Università degli Studi di Bergamo Department of Engineering Ph.D. in Economics and Management of Technology XXVII Cohort EUROPEAN ELECTRICITY DAY AHEAD MARKET A MULTIPLE TIME SERIES APPROACH Doctoral Dissertation Marta Trabucchi Supervisors: Prof. Luigi Buzzacchi Prof. Pia Saraceno November 2014 2 Abstract The energy market reform of the last decades is a complex restructuring process that first has opened up Member State electricity markets to competition and it gradually fosters them toward integration into the Single European Market. Even if national markets are still characterized by several differences in the production structures, regulation shapes a common market design at European level and voluntary measures have been adopted to promote market integration. The recent empirical literature highlights the presence of cointegration at least among the day ahead electricity markets of Central Western Europe. In this framework, Power Exchanges have taken a key role as shown by the growing volumes traded on their different segments and in recent years electricity price forecasting has become an interesting research field. However, up to now, most of the contributions on short term forecasting of day ahead electricity prices do not include the possibility of dynamic interactions between several interconnected electricity markets. After a primer on the economics of electricity markets and the analysis of the regulatory and market framework, the present work proposes a multiple time series approach for electricity price forecasting, joining the two strands of empirical literature on market integration and day ahead price forecasting. Accounting for the presence of market integration enlarges the model information set, so it may potentially improve the forecasting performance. This thesis considers hourly day ahead electricity prices for eight European countries (Austria, Belgium, France, Germany, Italy, Netherlands, Slovenia and Switzerland) for the period May 2010– July 2013. Multiple time series models have been used to forecast electricity prices for all the markets and an in-depth comparison between their accuracy and the one of simple time series models has been provided. At present the implemented forecasting exercise does not allow stating that estimating multiple time series models, and especially including potential cointegration relationships between day ahead electricity prices, greatly improve their forecasting performances compared to simple time series models. The adoption of multiple time series may lead to better results only in some hours and in other hours, simple time series models outperform multiple time series ones (especially ramp- up hours in the morning). Keywords: European electricity markets, electricity prices, forecasting, electricity market integration, multiple time series models 3 Acknowledgments This work has benefit from the support of several people. I gratefully acknowledge Luigi Buzzacchi, Pia Saraceno and Laura Solimene for their guidance and for insightful comments on this thesis. I am deeply in debt to Michele Dalena for patient day by day discussion and essential suggestions. I wish to express gratitude to all the members of the Istituto di Economia e Strategie d’impresa of the Catholic University of Milan: they pushed me on this road and constant support me during my doctoral studies. I would like to thanks also Pippo Ranci, who first taught me Energy Economics. Thanks also go to my PhD colleagues and all the Faculty Members of the Doctoral Programs in Economics and Management of Technology of the University of Bergamo. Finally, I heartily thank all my family and especially Carlo that every day reminds me that I “will never walk alone”. 4 Table of contents Introduction ........................................................................................................................................ 13 1 A primer on the economics of electricity market ........................................................................ 18 1.1 Electricity technical and economic features ........................................................................ 19 1.2 Market design ...................................................................................................................... 23 1.2.1 Energy transactions ...................................................................................................... 24 1.2.2 System Operations ....................................................................................................... 28 2 The regulatory Framework .......................................................................................................... 34 2.1 The electricity liberalization era .......................................................................................... 34 2.2 The European electricity reform .......................................................................................... 36 2.2.1 Toward a market based industry .................................................................................. 37 2.2.2 Toward the Single European Market ........................................................................... 40 2.3 The Electricity Target Model .............................................................................................. 43 2.3.1 Day ahead market coupling .......................................................................................... 43 2.3.2 Target Model for intraday, forward and real time timeframes .................................... 45 3 The market framework ................................................................................................................ 48 3.1 National electricity generation capacity .............................................................................. 48 3.2 Explorative analysis of wholesale markets ......................................................................... 51 3.2.1 Market liquidity............................................................................................................ 52 3.2.2 Price convergence ........................................................................................................ 53 4 Literature review ......................................................................................................................... 56 4.1 Forecasting electricity prices ............................................................................................... 56 4.2 Electricity markets integration in Europe ............................................................................ 60 5 Dataset description ...................................................................................................................... 63 Appendix A ........................................................................................................................................ 68 6 Methods ....................................................................................................................................... 