and factories now are involved in electricity generationBioenergygeothermalwindPVThermoelectric fossilHydroFurthermore, still speaking about Italy, we can see that wind+bio+PV production of energy is equal to the one produced by idro. This involves many aspects:1. The number of plants producing electricity passes in a decade from 1 thousand to 550,0002. The centralized system of generation tends towards a mixed system of generation: the usually called distributed generation or dispersed generation when the generation systems are very small in term of rated power.3. A growing number of households and factories now are involved in electricity generation.
9Introduction: the energy production forecasting and the role of RES in the world and in Italy The energy forecasting from RESWeather forecastingThe PV forecasting and error definition, some examplesThe wind forecasting, some examplesConclusion
10The energy forecasting from RES Distributed system:grid-connected RES installations are decentralizedRESs energy production has a stochastic behavior.RESs are much smaller than traditional utility generatorsToday's available transformation and storage capabilities for electric energy are limited and cost-intensive.In centralised system the produced energy is relatively independent on the plant location.Challenges of controlling and maintaining energy from inherently intermittet sources involves many aspects: efficicency, reliability, safety, stability of the grid and ability to forecast energy production.
11The energy forecasting from RES Forecasting of PV/wind is an estimation from expected power production of the plant in the future.For monitoring and maintenance purposesTo help the grid operators to better manage the electric balance between power demand and supply and to improve embedding of distributed renewable energy sources.In stand alone hybrid systems energy forecasting can help to size all the components and to improve the reliability of the isolated systems.
12Time scale classification for RES Forecasting horizonRangeApplicationsVery short-termFew seconds to30minutes ahead- Control and adjustment actionsShort-term30 minutes to 6hours ahead- Economic Dispatch Planning- Load Increment/Decrement DecisionsMedium-term6 hours to 1 dayahead- Generator Online/Offline Decisions- Operational Security in Day-Ahead- Electricity MarketLong-term1 day to 1 weekor more ahead- Unit Commitment Decisions- Reserve Requirement Decisions- Maintenance Scheduling to Obtain Optimal Operating CostThis table shows a possible time scale classification for forecasting of RES.This table has been set up with reference to wind but it can be also applied with reference to PV.For PV systems very short-term horizon it is not so important.Instead for wind systems and in particular farm it is really important.For the utility two applications are very important: the scheduling of power plants and the market.The typical time scales for start-up of conventonal power plant are between 20 min (for gas turbine) and 8 hours or more (for large coal or oil plants) long-term forecastUsually for electricity market you need medium term forecast: you have to sell today the energy for tomorrow.
13Introduction: the energy production forecasting and the role of RES in the world and in Italy The energy forecasting from RESWeather forecastingThe PV forecasting and error definition, some examplesThe wind forecasting, some examplesConclusion
14Forecasts of RES production is based on weather forecasts. This is an orthogonal step to a grid operator: weather data is usually obtained from meteorological services.The most influencing factor for output determination are:solar energy production: global irradiation forecast.wind energy production: wind speed amplitude and direction forecast, pressure forecastThe use of precise weather forecast models is essential before reliable energy output models can be generated.In order to make energy supply planning rational, forecasts of RES production have to be made based on the consideration of weather conditions.Although this step is orthogonal to a grid operators core activities (weather data is usually obtained from meteorological services), a basic understanding of the underlying principles is helpful when choosing a specific energy output model.
15Weather forecast models Numerical Weather Prediction (NWP)Complex global NWP models are used to predict a number of variables describing the dynamic of the atmosphere and then to derive the weather at a specific point of interest.Post processing techniques are applied to obtain down scaled models (1.5 km).European Center for Medium-Range Weather-Forecasts Model (ECMWF)Global Forecast System (GFS),North American Mesoscale Model (NAM)3-6 horsCloud ImageryInfluence of local cloudiness is considered to be the most critical factor for estimation of solar irradiation.The use of satellite provide high-quality medium term forecast.Satellite-based (METEOSAT),Total Sky Imager,24h-48hStatistical Methodsbased on historical observation data using time series regression modelsARIMAArticial Neural Networks (ANN),Fuzzy Logic (FL),ARMA/TDNNANFISRBFNNMLPLong term
16Meteorology remains a field of uncertainty. Weather forecastMeteorology remains a field of uncertainty.Time horizon is a crucial aspect. Sunshine and wind speed can only be predicted with accuracy a few days in advance.The number and type of variables describing the physics and dynamic of the atmosphere are fundamental topics. Cloudy index or irradiation are two indexes that can impact on the forecast in a different way.
17The energy forecasting from RES Weather forecasting Introduction: the energy production forecasting and the role of RES in the world and in ItalyThe energy forecasting from RESWeather forecastingThe PV forecasting and error definition, some examplesThe wind forecasting, some examplesConclusionAny output from the weather models described above must then be converted into electric energy output. According to the underlying methodology, the existing solutions can be itemized into the categories of physical, statistical and hybrid methods.
