Because of excellent superiorities, triple-electrode carbon nanotube sensor acts good in the detection of multi-component mixed gas. However, as one of the key factors affecting the accuracy of detection, the electrode separation of carbon nanotube gas sensor with triple-electrode structure is very difficult to decide. An optimization method is presented here to improve the mixed gas measurement accuracy. This method optimizes every separation between three electrodes of the carbon nanotube sensors in the sensor array when test the multi-component gas mixture. It collects the ionic current detected by sensor array composed of carbon nanotube sensors with different electrode separations, and creates the kernel partial least square regression (KPLSR) quantitative analysis model of detected gases. The optimum electrode separations come out when the root mean square error of prediction (RMSEP) of test samples reaches the minimum value. The gas mixtures of CO and NO<sub>2</sub> are measured using sensor array composed of two carbon nanotube sensor with different electrode separations. And every electrode separation of two sensors is optimized by above-mentioned method. The experimental results show that the proposed method selects the optimal distances between electrodes effectively, and achieves higher measurement accuracy.