A powerful concept of grid computing is Resource sharing in the large level high performance computing network at world wide. Main focus of grid computing is perfect Fault tolerant Resource selection for a job then only receives the good result. Selecting the perfect Grid resource for task prediction is the major problem in the dynamic grid computing. In this paper, we have designed the new optimization algorithm presented and named as hybrid GRP-PSO. Hybrid GRP- PSO is combination of Particle Swarm optimization (PSO) algorithm and Grid Resource Prediction Pattern (GRP) from two or more grid networks. Since Particle Swarm Optimization algorithm is weak in Local search, Grid Resource Prediction pattern has been used to improve the quality and selecting the perfect Fault tolerant grid resource. The experimental results show the important of proposed system to select the perfect grid resource. Here the end User doesn't need the grid resource knowledge only submit task to the grid service. This proposed grid service (portal) will care of all knowledge about the perfect resource selection automatically with secure and efficient via.