18The PV forecast: different Models. Physical Modelsto describe the relation between environmental data and powerhighly sensitive to the weather predictionhave to be designed specifically for a particular energy system and locationStatistical Modelsare based on persistent prediction or on the time series' historyPersistent prediction, Similar-days ModelStochastic Time SeriesMachine LearningArtificial neural network (ANN) learn to recognize patterns in data using training data sets.They need historical weather forecasting data and PV-plant measured data for their trainingAny combination of two or more of the previously described methods is a hybrid model. The idea is not new: we combine different models with unique features to overcome the single negative performance and finally improve the forecast.We can combine two different stochastic models or a stochastic model with a physical model.Hybrid Models are any combination of two or more of the previously described methods. They could be two different stochastic models or a stochastic model and a physical model.
19The PV forecast: Physical Models. Weather forecastPV energy forecastPV energy forecastPhysicalAlgorithmGlobal Irradiation, Cloud cover, Temperature, eccPlant Description; Monitoring SystemMeasured dataThe usage of such models requires detailed technical knowledge about characteristics and parameters of all underlying components, thus making them more relevant for energy plant owners or producers than for grid operatorsWe can improve the model adding temperature, or other meteorological parameters.
20The PV forecast: Statistical Models TRAINED NEURAL NETWORKEnvironmental temperatureANN are black box methods but they can consider many atmospherical paramiters to perform the energy forecast.
21Error DefinitionsIn order to correctly define the accuracy of the prediction and the relative error it is necessary to analyze different definitions of error. The starting point reference is the hourly error eh:Pm,h is the average power produced in the hour (or energy kWh)Pp,h is the prediction provided by the forecasting modelFrom this basic definition, other error definitions have been inquired:Absolute error based on the hourly output expected power (p=predicted) [AEEG]:absolute error based on the hourly output produced power (m=measured) [AEEG]:Some of the error definitions used in energy forecasting come from maths (statistics) theory. Others are introduced by regulatory authority for market issues (in Italy AEEG=Autorità energia elettrica e gas. Authority for Electricity and Gas)The error definitions are really different among them and it is not easy to chose one of these. Also the technical papers present a lot of this indexes.Here we report some of the most important error definitions.AEEG=Italian Authority for Electricity and Gas
22Error Definitions Mean absolute error [AEEG et al]: Normalized mean absolute error NMAE, based on net capacity of the plant C [AEEG et al]C could be the rated power, the maximum observed or expected power!!!!Look at C (net capacity of the plant): it could be the rated power, the maximun observed power or the expected power!!!! It depends on the Author of the papers!
23Error definitionsWeighted mean absolute error WMAE% based on total energy production [AEEG et al.]:Normalized root mean square nRMSE, based on the maximum observed power [Urlicht et al]:
24Some examples: Hybrid Models (ANN+Physical) Physical data: Theoretical Solar Irradiance (clear sky), Sunrise-, Sunset-hourweather forecastsPlant data validation: Theoretical Solar Irradiance (clear sky)Speaking about hybrid model I’m going to present and discuss a model developed at the Department of Energy of the Politecnico di Milano.My research group has made in the last three years great efforts for developing models that can improve the forecast and that can done the energy forecast usefull for the grid and the electricity market.Also a big work has been done about the error analysis.In fact these topics are really important for the application of the power forecasting both on the technical and economic point of views.This forecasting method for PV plants based on Neural Network can be described by the three following steps:Input data: the historical data (for training process) are first of all analyzed verifying their consistence on the point of view of physical behavior of the system. In particular the irradiance is compared with the clear sky theoretical model with the aim to individuate sunset sunrise the maximum irradiation. Furthermore not useful data are erased.The ann process: the neural network has to be trained by using historical data set. Then starting from weather forecasting data and from clear sky data set the power forecasting for the day X+1 is performed.Analysis of the error: taking into account the error definition some evaluation can be made. Usually an assessment of the obtained results and a comparison between different methods.weather forecasts data analysis: evaluation of their reliability.Comparison between ANN forecasts and other methodsEnsembled methodsError definitionsAccuracy assessment of the obtained results
26A. Some Results: Solar Tech Lab NMAEp%= 3.08%Figure on the left:Red line: predicted powerBlue line: mesured powerFigure on the right:Green line: Global tilted irradiation 30°Blue line: Global horizontal irradiationBlack line: theoretical global irradiation 30°Pink line: cloud cover forcastedRed line: measured powerLooking at the second day (cloudy day) we can see, observing the pink line, that there was an error in the weather forecast.In fact the forecasted cloud cover (pink line) shows a riduction in the global radiation starting from sunrise to 12 and then during the night. Instead the measured radiation shows that cloud cover was high only from 9 to 12.NMAEp% = 30.1%pink line: there was an error in the weather forecast.
