PAVAN is an atmospheric dispersion program for evaluating design basis accidental releases of radioactive materials from nuclear power plant. It was developed by Pacific Northwest Laboratory on the basic of the atmospheric dispersion models described in RG 1.145 by NRC (U.S. Nuclear Regulatory Commission). Using the joint frequency of wind direction, wind speed and atmospheric stability, the atmospheric relative concentration values for the exclusion area boundary and outer LPZ boundary of nuclear power plant are calculated and given by the program. Once the program was introduced, it has been widely used in the radioactive accident consequence assessment, especially in the FSAR (Final Safety Analysis Report) and Report of EIA (Environmental Impact Assessment) of NPPs in China.
The theory basis and general method of PAVAN is introduced in this paper. And specialty of the X/Q points based on joint frequency data is discussed. The envelope algorithm of PAVAN is also introduced and discussed.
The paper presents an improved algorithm based on PAVAN which uses the hourly meteorological data as input instead of joint frequency data. In this algorithm, the size of X/Q points is related to the quantity of the hourly meteorological data. When the quantity is large enough, e.g. 17520 sets of hourly meteorological data in two years, the envelope curve for X/Q points fit more exactly than PAVAN.
Using the observed meteorological data, the improved algorithm is compared with PAVAN. The result proves that the former is more accurate. In general, the improved algorithm is relatively conservative. In some situation, the conservativeness is not certain. The factors which result in the uncertainty are deeply discussed.
Further optimized are performed by the algorithm. The number of points to seek in envelope curve fitting is set to be dynamic and be a quarter of total number of X/Q points to be fitted. The result shows that increasing the number of points to seek in the iteration process of envelope curve fitting will lead to more conservative X/Q values.
Additionally, the optimized algorithm provides X/Q value of 50% probability level for overall site. The value is not relatively conservative. From the standpoint of statistical probability, it is more realistic and is acceptable for potential accident consequence assessment. Especially, when X/Q value of 95% probability level for overall site is too conservative to accept, the value of 50% probability level can be used to replace the conservative value.