This item is available under a Creative Commons License for non-commercial use only
Civil engineering, Municipal and structural engineering
This research is centred on three pillars of EU energy policy that aim to improve: 1) energy efficiency, in order to reduce CO2 emissions and therefore limit climate change; 2) security of energy supplies, in order to protect economic output and vulnerable citizens in extreme weather; and 3) market integration, in order to increase energy supplier competition and consumer choice in each member state. To help deliver on these policies, the EU has recently mandated that: 1) gas smartmeters are to be provided to consumers to help improve energy efficiency; 2) network operators ensure adequate gas supplies during extreme cold weather; and 3) network operators provide energy suppliers with forecasts of the volume of gas they should purchase each day in wholesale markets in order to limit the risk to suppliers when entering new markets. Gas Networks Ireland has part-funded this research and has provided smart-metering and network gas consumption data, so that bottom-up and top-down models of gas consumption can be developed to assist with these EU requirements. Bottom-up models can be used to assess building energy efficiency and to forecast the daily volume of gas to be purchased by an energy supplier for its consumer portfolio. Top-down models can be used to forecast peak-day consumption on the network during extreme weather, and to improve the accuracy of bottom-up portfolio forecasts. This research develops such models using both ordinary and non-linear least squares (OLS and NLS) regression modelling methods. Each of the resulting models is either based on or develops upon standard heating degree day (HDD) theory used to model iii building heating system fuel consumption. It is shown that HDDs are used as an explanatory variable in linear regression models of building gas consumption and that these models can be used to infer building energy performance. This is used as a basis on which to develop a new energy efficiency benchmarking tool for domestic dwellings. This tool is for the use of energy suppliers who must assist their consumers in making energy savings. It is also shown that the HDD approach can be extended to include other variables such as wind speed and solar radiation. This is used as a basis to develop adapted HDD variables to improve estimates of daily gas consumption of individual buildings and of the Irish domestic and SME gas market. These variables are used to develop improved models for bottom-up portfolio and peak-day network forecasting. The development of the new benchmarking tool is based on the availability of gas smart-metering and household survey data for a sample of dwellings. It is shown that these data allow each parameter of a HDD linear regression model to be estimated using non-linear regression methods rather than the traditional ‘trial and error’ methods applied to monthly or longer fuel consumption data. This improved method is used to estimate HDD models for the dwelling sample and the resulting distribution of independent parameters are presented. These parameter distributions are then characterised by multinomial logistic regression (MLR) models using descriptive household variables. These MLR models are then used to demonstrate a new energy efficiency benchmarking method by comparing the inferred energy end-use of similar buildings.
The NLS regression modelling method is also used to develop an adapted HDD variable to improve estimates of total daily domestic and SME gas market consumption. The resulting model is based on the availability of recent market consumption data and accounts for numerous effects on gas consumption in addition to those currently estimated by the HDD variable. The improvement in modelling accuracy is quantified by applying a comparative analysis for each of the additional effects accounted for by the new adapted HDD variable. It is found that solar radiation significantly affects gas consumption and should be considered in market consumption models. The new model is used to predict year-ahead peak-day market consumption to alternative supply standards.
Finally, the research develops new models of daily gas consumption for individual consumers based on smart-metering data. These models are developed using SME smart-metering data. This is challenging because their consumption is unpredictable relative to domestic consumers, leading to forecasting difficulties for network operators and energy suppliers. Two modelling options are investigated: one that applies an adapted HDD variable (similar to that referred to above) to estimate the daily gas consumption of individual enterprises using the NLS method; and a second that applies the same market consumption estimator to each enterprises using the OLS method. It is found that OLS models are the most suitable for individual consumer forecasting in terms of the practicality of their implementation and accuracy of their forecasts.
Oliver, R. (2016) Statistical methods of domestic and SME daily gas consumption - applications to Gas Network Planning and Management. Doctoral Thesis, sTechnological University Dublin. doi:10.21427/D78S3R