Abstract: In order to meet the request of data fusion for greenhouse wireless sensor networks (WSN) system, an algorithm of multi-sensors data fusion in real-time was put forward. First, sensor data sequence was checked for consistency and then exponentially smoothed three times on sensor nodes; Then, the smoothed data was sent to the router or gateway, fusion by a power mean square, which works based on the proposed new-model support function without operating exponent in this paper. Algorithm was tested in the glass greenhouse of Jiangsu Agriculture Expo Garden, experimental results showed that cubic exponential smoothing can significantly reduce data volatility; data fusion precision of new-model support function and Gaussian support function had no significant difference, while actual running time of new algorithm was less 83.6% than the latter, and the fusion effect was better than the arithmetic average operator.