Statistical analysis of flow rate for the design optimisation of a proposed CO2 pipeline network

Abstract

Coal power plants produce about 38% of the electricity around the world and are a major source of CO2. The CO2 emitted by these plants has to be captured and transported to a storage site as part of the CO2 mitigation strategy defined by Carbon Capture and Storage (CCS). The most economical method for long-term onshore CO2 transport is by pipeline, and it is recognised that the CO2 flow rate assumed for pipeline design has an impact on transport costs. Therefore, it is important to investigate the impact of the variability of flow rate in a CO2 pipeline network on its design and economics. CO2 flow rate is a strong function of power plant load. The latter is affected by seasonal and diurnal changes, rarer events such as a global recession, variations in coal prices and climate change, which can impact the demand. Besides plant load, the instantaneous CO2 flow rate is also a function of capture rate and emission intensity. The combination of all these factors makes power plant load behaviour complex and random, making it a candidate for stochastic treatment. Several studies have been conducted for CO2 pipeline optimisation but none of these investigate the randomness in behaviour of CO2 flow rate. Therefore, this paper presents a stochastic analysis of the flow rates used for the optimisation of the design of a proposed CO2 pipeline network. The CO2 sources considered for this analysis are the black coal-fired power plants located in NSW, Australia. These plants use bituminous coal as fuel, which has a high HHV/LHV, i.e. the amount of heat released by unit mass or volume of fuel (initially at 25 °C) once it is combusted and the products have returned to a temperature of 25 °C. Therefore, these power plants are good candidates for CCS, having high mitigation potential. The analysis in this work is applicable to the expected range of CO2 flow rates around the world, as the CO2 sources have a rated capacity ranging from 1,320 MW to 3,000 MW, equivalent to CO2 flow rates of 1 to 20 Mtpa. The load data is investigated for a period of two consecutive years, spaced over 5-minute intervals, to observe the trends in the flow behaviour such as seasonality and diurnal patterns. For real cases, the CO2 flow is expected to change with time as the power plant load varies, which necessitates a dynamic flow analysis. However, if the variation in load is gradual, the dynamic state can also be treated as quasi-steady. Therefore, this analysis is conducted assuming quasi-steady state conditions. The approach used involves fitting multi-modal probability models to the distribution of flow rate of CO2 captured from the power plant. Three distributions types are evaluated; Normal, Logistic-normal and Gamma, which are used to model the cumulative probability distribution of the flow in terms of time. The load data are compared using goodness of fit measures to observe their conformance with analytical probability distribution behaviour. The probability distribution with the best goodness of fit to the actual plant data is then chosen for predicting the percentage of time the flow rate takes a specific range of values, and this information is used in the design of the CO2 pipeline network. The probability distribution of flow rates is then used to determine the probability distribution of the operating costs of the pipeline network. Depending on the flow rate chosen for pipeline design, the pipeline may be over or under designed over a given time period and this impacts the annual operating costs. For example, if the average flow rate is chosen as the design flow rate, then the pipeline will be under designed for the times when flow rate is higher, and vice versa. During the time when operating flow rate exceeds the design flow rate, the operating costs will be higher than estimated. However, there is a trade-off between the operating costs and the time for which the pipeline is under-utilized (over-designed) or under-designed. This analysis helps to identify the conditions under which is it more suitable to under or over design the pipeline, given the probability distribution.