Multi-scale analysis of the energy performance of supermarkets

The retail sector accounts for more than 3% of the total electricity consumption in the UK and approximately 1% of total UK CO2 emissions. The overarching aim of this project was to understand the energy consumption of the Tesco estate (the market leader), identify best practice, and find ways to identify opportunities for energy reduction. The literature review of this work covered the topic of energy consumption in the retail sector, and reviewed benchmarks for this type of buildings from the UK, Europe and the US. Related data analysis techniques used in the industry or presented in the literature were also reviewed. This revealed that there are many different analysis and forecasting techniques available, and that they fall into two different categories: techniques that require past energy consumption data in order to calculate the future consumption, such as statistical regression, and techniques that are able to estimate the energy consumption of buildings, based on the specific building’s characteristics, such as thermal simulation models. These are usually used for new buildings, but they could also be used in benchmarking exercises, in order to achieve best practice guides. Gaps in the industry knowledge were identified, and it was suggested that better analytical tools would enable the industry to create more accurate energy budgets for the year ahead leading to better operating margins. Benchmarks for the organisation’s buildings were calculated. Retail buildings in the Tesco estate were found to have electrical intensity values between 230 kWh/m2 and 2000 kWh/m2 per year. Still the average electrical intensity of these buildings in 2010-11 was found to be less than the calculated UK average of the 2006-07 period. The effect of weather on gas and electricity consumption was investigated, and was found to be significant (p<0.001). There was an effect related to the day-of-the-week, but this was found to be more related to the sales volume on those days. Sales volume was a proxy that was used to represent the number of customers walking through the stores. The built date of the building was also considered to be an interesting factor, as the building regulations changed significantly throughout the years and the sponsor did not usually carry out any fabric work when refurbishing the stores. User behaviour was also identified as an important factor that needed to be investigated further, relating to both how the staff perceives and manages the energy consumption in their work environment, as well as how the customers use the refrigeration equipment. Following a statistical analysis, significant factors were determined and used to create multiple linear regression models for electricity and gas demands in hypermarkets. Significant factors included the sales floor area of the store, the stock composition, and a factor representing the thermo-physical characteristics of the envelope. Two of the key findings are the statistical significance of operational usage factors, represented by volume of sales, on annual electricity demand and the absence of any statistically significant operational or weather related factors on annual gas demand. The results suggest that by knowing as little as four characteristics of a food retail store (size of sales area, sales volume, product mix, year of construction) one can confidently calculate its annual electricity demands (R2=0.75, p<0.001). Similarly by knowing the size of the sales area, product mix, ceiling height and number of floors, one can calculate the annual gas demands (R2=0.5, p<0.001). Using the models created, along with the actual energy consumption of stores, stores that are not as energy efficient as expected can be isolated and investigated further in order to understand the reason for poor energy performance. Refrigeration data from 10 stores were investigated, including data such as the electricity consumption of the pack, outside air temperature, discharge and suction pressure, as well as percentage of refrigerant gas in the receiver. Data mining methods (regression and Fourier transforms) were employed to remove known operational patterns (e.g. defrost cycles) and seasonal variations. Events that have had an effect on the electricity consumption of the system were highlighted and faults that had been identified by the existing methodology were filtered out. The resulting dataset was then analysed further to understand the events that increase the electricity demand of the systems in order to create an automatic identification method. The cases analysed demonstrated that the method presented could form part of a more advanced automatic fault detection solution; potential faults were difficult to identify in the original electricity dataset. However, treating the data with the method designed as part of this work has made it simpler to identify potential faults, and isolate probable causes. It was also shown that by monitoring the suction pressure of the packs, alongside the compressor run-times, one could identify further opportunities for electricity consumption reduction.