With the rapid development of social economy, the scale of the grid gradually expanded, the power structure is becoming more complex and gradually from extensive management to intensive management development, unit maintenance problem size is also increasing rapidly. Unit maintenance issues related to the power grid output and operation mode, and have great impact on system reliability and economy, How to establish a scientific and rational model, effectively solution and choose the best as the focus of grid scheduling. Scientifically rational maintenance scheme has important role in enhance utilization ratio of equipment, gain higher economic benefit and guarantee the safe operation of the grid reliability.This paper takes Yunnan Power Grid Unit Maintenance as the study of the background; establish an equivalent reserve capacity model by means of heuristic search algorithm to solve. Driven by the deficiency of cannot guarantee the result is optimal using the fixed order of power plant to calculate, this paper proposes a model for optimizing maintenance based on Fork/Join parallel framework and fuzzy optimization algorithm. The main works are as follows:(1) Complex structure of the provincial power grid leading to a number of constraints involved in maintenance models, not only have the conventional repair constraint, but also have the cascade hydropower and inter-basin power station constraint. So use the principle of mathematical set to process these constraints makes the model easier to solve. This paper defined all the cascade power plant units as a whole involved in the calculation if they need to simultaneous overhaul. When a single station involved in the calculation, separate the combination unit.(2) Analysis and design the model for parallel, and realized multi-core parallel using the Fork/Join parallel framework for get the optimal solution.(3) Because there are many evaluation indexes, single indicator optimal cannot guarantee that other indicators are optimal or feasible.so establishing fuzzy optimization model for the feasible solution space with multiple solutions. Then solve the model and get the relative optimum membership degree of the scheme order, the front is the best scheme. Finally, using multi-core parallel fuzzy optimization algorithm, Yunnan Power Grid long-term unit maintenance optimization system is built. And then describes the overall system design and other key technologies. The actual operation shows that the system have many excellent characteristics, such as scientific and reasonable calculation, friendly interface and stronger practicability. |