Research On Temperature Compensation Algorithm In Ultrasonic Heat Meter

With the deepening reformation of heating system, ultrasonic heat meter has played an important role in the charging of heating-metering. Ultrasonic heat meter calculates the heating flow by ultrasonic detection technology, then computes and displays the released heat of heating flow in the heat exchange system. Therefore, flow measurement is the key link that determines the metering accuracy of ultrasonic heat meter. However, in practical application, there is an error cannot be ignored in flow measurement which caused by the temperature of heating flow.Currently, the correction look-up table algorithm is widely used at home and abroad to improving the effect of temperature on flow measurement. But the error of flow measurement is still large after being compensated, and the accuracy of measurement is difficult to achieve the 2nd grade of standard meter. In order to solve it, this paper analyzed temperature influence on DN20 ultrasonic heat meter flow measurement, and then proposed a temperature compensation algorithm based on LMBP neural network to reduce the effect of temperature on flow measurement accuracy. The main contents of this paper are as follows:Firstly, according to the Heat Meter(CJ128-2007) industry standard, we developed the software on the DN20 household ultrasonic heat meter hardware designed by our research group, and some computing, real-time display and storage functions of temperature, flow, heat and other information were realized. Then we carried out flow measurement experiment. Verification experiment results showed that the accuracy of flow measurement was affected by temperature significantly, and the error of flow measurement was nonlinear.Secondly, for the strong error fluctuations of flow measurement affected by temperature and the nonlinear error of small diameter flow measurement, LMBP temperature compensation algorithm was designed and implemented in this paper. The training rule of LMBP temperature compensation algorithm consisted of steepest descent algorithm and Newton-Gauss algorithm, and it was used to learn the experimental data which included the information of flow metering. Then the nonlinear mapping relationship between temperature and flow measurement was built. Simulation results showed that LMBP temperature compensation algorithm was able to integrate and predict all experimental data, and had accurate generalization capability.Finally, in order to verify the practicality and stability of LMBP temperature compensation algorithm, we carried out verification test after transplanting the algorithm into ultrasonic heat meter. Experiment results showed that compared to the existing look-up table correction algorithm, LMBP temperature compensation algorithm can more accurately predict the magnitude of flow measurement under each temperature, and effectively inhibit the effect of temperature on flow measurement.Theoretical analysis and experimental studies had shown that the temperature compensation algorithm based on LMBP neural network, designed in this paper, performed stable compensation capacity. It reached the 2nd grade of standard meter with strong flow correction capability and high engineering value.