Thermal network model-based data assimilation for estimation of temperature characteristic coefficients for Peltier device

Abstract

To achieve precise temperature control, Peltier devices, which are also called thermoelectric coolers, are widely used. However, simulating the unsteady temperature history caused by the Peltier devices is difficult because the amounts of heat absorption and generation are affected by their temperature. The temperature dependence reportedly can be calculated using the quadratic equation of temperature and the typical temperature characteristic coefficients. However, as the temperature dependence varied among the manufacturers, the temperature characteristic coefficients had to be modified for each of the devices. We developed a technique to determine the temperature characteristic coefficients of the Peltier device automatically by using data assimilation. We integrated the particle filter, one of the data assimilation algorithms, into the thermal network method and enabled estimating the suitable temperature characteristic coefficients. To demonstrate the estimation, we evaluated a Peltier device. The constant current 1.0 A and its inverse current were applied to the sample device repeatedly, and the temperature of the control object fluctuated repeatedly between 40 °C and 90 °C was measured. The temperature change was simulated using the thermal network method with the typical temperature characteristic coefficients and the history was compared with the measurement results. The root mean square error of temperature between the measurement and the calculation results was 3.20 K. Then, we estimated the applicable value of the temperature characteristic coefficients by applying the particle filter combined with the thermal network model. When the estimated coefficients were applied to the thermal network model, the root mean square error of temperature decreased to 1.39 K.