The Research Of The High-Speed Parallel Wavelet And Its Applications In Power System

With the scale of the power systems constantly expanding and the network structure complicating, the variety of studies in the large power systems may require the larger scale computations than before. The wavelet transforms are important tools in this area. Largely as a result of the complex calculations, slow velocity and serious time-consuming, they could not exploit their advantages to the practical engineering. Therefore, using the parallel technology to solve this problem is great significant.The current research on the parallelism has many aspects. The multi-core technology can not only calculate in parallel but also can avoid the redundancy costs by the communication and maintenance of the multi-machine network structure. The popularity of the multi-core platform also provides excellent conditions for its research. This paper uses the multi-core technology to achieve the parallelism of the wavelet and wavelet-packet transforms, so that the speed of calculations in the practical is increasing and the time is eliminating. The main work in this paper includes:1. The first thing is to introduce the parallel technology of the multi-core, the way of programming based on two environments-- Pthreads and OpenMP, and then to write the serial programs on the Visual studio 2008.2. That is applying the Intel Parallel Amplifier to find the hot parts of the programs and analyzing if there is any parallelism during the most time-consuming codes of the wavelet and wavelet-packet programs. This paper proposed one parallel strategy in task and anther in data for the convolution calculations and the data restoring.3. This paper realizes the parallel algorithms of the wavelet and the wavelet-packet based on the Pthreads and OpenMP respectively. The DB4 wavelet transforms is used in the experiments and the periodic extension method in the data pre-processing, the author make the test under five kinds of data level and on the platform of single machine with two cores. In the Pthreads, the process of the decomposition is parallel in the data-level and the reconstruction in the task-level. The parallel results meet the requirements by comparing parallel algorithms with the serial algorithms of the single-layer transform and multi-layer. In the OpenMP, the author takes the guidance statements to parallel the loops of the serial programs. In addition, this paper proposed one nesting and anther non-nesting parallel ways of the wavelet-packet and compared the parallel results with the serial. The tests show that the speedup is increasing with the amount of the data growing and finally closing to 2. The effect is obvious and it can improve the transform's speed significantly.4. This paper applies these parallel algorithms to the harmonic analysis and the data compression of the power system and proposed a parallel strategy for compression of the bitmap. The results also show that the parallel transforms of the wavelet and wavelet-packet can improve the analysis speed in power system significantly.