In the service Internet environment, different requirements may be presented simultaneously by different users. It is still a great challenge to structure composite services quickly and effectively that fulfills personalized requirements. This paper present a personalized service customization method which is based on patterns. The method first identifies requirement patterns and service patterns from historical service request and service solutions, establish the matching relationship between the requirement patterns and service patterns. Then an algorithm called LPSC is proposed to deal with personalized requirements. LPSC first classify personalized requirements according to the similarity between requirements and constructing virtual requirements to replace them. Then each virtual requirement can be covered by a limited number of requirement patterns. Through the matching relationship which is established between requirement patterns and service patterns, we can find optimal service pattern set that suit the requirement patterns. Finally, service solutions can be obtained by the composition of service patterns. Experimental results prove the effectiveness of the proposed algorithm.