Single　neuron　PID　controller　has　self-learning　and　self-adaptive　abilities,　and　it＇s　widely　applied　to　industrial　control　instruments　exposed　to　all　kinds　of　disturbances.　However,　neural　networks　is　low　at　convergence　rate,　thus　is　difficult　to　meet　the　requirement　for　being　real-time.　Based　on　the　analysis　of　single　neuron　PID　control,　it　is　proposed　in　this　paper　that　an　algorithm　of　single　neuron　PID　adaptive　controller　based　on　expert.　According　to　characteristic　parameters　calculated　by　the　performance　identification　module,　this　new　algorithm　can　automatically　adjust　the　most　sensitive　gain　coefficient　K　of　single　neuron　by　importing　empirical　knowledge,　and　modify　the　output　of　single　neuron　PID　controller.　Then　a　sort　of　temperature　control　system　has　been　stimulated　using　the　tool　of　Microsoft　Visual　C++.　The　experiment　results　show　that　this　new　algorithm　has　satisfactory　static　and　dynamic　performances,　as　well　as　better　stability　and　faster　speed　of　response　than　the　conventional　single　neuron　PID　controller.　Furthermore,　this　new　algorithm　has　strong　robustness　and　adaptation　with　changing　plant　parameters.