A look into the future of electricity price forecasting (EPF)

Size: px

Start display at page:

Download "A look into the future of electricity price forecasting (EPF)"

Mark Barnett
6 years ago
Views:

1 A look into the future of electricity price forecasting (EPF) Rafa l Weron Department of Operations Research Wroc law University of Technology (WUT) Wroc law, Poland 11 December 214 Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 1 / 31

2 Based on (IJF 3(4), 214, pp ): Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 2 / 31

3 IJF: The longest paper we have ever published... Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 3 / 31

4 Bibliometrics of electricity price forecasting (EPF) EPF journal articles and citations to those articles Number of WoS indexed articles and citations Journal articles Citations ( 25) Number of Scopus indexed articles and citations Journal articles Citations ( 25) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 4 / 31

5 Bibliometrics of electricity price forecasting (EPF) Ten most popular journals IEEE Transactions on Power Systems Electric Power Systems Research Int. J. Electrical Power & Energy Systems Energy Conversion and Management Energy Economics IET Generation Transmission & Distribution International Journal of Forecasting Applied Energy Energy Policy IEEE Transactions on Smart Grid Neural network & time series Neural network only Time series only Other methods Number of articles (2 213) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 5 / 31

6 What and how are we forecasting? Forecasting horizons Short-term From a few minutes up to a few days ahead Of prime importance in day-to-day market operations Medium-term From a few days to a few months ahead Balance sheet calculations, risk management, derivatives pricing Inflow of finance solutions Long-term Lead times measured in months, quarters or even in years Investment profitability analysis and planning Beyond the scope of this review Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 6 / 31

7 What and how are we forecasting? A taxonomy of modeling approaches Electricity price models Multi-agent Fundamental Reduced-form Statistical Computational intelligence Cournot- Nash framework Parameter rich fundamental Jumpdiffusions Similar-day, exponential smoothing Feed-forward neural networks Supply function equilibrium Parsimonious structural Markov regimeswitching Regression models Recurrent neural networks Strategic productioncost AR, ARX-type Fuzzy neural networks Agent-based Threshold AR Support vector machines GARCH-type Hybrid Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 7 / 31

8 Agenda A look into the future of EPF 1 Fundamental price drivers and input variables Modeling and forecasting the trend-seasonal components The reserve margin and spike forecasting 2 Probabilistic forecasts ( Jakub Nowotarski s talk) Interval & density forecasts 3 Combining forecasts ( Jakub Nowotarski s talk) Point & probabilistic forecasts 4 Multivariate factor models ( Katarzyna Maciejowska s talk) 5 The need for an EPF-Competition A universal test ground Guidelines for evaluating forecasts Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 8 / 31

9 Modeling the trend-seasonal components Standard approach decompose a time series of prices P t into the long-term trend-seasonal component (LTSC) T t, the short-term seasonal component (STSC) s t, and the remaining variability, error or stochastic component X t The hourly/weekly STSC is usually captured by autoregression & dummies forecasting is straightforward Annual seasonality is present in spot prices, but in most cases the LTSC is dominated by a more irregular cyclic component Due to fuel prices, economic growth, long-term weather trends See e.g. Janczura et al. (213), Nowotarski et al. (213b) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 9 / 31

10 Modeling the LTSC Nord Pool spot price [EUR/MWh] Spot price Wavelet based LTSC Sine Monthly dummies Days [ ] Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 1 / 31

11 Adequate seasonal decomposition is important! 1 5 Wavelet based: α=9.6, β=.48, (α/β=2.), σ=6.17, µ=71.98, γ=.13, λ=.1 Simulated stochastic component (X t ) Sine: α=1.7, β=.6, (α/β=17.11), σ=2.75, µ=1.52, γ=26.96, λ= Monthly dummies: α=1.11, β=.6, (α/β=2.13), σ=2.7, µ=1.49, γ=23.75, λ= Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 11 / 31

12 Forecasting a wavelet-based LTSC (Nowotarski, Tomczyk & Weron, 213, Energy Economics) Spot price Median Last observation Spike filtered last obs. EEX price [EUR/MWh] Days [Dec 2, 24 Dec 19, 27] Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 12 / 31

