Abstract/Summary

This thesis explores the potential application of subseasonal weather forecasts for the energy industry. Power contracts that energy companies buy and sell are subject to price and volume risk. A significant component of these risks is driven by weather variability. Accurate weather forecasts can help increase profits whilst reducing price and volume risk. However, meteorological research to date (in relation to the energy sector) has focused on forecasting up to 10 days ahead, as weather forecasts were traditionally considered to have limited to no skill thereafter. The aim of this research is therefore to show that meteorological forecasts can be used to make quantitative skillful predictions that can reduce risk within the energy sector at the subseasonal timescale. Although there is a large body of literature using NWP model output at lead times up to 10 days, there appears to be no prior research investigating the potential for subseasonal weather forecasts on the energy sector. A three step process was pursued in order to achieve this. Firstly, the forecast skill of wind speed and temperature (two key meteorological variables for the energy industry) was evaluated. Then wind power, demand and power price models were developed allowing the explicit incorporation of weather into the power price. This allowed quantification of the weather related skill and impacts on the power price and subsequent evaluation of applications that are contingent on the power price. These applications were evaluated using the forecasts to inform trading strategies in an effort to increase profits and reduce risk. The first section of research demonstrates that there is forecast skill of wind speed and temperature within the ECMWF monthly forecast model up to week 3 weeks ahead (Le. a weekly average over a lead time of day 14-21). The ECMWF model demonstrated cor¬relations of approximately 0.6 for the operational forecast and 0.3 for the hindcasts when forecasting week 3 UK winds speeds. Similar results were found for temperatures. By using the weather information from the ECMWF monthly forecast, skillful predictions of UK wind power, demand and electricity price were obtained for week 3 during the winter period over the years 2008 to 2014. Anomaly correlations in the range of 0.5-0.6 and CRPS skill scores of 0.10-0.16 were obtained for all three of these variables when comparing the subseasonal forecast with a forecast based on climatological weather in¬formation. The added value of using the subseasonal weather forecast information for a number of trading strategies was evaluated. A speculative trading strategy using the subseasonal weather forecast to value futures contracts demonstrated that positive re¬turns were achieved when systematically trading over 5 winters. When choosing the volume of power to buy in order to hedge retail demand risk, it was found that in some instances the subseasonal forecast outperforms the version using climatological weather to hedge the risk. The final conclusion is that skillful subseasonal forecasts of the meteorological variables exist and this skill propagates through to the energy system variables (wind power, demand and price) which should allow a range applications within the energy sector to potentially reduce risk.