76 5 6.1 An introduction to Vector Autoregressive Models for Multivariate Time Series ............... 76 6.2 Model specification ............................................................................................................. 79 6.2.1 Unit root and stationarity tests ..................................................................................... 79 6.2.2 The models implemented ............................................................................................. 84 Appendix B ........................................................................................................................................ 92 7 Day ahead electricity price forecasting ....................................................................................... 94 7.1 Short term forecasting ......................................................................................................... 94 7.1.1 Models setting .............................................................................................................. 94 7.1.2 Results .......................................................................................................................... 97 7.1.3 Conclusion ................................................................................................................. 105 7.2 Pre-filtered short term forecasting ..................................................................................... 107 7.2.1 Spike detection and substitution ................................................................................ 107 7.2.2 Filtered dataset description ........................................................................................ 108 7.2.3 Models settings .......................................................................................................... 111 7.2.4 Results ........................................................................................................................ 119 7.2.5 Conclusion ................................................................................................................. 125 7.3 Scenario based conditional forecasting ............................................................................. 126 Appendix C ...................................................................................................................................... 129 Conclusion and further developments .............................................................................................. 153 References ........................................................................................................................................ 155 6 List of Figures Figure 1.1: European Electricity Production (TWh) - 2013 ............................................................. 18 Figure 1.2: Hourly load values for Italy (MW) ................................................................................. 20 Figure 1.3: Hourly load (left) and Load duration curve (right)......................................................... 21 Figure 1.4: Optimal generation mix .................................................................................................. 22 Figure 1.5: Timeline of electricity transactions ................................................................................ 24 Figure 1.6: Market clearing price ...................................................................................................... 25 Figure 2.1: Main Steps in Electricity Reform ................................................................................... 35 Figure 2.2: EU Electricity Directives ................................................................................................ 42 Figure 3.1: Net generating capacity mix - 2013 ................................................................................ 49 Figure 3.2: RES plant evolution in net generating capacity mix (%) ............................................... 51 Figure 3.3: Wholesale market liquidity (%) ...................................................................................... 52 Figure 4.1: Empirical literature on price forecasting ........................................................................ 57 Figure 4.2: Empirical literature on European market integration (cointegration) ............................ 62 Figure 5.1: Average price by countries (May, 11th 2010 – July, 29th 2013) ..................................... 64 Figure 5.2: Average hourly load by countries in 2012 (MW) .......................................................... 66 Figure 7.1: Average hourly RMSE (€/MWh) ................................................................................... 97 Figure 7.2: Supply and demand structure ......................................................................................... 98 Figure 7.3: Hourly multiple times series model vs simple time series model ................................ 104 Figure 7.4: Hourly average Delta RMSE ........................................................................................ 105 Figure 7.5: EPEX France Spot price time series (Hour 10th) .......................................................... 109 Figure 7.6: Standard deviation by countries on the original (a) and on the filtered (b) dataset ..... 111 Figure 7.7: Average hourly RMSE (Pre-filtered dataset) ............................................................... 119 Figure 7.8: Hourly average Delta RMSE (Pre-filtered dataset) ...................................................... 125 Figure 7.9: Price changes across scenarios (VEC-X) ..................................................................... 128 7 List of Tables Table 1.1: Electricity demand by region and scenario (TWh) .......................................................... 19 Table 3.1: Net generating capacity (MW) ......................................................................................... 49 Table 3.2: Wholesale price convergence 2010-2013 ........................................................................ 54 Table 5.1: Descriptive statistics EPEX France price ......................................................................... 65 Table A.1: Descriptive statistics EXAA price ................................................................................... 68 Table A.2: Descriptive statistics Belpex price .................................................................................. 69 Table A.3: Descriptive statistics EPEX Germany price .................................................................... 69 Table A.4: Descriptive statistics IPEX price ..................................................................................... 70 Table A.5: Descriptive statistics APX price ...................................................................................... 70 Table A.6: Descriptive statistics BSP price....................................................................................... 