Locally stationary Energy Time Series (LETS)

It is difficult to think of any aspect of everyday life which does not rely in some way on energy supply and use. Behind every energy source is a complex network of stakeholders ensuring a reliable supply from generation through to distribution and use. In recent years, there has been an increasing focus on low carbon energy & renewables and also increasing marketisation, reorganisation and privatisation in the sector, particularly with large utilities.Time series analysis is a statistical cornerstone, of vital importance to many energy related challenges. For example, short-term wind speed forecasting is key for utilities aggregating many sources of supply, as is predicting the future energy use of groups of customers. Time series analysis is also critical to the planning of proposed wind farms to see if the predicted wind power is likely to be efficient and reliable. Over the last decade, the nature of time series encountered by stakeholders has changed. In the past, series were assumed to be stationary (i.e. that their statistical properties did not change over time). Much of what is now experienced is non-stationary. This becomes ever clearer as increasing flows of high-quality data enable new models to be proposed, studied and considered.Compare, for example, wind and gas-fired power. Wind is intermittent and not controllable. Gas powered stations, by comparison, are highly controllable and can produce almost constant power. Incorporating large quantities of wind power into the grid can be problematic as there can be sustained periods without wind, or periods of highly variable wind. Another issue is increasing marketisation: across Europe people are now able to purchase power from a variety of suppliers and modes of supply, distributors supply to different, fragmented parts of the market. Consequently, data collected on consumers or generators is less stable and much less stationary than in previous years.Our proposal addresses this new world of non-stationarity head-on. For several years our team has been at the forefront of developments in non-stationary time series: introducing new classes and using them in new and innovative ways. Our proposal will develop novel techniques to revolutionize the way that such time series are analyzed and hence be of considerable use to our industrial partners and the energy industry more widely. For example, we shall investigate and develop new methods for (i) handling more than one non-stationary series simultaneously; (ii) identifying appropriate sampling rates for series and whether any series have been compromised by inappropriate sampling rates; (iii) dealing with the common problem of data dropouts and irregularly spaced time series but still obtain meaningful insights; (iv) improved methods for forecasting and enabling predictions of one time series from another; (v) improving robust measures of uncertainty of our estimates. Even small improvements in any of these quantitative areas can lead to massive financial, environmental and reliability benefits of value to our partners and society more generally. We intend to create a step-change in the methods and procedures used by energy stakeholders by moving to the non-stationary world.

**Who will benefit?** This research programme will generate considerable impact for a wide range of academic and non-academic beneficiaries, principal among whom are: a) The Energy community including academia and industry; b) Our collaborating industrial and academic partners as listed in section 1 of the Case for Support; c) The statistical research community, particularly in the area of locally stationary methods; d) The project personnel: the PDRAs and PhD students; e) Recruiters of doctoral graduates in statistics; f) Society in general. **How will they benefit?** New techniques: (a, b, f) The research undertaken by this proposal will develop a number of exciting new statistical techniques which will be disseminated to our partners and the energy community more widely. Our methods are intended to result in more efficient and cost-effective ways of marshalling precious resources, by making better predictions and reducing the impact of intermittency. These benefits will flow through to Society in terms of a cheaper, cleaner, more sustainable and reliable energy supply. Targeted Knowledge Exchange: (b) Significant further benefit will accrue to beneficiary group (b) through their partnership on this project. The supporting letters from Shell and GL Garrad Hassan give examples of how understanding in the areas of aliasing and locally stationary time series are vital for the longer-term development of their businesses. Through workshops and engagement sessions with our partner organisations (EDF, Garrad Hassan, Nuon and Shell) we will seek to extend the impact of this research across the sector (c). Generic Knowledge Exchange: (c) Our intention is to develop methods which are of considerable interest to the academic community both in statistics and the energy fields. As well as the traditional routes of journal publication, workshops and conferences we shall develop open-source R software that embodies our techniques: these will benefit the academic community and beyond. Further, we shall utilize our advisory group to promulgate our techniques to a wider audience where appropriate and where they see fit, allowing us to plug-in to the international community through an innovative route. Developing good people: (d). The project personnel will themselves benefit from a supportive training, research and development environment, given the opportunity to create new techniques and see them employed in a productive and worthwhile setting, given the opportunity to work with industry on learning new challenges and a team-based approach to solving them, and be ideally positioned to seek future employment in a field/industry of such great importance to society. Contributing to the future supply of people: (all) This proposal will secure an increase in the number and quality of post-doctoral and PhD researchers in statistics with appropriate training to make an impact in the area of energy. In particular, the renewable energy sector is a comparatively new, highly instrumented industry. Data monitoring and modelling are all pervasive within this sector. Given its data-rich character it is perhaps surprising to read in the letters of support how few statistically trained researchers work in this area. As Jonathan (Shell) describes The number of suitably qualified doctoral and post-doctoral statisticians with the skills and passion in this area are relatively low , but as Landberg (GL Garrad Hassan) concludes we believe the proposal will go a long way to helping raise the profile of the renewable energy sector as an application area for statistics researchers within the UK. Beneficiaries (a,b,c,e) will consequently be able to recruit outstanding scientists equipped with the knowledge and skill-set to prosper in this sector. Beneficiary (c) will also benefit from the training of career young researchers in Statistics, a key STEM discipline which is in skills shortage.