Research on a Hybrid Intelligent Method for Natural Gas Energy Metering

In this paper, a Comprehensive Diagram Method (CDM) for a Multi-Layer Perceptron Neuron Network (MLPNN) is proposed to realize natural gas energy metering using temperature, pressure, and the speed of sound from an ultrasonic flowmeter. Training and testing of the MLPNN model were performed on the basis of 1003 real data points describing the compression factors (Z-factors) and calorific values of the three main components of natural gas in Sichuan province, China. Moreover, 20 days of real tests were conducted to verify the measurements' accuracy and the adaptability of the new intelligent method. Based on the values of the Mean Relative Errors and the Root Mean Square errors for the learning and test errors calculated on the basis of the actual data, the best-quality MLP 3-5-1 network for the metering of Z-factors and the new CDM methods for the metering of calorific values were experimentally selected. The Bayesian regularized MLPNN (BR-MLPNN) 3-5-1 network showed that the Z-factors of natural gas have a maximum relative error of -0.44%, and the new CDM method revealed calorific values with a maximum relative error of 1.90%. In addition, three local tests revealed that the maximum relative error of the daily cumulative amount of natural gas energy was 2.39%.

Automated summary

In this paper, a Comprehensive Diagram Method (CDM) based on a Multi-Layer Perceptron Neuron Network (MLPNN) is proposed for natural gas energy metering using temperature, pressure, and the speed of sound from an ultrasonic flowmeter. The MLPNN model is trained and tested using real data points of compression factors (Z-factors) and calorific values of natural gas in Sichuan province, China. Experimental results show that the selected MLP 3-5-1 network for Z-factor metering and the new CDM method for calorific value metering have high accuracy and adaptability.