13 Forecasting a wavelet-based LTSC cont. (Nowotarski, Tomczyk & Weron, 213, Energy Economics) Spot price Exponential decay Exp. decay (spike filtered) EEX price [EUR/MWh] Days [Dec 2, 24 Dec 19, 27] Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 13 / 31

14 Forecasting a wavelet-based LTSC cont. (Nowotarski, Tomczyk & Weron, 213, Energy Economics) Spot price Wavelet Wavelet (spike filtered) EEX price [EUR/MWh] Days [Dec 2, 24 Dec 19, 27] Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 14 / 31

15 Forecasting a wavelet-based LTSC cont. (Nowotarski, Tomczyk & Weron, 213, Energy Economics) Spot price Wavelet forecast Wavelet forecast (spike filtered) EEX price [EUR/MWh] Days [Dec 2, 24 Dec 19, 27] Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 15 / 31

16 Wavelets beat sines and monthly dummies (Nowotarski, Tomczyk & Weron, 213, Energy Economics) The number of times models from a given family are ranked in the top 5, 2 and 5 of all 34 models according to GM(MAE h, ), GM(MSE h, ) and MAPE h, for each of the six forecast horizons h = 1,...,6 #times in "top 5" #times in "top 2" GM(MAE h,* ) Expected 2 year 3 year GM(MSE h,* ) MAPE h,* #times in "top 5" Model class Model class Model class Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 16 / 31

17 The Hodrick-Prescott (198, 1997) filter A simple alternative to wavelets Originally proposed for decomposing GDP into a long-term growth component and a cyclical component Returns a smoothed series τ t for a noisy input series y t : { T } T 1 [ ] 2 min (y t τ t ) 2 + λ (τ t+1 τ t ) (τ t τ t 1 ), τ t t=1 t=2 Punish for: deviating from the original series roughness of the smoothed series (1) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 17 / 31

18 HP-smoothing: EEX and PJM (Weron & Zator, 215, Energy Economics) Price [EUR/MWh] EEX spot price HP, λ =5x1 4 HP, λ =5x1 5 HP, λ =5x1 7 Price [USD/MWh] Days ( ) PJM spot price HP, λ = 5x1 4 HP, λ = 5x1 5 HP, λ = 5x Days ( ) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 18 / 31

19 HP provides a better fit than the nominal LTSC (Weron & Zator, 215, Energy Economics) Identification technique (estimated LTSC model) HP filter-based (λ =...) Wavelet-based sin- 5x x x x1 7 S 5 S 6 S 7 S 8 EWMA Nord Pool market (3 years: ) S S S S sin EEX market (5 years: ) S S S S sin PJM market (8 years: ) S S S S sin Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 19 / 31

20 A look into the future of EPF 1 Fundamental price drivers and input variables Modeling and forecasting the trend-seasonal components The reserve margin and spike forecasting 2 Probabilistic forecasts ( Jakub Nowotarski s talk) Interval & density forecasts 3 Combining forecasts ( Jakub Nowotarski s talk) Point & probabilistic forecasts 4 Multivariate factor models ( Katarzyna Maciejowska s talk) 5 The need for an EPF-Competition A universal test ground Guidelines for evaluating forecasts Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 2 / 31

The reserve margin and spike forecasting Reserve margin, also called surplus generation, relates the available capacity (generation, supply), C t, and the demand (load), D t, at a given moment in

21 The reserve margin and spike forecasting Reserve margin, also called surplus generation, relates the available capacity (generation, supply), C t, and the demand (load), D t, at a given moment in time t The traditional engineering notion: RM = C t D t Some authors prefer to work with dimensionless ratios: ρ t = Dt R t = Ct D t 1 or the so-called capacity utilization CU = 1 Dt C t C t, Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 21 / 31

22 The reserve margin and spike forecasting cont. Consider ρ(t 1, t 2 ) = D(t 1,t 2 ) C(t 1,t 2 ) calculated at time t 1 (e.g. today) for an upcoming period t 2 D(t 1, t 2 ) is the National Demand Forecast (Indicated Demand) C(t 1, t 2 ) is the predicted Generation Capacity (Indicated Generation, see See Cartea et al. (29), Maryniak and Weron (214) Plot P(spike ρ(t τ, t)) for different τ s Check how it depends on spike identification Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 22 / 31

