The case for geographical diversification in renewable energy portfolios - Does the application of volatility forecasting and Mean Variance Optimization provide more value than equally weighted portfolios?
Author
Jónsson, Jón Frímann
Term
4. semester
Publication year
2023
Submitted on
2023-05-30
Abstract
Specialet undersøger, om finansielle porteføljemetoder kan øge værdien af geografisk diversificerede vindporteføljer i elmarkederne. Med udgangspunkt i den voksende udbredelse af vedvarende energi, BRP’ers rolle og udfordringer som pris-kannibalisering, vurderes om volatilitesprognoser via DCC-GARCH kombineret med Mean Variance Optimization (MVO) kan skabe mere værdi end en naiv ligevægtet tilgang. Analysen omfatter vindeksponering i Danmark, Finland, Italien, Belgien og Tyskland. Metodisk modelleres afkast og tidsvarierende kovarianser med ARMA/GARCH og DCC-GARCH, der valideres med stationaritets- og diagnostiske tests (bl.a. Phillips-Perron, Ljung-Box og ARCH-LM). De prognosticerede kovariansmatricer anvendes i MVO over tre horisonter inden for et år, og resultaterne sammenlignes out-of-sample med en MVO baseret på simpel stikprøve-kovarians samt en ligevægtet portefølje. Porteføljernes værdi måles primært ved volumen-vægtet gennemsnitspris (VWAP) pr. produceret MW og risiko ved varians. Resultaterne peger på, at DCC-GARCH-baseret MVO giver højere værdi end en ligevægtet portefølje, men at varians som risikomål kan være utilstrækkeligt under ekstreme prisforhold i elmarkederne, hvilket indikerer behov for mere avancerede risikomål.
This thesis examines whether financial portfolio methods can increase the value of geographically diversified wind portfolios in power markets. Motivated by the expansion of renewables, the role of Balance Responsible Parties (BRPs), and challenges such as price cannibalization, it evaluates if volatility forecasting using DCC-GARCH combined with Mean Variance Optimization (MVO) outperforms a naive equally weighted approach. The analysis focuses on wind exposure across Denmark, Finland, Italy, Belgium, and Germany. Methodologically, returns and time-varying covariances are modeled with ARMA/GARCH and DCC-GARCH, supported by stationarity and diagnostic tests (including Phillips-Perron, Ljung-Box, and ARCH-LM). Forecast covariance matrices are fed into MVO at three horizons within a year and compared out-of-sample with an MVO using sample covariance and a naive equal-weight portfolio. Portfolio value is assessed mainly via volume-weighted average price (VWAP) per produced MW, with variance as the risk metric. Findings indicate that DCC-GARCH-based MVO yields higher value than equal weighting, while variance alone may not adequately capture risk under extreme electricity price events, suggesting the need for more comprehensive risk measures.
[This summary has been generated with the help of AI directly from the project (PDF)]
Documents