A new method which schemes of conditional data processing named standard deviation (SD) method is present and evaluated for identifying spectral data of gas-solid particle flow based on laser-induced plasma spectroscopy. The SD method is also compared with two conditional analysis methods called the SNR method and the absolute intensity method. First, the effect of three methods identifying the spectral data of the same fly ash sample is compared. Then, the threshold stability of the three methods under different conditions was observed by screening the spectral data of a set of 12 coal samples. The rejection rate, error rejection rate (or error rejection) and missing rejection rate (or missing rejection) were used to evaluate these three different approaches. And probability of false hits is determined by evaluating various conditional processing thresholds. The characteristic peaks used for the analysis of fly ash and coal samples are selected as Si 288.16 nm and C 247.86 nm, respectively. The result shows that for all approaches the SD method has the minimum rejection rate and error rejection (both 0%) when the SD threshold value is 120, the total rejection rate of the SD method is also the lowest of the three under this condition, which is 37.67%. It means that true data and false data could be completely and accurately identified by using the SD method. Moreover, for different samples and experimental conditions, the thresholds of absolute intensity method and SNR method fluctuate in the range of 83 - 135 counts and 2.05 - 4.00, respectively, while the threshold of the SD method has been stable at 55. Compared with other two methods, the SD method is more universal when faced with variable detection conditions. So it has more advantages in the rapid detection and analysis of power plants.