In　this　paper,　the　self-organizing　heuristic　dynamic　programming　algorithm　is　established　to　solve　the　approximate　optimal　control　issue　for　affine　nonlinear　systems.　A　self-organizing　neural　network　modeling　method　based　on　the　particle　swarm　optimization　algorithm　is　introduced　to　construct　the　model　network.　In　contrast　to　the　traditional　backpropagation　neural　network,　it　has　stronger　adaptive　ability　and　higher　modeling　precision　for　a　variety　of　different　complex　systems,　which　substantially　boosts　the　efficiency　of　the　method.　In　addition,　the　action　network　and　the　critic　network　are　constructed　to　obtain　the　approximate　optimal　control　strategy　and　the　optimal　cost　function,　respectively.　The　convergence　of　the　cost　function　is　proved.　It　also　proved　that　the　state　estimation　errors　and　the　weight　vector　estimation　errors　are　uniformly　ultimately　bounded.　Several　nonlinear　complex　systems　are　selected　in　the　experimental　simulation　to　prove　the　efficiency　of　the　method.