Search

Predicting the electricity demand of new supermarkets will help with design, planning, and future energy management. Instead of creating complex site-specific thermal engineering models, simplified statistical energy prediction models as we propose can be useful to energy managers. We have designed and implemented a data-driven method to predict the ‘electricity daily load profile’ (EDLP) for new stores. Our preliminary work exploits a data-set of hourly electricity meter readings for 196 UK supermarkets from 2012 to 2015. Our method combines the most similar stores on a feature space (floor area split by usage such as general merchandise, food retail and offices and geographical location) to obtain a prediction of the EDLP of a new store. Computational experiments were performed separately for subsets of supermarkets that consume only electricity, both electricity and gas, and by season. The best results were obtained when predicting Summer EDLPs with stores using electricity only. In this case, the average Manhattan difference and the percentage difference are 234 kWh and 16%, respectively. We aim to develop an application tool for supermarket energy managers to automatically generate EDLP for potential new stores. EPSRC Impact Acceleration Account - University of Oxford EP/R511742/1

Predicting the electricity consumption of proposed new supermarkets is helpful to design and plan future energy management. Instead of creating complex site-specific thermal engineering models, data-driven energy prediction models can be useful to energy managers. We have designed and implemented a data-driven method to predict the future ’electricity daily load profile’ (EDLP) of new supermarkets using historical EDLPs of existing supermarkets of the same type. The supermarket features used for the prediction are 10 types of floor areas divided by usage ( m 2 ) and its location. Four data-driven regression models are used and compared to predict EDLPs: Artificial Neural Networks, Support Vector Machines, k-Nearest Neighbours and OLS. Prediction computational experiments were performed over 1-h electricity readings of 213 UK supermarkets gathered during six years. Prediction error mainly varies between 12 and 20% depending on method, year, supermarket type, and division of the data (season or temperature intervals). EDLPs computed over warm periods are better predicted than over cold periods and supermarkets only with electricity are better predicted than supermarkets with electricity and gas. The three features with more weight in the prediction are Food, Chilled produce and Cafeteria areas. The limitations of machine learning methods to solve this problem are discussed.

By using smart meters, more data about how businesses use energy is becoming available to energy retailers (providers). This is enabling innovation in the structure and type of tariffs on offer in the energy market. We have applied Artificial Neural Networks, Support Vector Machines, and Naive Bayesian Classifiers to a data set of the electrical power use by 12,000 businesses (in 44 sectors) to investigate predicting which businesses will gain or lose by switching between tariffs (a two-classes problem). We have used only three features of each company: their business sector, load profile category, and mean power use. We are particularly interested in the switch between a static tariff (fixed price or timeof-use) and a dynamic tariff (half-hourly pricing). We have extended the two-classes problem to include a price elasticity factor (a three-classes problem). We show how the classification error for the two- and three-classes problems varies with the amount of available data. Furthermore, we used Ordinary Least Squares and Support Vector Regression models to compute the exact values of the amount gained or lost by a business if it switched tariff types. Our analysis suggests that the machine learning classifiers required less data to reach useful performance levels than the regression models. 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://

The HiPerDNO project is a three-year EU-funded research projects which studies novel applications for Distribution Network Operations (DNOs). These applications (distributed state estimation, power restoration, monitoring of assets) emerge from the large-scale deployment of sensors and instrumentation devices, responsive loads and embedded generation. Their introduction requires HPC strategy. A HiPerDNO HPC platform, which takes into account DNOs' and applications requirements, has been built and several applications were developed and tested on it. In this paper we report on the results achieved.