27B. Hybrid Models (ANN+Physical) PV Plant in Cuneo 285kW PV plant, Cuneo (Italy)Meteo dataset: Day, hour, Environmental temperature, wind direction, wind speed, global solar irradiationGoals:Analysis of the error due to the weather forecastingEnsembles method: use more than one trials of stochastic methods to make the forecastAbsolute hourly error based on predicted power vs measured powerANNs are stochastic methods then we can use more than one trials to make the forecast
28B. Hybrid Models (ANN+Physical) PV Plant in Cuneo Error due to the weather forecasting: difference between the irradiation given by weather service and the irradiation measuredThis allows us to analyze the error due to the weather forecasting.In this figure we can see that there is a difference in the irradiation given by weather service and the one measured.This difference is very high in the second day. Here the meteo service forecasted a cloudy day in the morning. Instead we had a sunny day.
29B. Hybrid Models (ANN+Physical) PV Plant in Cuneo Error due to the weather forecasting: Absolute hourly errors of GI are sorted from largest to smallest.Absolute hourly error based on expected global irradiation (predicted) and on the measured global irradiation.To better understand the entity of the error in this picture we can see the absolute hourly error based on expected irradiation (predicted) and on the measured irradiation.Hourly errors are sorted from largest to smallest to allow us an easier comparison.We can see that the hourly error could be huge: when weather forecast give very low or zero at all irradiations.The normalized error is good enough: about 13%.Solar Radiation forecastings are affected by a great error!
30Some Results: Power Plant ANN are stochastic methods: Different trials give different forecasting curves.Ensemble: power/energy forecast is calculated considering the hourly average value of different (here 10) trials.Absolute hourly error based on expected output power (predicted) and on the measured output power.Ensemble methods reduce the error!The error based on the measured power is bigger than the one based on the predicted!The ANN are stochastic methods than different trials give different forecast.The forecast can be evaluated considering not a single trial but the average value of more trials.Hourly sample (from sunrise to sunset)
31Some Results: Power Plant It is very difficult to obtain a day error lesser than 10%, in particular in cloudy days.NMAEp% = 10NMAEr% = 5.86WMAEp% = 16.58NMAEp% = 29.14NMAEr% = 15WMAEp% = 50NMAEp% = 16NMAEr% = 7.33WMAEp% = 28.7
32Some Results: Power Plant expected output power (predicted) and versus measured output power.1 year: NMAEp = 12.15%, NMAEr%=7,34%
33Introduction: the energy production forecasting and the role of RES in the world and in Italy The energy forecasting from RESWeather forecastingThe PV forecasting and error definition, some examplesThe wind forecasting, some examplesConclusion
34Wind ForecastingForecasting of wind is an estimation from expected power production of the wind turbines in the future. This power production is expressed in kW or MW depending on the nominal capacity of the wind farm.Forecasting methods described for PV can be appliedError definitions described for PV are usedKalman or Kolmogorov-Zurbenko are usually adopted to better extimate the wind speed eliminating the effects of noise and systematic errorsHybrid approaches (ANN + CFD computational fluid dynamics software) can improve the accuracy of the forecasting
35Example: Castiglione Messer Marino Wind Farm Input parameters:Inviromental temperature [°C]Atmospheric pressure [hPa]Wind speed intensity [m/s]Humidity [%]Cloud coverage [%]Performance parametersWMAENMAEImplemented feed-forward ANN with details on input, output, and hidden layers.
37Hybrid methods: computational fluid dynamics software The use of tools of CFD (computational fluid dynamics software) can improve the predictive capability of forecasting systems.The computational cost greatly limits its practical applicability for wind farms with a large number of wind turbines.Expensive measurement systems (see anemometer towers) to model the field.
38The most promising method: Hybrid methods Plant DescriptionANNHistorical Wind dataHistorical Power dataGround descriptionPhysical algorithmCFD Analysisby GSE, ANEMOS.plus
39Introduction: the energy production forecasting and the role of RES in the world and in Italy The energy forecasting from RESWeather forecastingThe PV forecasting and error definition, some examplesThe wind forecasting, some examplesConclusions
40ConclusionsThe meteorological services have an important influence on the power forecasting system for PV and wind energy.The input data analysis is very important and cost-intensiveHybrid forecasting method are the most promising methods both for PV and Wind energy forecastingPV. Clear sky data are very useful to reduce error.Wind. The use of special filters (eg Kalman or KZ) may be useful for the removal of systematic errors of the forecasts of wind speed provided by the NWP and used as input to statistical methods.The performance of the forecasting models are strongly related to the time horizon of the forecast and to the characteristics of the land on which the plant/farm is placed.The need for energy forecasting from RES is a recent topic!!!as the challenge in RES penetration has emerged just a couple of years ago.
41THANK YOU! www.solartech.polimi.it Diapartimento di Energia Via Lambruschini, 420133 MilanoTel (Centralino)3709 (Leva) – 3810 (Manzolini)
42Some Results: Power Plant Absolute hourly error based on expected output power (predicted) and on the measured output power.