71 Table A.7: Descriptive statistics EPEX Switzerland price ................................................................ 71 Table A.8: Descriptive statistics Austrian Load ................................................................................ 72 Table A.9: Descriptive statistics Belgian Load ................................................................................. 72 Table A.10: Descriptive statistics French Load ................................................................................ 73 Table A.11: Descriptive statistics German Load ............................................................................... 73 Table A.12: Descriptive statistics Italian Load ................................................................................. 74 Table A.13: Descriptive statistic Dutch Load ................................................................................... 74 Table A.14: Descriptive statistics Slovenian Load ........................................................................... 75 Table A.15: Descriptive Statistics Swiss Load ................................................................................. 75 Table 6.1: Augmented Dickey-Fuller test ......................................................................................... 81 Table 6.2: Phillips-Perron test ........................................................................................................... 81 Table 6.3: Kwiatkowsky-Phillips-Schmidt-Shin test ........................................................................ 82 Table 6.4: Lag selection VAR models .............................................................................................. 85 Table 6.5: Cointegration rank ............................................................................................................ 89 Table B.1: Lag selection VAR-X models ......................................................................................... 92 Table 7.1: The average SMAPE errors in percentages for all the hours of the day (%) ................... 99 Table 7.2: SMAPE errors from AR and VAR models (%) ............................................................. 100 Table 7.3: SMAPE errors from AR-X and VAR-X models (%)..................................................... 101 Table 7.4: SMAPE errors from ARI and VEC models (%) ............................................................ 102 Table 7.5: SMAPE errors from ARI-X and VEC-X models (%) ................................................... 103 Table 7.6: Summary statistics for EPEX France price (Pre-filtered dataset) .................................. 110 8 Table 7.7: Augmented Dickey- Fuller test (Pre-filtered dataset) .................................................... 112 Table 7.8: Phillips-Perron Test (Pre-filtered dataset) ...................................................................... 113 Table 7.9: Kwiatkowsky-Phillips-Schmidt-Shin test (Pre-filtered dataset) .................................... 113 Table 7.10: Lag selection for VAR models estimated (Pre-filtered dataset) .................................. 115 Table 7.11: Cointegration rank (Pre-filtered dataset) ...................................................................... 117 Table 7.12: The average SMAPE errors for all the hours of the day (%) (Pre-filtered dataset) ..... 120 Table 7.13: SMAPE errors from AR and VAR models (Pre-filtered dataset) ................................ 121 Table 7.14: SMAPE errors from AR-X and VAR-X models (Pre-filtered dataset) ....................... 122 Table 7.15: SMAPE errors from ARI and VEC models (Pre-filtered dataset) ............................... 123 Table 7.16: SMAPE errors from ARI-X and VEC-X models (Pre-filtered dataset) ....................... 124 Table 7.17: Average monthly price values (VEC-X model) – August, 2013 (€/MWh) ................. 127 Table 7.18: Price changes across scenarios (VEC-X) ..................................................................... 128 Table C.1: MAPE errors from AR and VAR models (%) .............................................................. 130 Table C.2: MAPE errors from AR-X and VAR-X models (%) ...................................................... 131 Table C.3: MAPE errors from ARI and VEC models (%) ............................................................. 132 Table C.4: MAPE errors from ARI-X and VEC-X models (%) ..................................................... 133 Table C.5: RMSE errors from AR and VAR models...................................................................... 134 Table C.6: RMSE errors from AR-X and VAR-X models ............................................................. 135 Table C.7: RMSE errors from ARI and VEC models .................................................................... 136 Table C.8: RMSE errors from ARI-X and VEC-X models ............................................................ 137 Table C.9: Descriptive statistics EXAA price (Pre-filtered dataset) ............................................... 138 Table C.10: Descriptive statistics BELPEX price (Pre-filtered dataset) ......................................... 138 Table C.11: Descriptive statistics EPEX Germany price (Pre-filtered dataset) .............................. 139 Table C.12: Descriptive statistics IPEX price (Pre-filtered dataset) ............................................... 139 Table C.13: Descriptive statistics APX price (Pre-filtered dataset) ................................................ 140 Table C.14: Descriptive statistics BSP price (Pre-filtered dataset) ................................................. 140 Table C.15: Descriptive statistics EPEX Switzerland price (Pre-filtered dataset) .......................... 141 Table C.16: Lag selection VAR-X models (Pre-filtered dataset) ................................................... 142 Table C.17: MAPE errors from AR and VAR models (%) (Pre-filtered dataset) .......................... 144 Table C.18: MAPE errors from AR-X and VAR-X models (%) (Pre-filtered dataset) .................. 145 Table C.19: MAPE errors from ARI and VEC models (%) (Pre-filtered dataset) ......................... 146 Table C.20: MAPE errors from ARI-X and VEC-X models (%) (Pre-filtered dataset) ................. 147 Table C.21: RMSE errors from AR and VAR models (Pre-filtered dataset).................................. 148 9 Table C.22: RMSE errors from AR-X and VAR-X models (Pre-filtered dataset) ......................... 149 Table C.23: RMSE errors from ARI and VEC models (Pre-filtered dataset) ................................ 150 Table C.24: RMSE errors from ARI-X and VEC-X models (Pre-filtered dataset) ........................ 151 Table C.25: Average monthly price values (VAR-X model) – August, 2013 (€/MWh)................ 152 Table C.26: Price changes across scenarios (VAR-X).................................................................... 152 10
Description: