World experience shows that important factor in the calculations for natural gas consumption between suppliers and consumers is not only the volume of natural gas, but the quality indicators. With gas market liberalization, gas properties are expected to vary more frequently and strongly (composition, heating value etc.). Quality of natural gas is currently a topical issue, considering the steady increase of gas consumption in the world in recent decades. Existent chromatographs and calorimeters are very accurate in gas quality determination, but general expenditure and maintenance costs are still considerable. Market demands alternative lower cost methods of natural gas quality determination for transparent energy billing and technological process control.
Investigation results indicate that heating value (HV) is a nonlinear function of such parameters as sound velocity in gas, N2 and CO2 concentration. Those parameters show strong correlation with natural gas properties of interest (HV, density, Wobbe index), during analysis conducted on natural gas sample database. For solving nonlinear multivariable approximation task of HV determination, artificial neural networks were used. Proposed approach allowed excluding N2 concentration from input parameters with maintenance of sufficient accuracy of HV determination equal to 3.7% (with consideration of N2 concentration – 2.4%) on sample database. For validating of received results corresponding experimental investigation was conducted with reference analysis of physical and chemical parameters of natural gas samples by gas chromatography and followed superior HV calculation according to ISO 6976:1995. Developed experimental setup consist of measuring chamber with ultrasonic transducer, reflector, pressure, temperature and humidity sensors, ultrasonic inspection equipment for sound velocity measurements and CO2 concentration sensor with relevant instrument. The experimental setup allows measurement of sound velocity at 1MHz frequency and CO2 concentration in natural gas sample along with parameters control (temperature, humidity, pressure). The HV calculation algorithm was based on specially designed and trained artificial neural networks.
Experimental investigation of proposed approach was conducted on 40 real samples of locally distributed natural gas. Obtained results, in comparison to reference values, showed absolute error in Lower HV (net calorific value) determination equal 166 kJ/m3, while relative error was equal 4.66%. Developed technology allows construction of autonomous instrument for instant natural gas quality determination, which can be combined with volume meters in order to provide transparent energy flow measurement and billing for gas consumers. Additionally it can be used for gas sensitive technological process control.