There are a lot of data redundancy in wireless sensor networks. By compressing the original sampling data, the data fusion technology eliminates redundancies in data, reduces the amount of data sent by nodes effectively and prolongs lifetime of sensor networks. This paper proposed a data fusion algorithm that combined data forwarding and compressed sensing. During the process of collecting sampling data in sensor networks, the algorithm selects using compressed sensing to compress original sampling data or simply storing and forwarding sampling data according to the amount of nodes' child nodes. Simulations indicate that compared with the data fusion algorithm based on compressed sensing, the data fusion algorithm that combined data forwarding and compressed sensing achieved both network load balance and data compression effectively.