In order to meet the requirements for automatic navigation operations and optimize the efficiency of agricultural machinery operations, a path planning algorithm formed by mixed rules was proposed based on simulated annealing method, after dealing with the constraints of various turning methods and agricultural plots. In terms of multi-constraint processing, the adjacency matrix of agricultural machinery turning cost was introduced to quantify the influence of different turning modes, and the block boundary and obstacle boundary were respectively dealt by using improved Douglas-Peucker fitting algorithm and the minimum convex hull of sampling points was solved. After the reserved area for turning were obtained by using the parallel migration of angle bisectors, the optimal crop rows of multi-shape parcels were generated in order to reduce the cost of turning. In traversal order of agricultural machinery, full coverage traversal order was formed by disassembling crop rows into units, solving the optimal units with simulated annealing method and assembling to a whole path. The algorithm solved the problem that traditional planning methods were hard to adapt to different conditions and classical simulated annealing method fell into local optimal solution easily in large scale planning. The experiment results showed that coverage of the paths obtained by proposed method was up to 90.78% in average and the average duty cycle was 85.10%. Under the same conditions, the proposed path can save up to 30.3% distance consumption compared with the traditional path, and 6.9% compared with the path formed by simulated annealing algorithm. The results showed that the proposed algorithm can plan the operation path for agricultural machinery under various constraints and reach a better planning effect.