No abstract

No abstract

Multi-objective optimizations were conducted for a compressor station comprising two dissimilar compressor units driven by two dissimilar gas turbines, two coolers of different size, and two parallel pipeline sections to the next station. Genetic Algorithms were used in this optimization along with detailed models of the performance characteristics of gas turbines, compressors, aerial coolers, and downstream pipeline section. Essential in these models is the heat transfer between the gas and soil as it affects the pressure drop along the pipeline, and hence relates back to the coolers and compressor flow/pressure settings. Further investigative techniques were developed to refine the methodology as well as to minimize the downstream gas temperature at the suction of the next station. Current operating conditions at the station were compared to the optimized settings, showing that there is room for improving the efficiency of operation (i.e. lower energy consumption) with minimum effort on the station control strategy. Two threshold throughput conditions were determined in so far as single vs. multi-unit operations due to the dissimilarity in the compressor units and associated gas turbine drivers. The results showed that savings in the energy consumption in the order of 5–6% is achievable with slight adjustment to unit load sharing and coolers by-pass/fan speed selections. It appears that most of the savings (around 70–75%) are derived from optimizing the load sharing between the two parallel compressors, while the balance of the savings is realized from optimizing the aerial coolers settings. In particular, operating the aerial coolers at 50% fan speed (if permitted) could lead to substantial savings in electric energy consumption in some cases.

No abstract

This paper presents application of Genetic Algorithm (GA) methodologies to multi-objective optimization of two complex gas pipeline networks to achieve specific operational objectives. The first network contains 10 compressor stations resulting in 20 decision variables and an optimization space of 6.3 × 1029 cases. The second system contains 25 compressor stations resulting in 54 decision variables and an optimization space of 1.85 × 1078 cases. Compressor stations generally included multiple unit sites, where the compressor characteristics of each unit is taken into account constraining the solution by the surge and stonewall limits, maximum and minimum speeds and maximum power available. A key challenge to the optimization of such large systems is the number of constraints and associated penalty functions, selection of the GA operators such as crossover, mutation, selection criteria and elitism, as well as the population size and number of generations. The paper discusses the approach taken to arrive at optimal values for these parameters for large gas pipeline networks. Examples for two-objective optimizations, referred to as Pareto fronts, include maximum throughput and minimum fuel, as well as, minimum linepack and maximum throughput in typical linepack/throughput/fuel envelopes.