23 Spikes identified in UK spot prices (23-212) Price [GBP/MWh] Price [GBP/MWh] Spikes (VPT) VPT filtered price Spikes (RSC) RSC filtered price Days 6[ ] #spikes in a cluster CF VPT RFP RSC #clusters Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 23 / 31

24 Number of spikes and spike probability #spikes CF τ=2d τ=3d τ=1w τ=2w P(spike CF ρ) τ=2d τ=3d τ=1w τ=2w ρ(t τ,t) ρ(t τ,t) Anderson and Davison (28): ρ = 85% is the industrial standard warranting a safe functioning of the power system Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 24 / 31

25 Spike probability... is roughly exponential P(spike CF ρ) τ=2d τ=3d τ=1w τ=2w P(spike VPT ρ) τ=2d τ=3d τ=1w τ=2w ρ(t τ,t) ρ(t τ,t) P(spike ρ) CF VPT α exp(2.91 ρ) P(spike ρ) RFP RSC α exp(19.4 ρ) ρ(t 2D,t) ρ(t 2D,t) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 25 / 31

26 Agenda A look into the future of EPF 1 Fundamental price drivers and input variables Modeling and forecasting the trend-seasonal components The reserve margin and spike forecasting 2 Probabilistic forecasts ( Jakub Nowotarski s talk) Interval & density forecasts 3 Combining forecasts ( Jakub Nowotarski s talk) Point & probabilistic forecasts 4 Multivariate factor models ( Katarzyna Maciejowska s talk) 5 The need for an EPF-Competition A universal test ground Guidelines for evaluating forecasts Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 26 / 31

27 5. The need for an EPF-Competition The need for an EPF-Competition Many of the published results seem to contradict each other Misiorek et al. (26) report a very poor forecasting performance of a MRS model, while Kosater and Mosler (26) reach opposite conclusions for a similar MRS model but a different market and mid-term forecasting horizons On the other hand, Heydari and Siddiqui (21) find that a regime-switching model does not capture price behavior correctly in the mid-term Cross-category comparisons are even less conclusive and more biased Typically advanced statistical techniques are compared with simple AI methods, see e.g. Conejo et al. (25a), and vice versa, see e.g. Amjady (26) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 27 / 31

5. The need for an EPF-Competition A universal test ground This calls for a comprehensive and thorough study involving 1 the same datasets 2 the same robust error evaluation

28 5. The need for an EPF-Competition A universal test ground This calls for a comprehensive and thorough study involving 1 the same datasets 2 the same robust error evaluation procedures 3 statistical testing and forecast evaluation GEFCom214 included an EPF track this year: Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 28 / 31

29 5. The need for an EPF-Competition Guidelines for evaluating forecasts A selection of the better performing measures for point forecasts weighted-mae, like the weekly-weighted WMAE seasonal MASE (Mean Absolute Scaled Error) RMSSE (Root Mean Square Scaled Error) should be used exclusively or jointly with the more popular ones (MAPE, RMSE), see e.g. Hyndman and Koehler (26) For probabilistic forecasts The interval or Winkler score can be used to evaluate PI and the Continuous Ranked Probability Score (CRPS) to evaluate density forecasts, see e.g. Gneiting and Raftery (27), Maciejowska et al. (214) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 29 / 31

30 5. The need for an EPF-Competition Guidelines for evaluating forecasts cont. Statistical tests for evaluating point forecasts The Diebold and Mariano (1995) test for the significance of the difference in forecasting accuracy; for uses and abuses see Diebold (213) The model confidence set approach of Hansen et al. (211) The test of forecast encompassing, see Harvey et al. (1998) And probabilistic forecasts The conditional coverage test of Christoffersen (1998); for extensions and alternatives see Berkowitz et al. (211) The Berkowitz (21) approach to the evaluation of density forecasts (popular in VaR backtesting) Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 3 / 31

31 The end Thank you Rafa l Weron (WUT) A look into the future of EPF , EFC14, St.Gallen 31 / 31

Importance of the long-term seasonal component in day-ahead electricity price forecasting: Regression vs. neural network models

Importance of the long-term seasonal component in day-ahead electricity price forecasting: Regression vs. neural network models Rafa l Weron Department of Operations Research Wroc law University of Science