The current domestic provincial power market transactions are still dominated by medium-term and long-term power trading, mainly including annual negotiation trading and monthly centralized trading. Monthly centralized trading is the last transaction window followed by power dispatch operations. The transaction and clearing mechanism design must fully consider power grid security and unit operation constraints in order to guarantee the feasibility of transaction results. Based on the design and practice of Guangdong monthly centralized trading, this paper concludes a method for determining and adjusting the maximum amount of electricity for various type of units, while considering the characteriscs of units such as combined heat and power units and gas power generation, the power plant load rate limit, and the minimum/maximum transaction limit of power plants. This paper also implements a series of mechanisms to deal with power plant group transaction limits and the impact of must-run units on centralized trading, including improved market clearing algorithm and must-run unit bidding price processing. The reasonable compatibility between the design of competitive transaction mechanism and the operational constraints of the grid has been achieved. Finally, the real case analysis of the operation of Guangdong power market proves the practicability of the proposed mechanism.