Aimed at the voltage-sensitive problem of EC-5 sensor which was suitable for WSN soil moisture acquisition nodes, the sources of detection error was analyzed according to the sensor working principle. It was found that the error would increase with the soil moisture when sensor parameters were determined. Using CC2430 platform as WSN information transmission node, WSN soil moisture acquisition system with multiple nodes based on controller and host computer was established. Taking node voltage u and the real soil moisture θ as parameters, the inverse modeling of nonlinear soil moisture sensor system was performed by neural network, and the compensation of EC-5 sensor detection error caused by supply voltage was implemented in host computer. Experimental results showed that this method could sharply reduce the influence caused by node voltage variation and improve the detection precision. 