Aiming at the issues that the field mill type electric field sensor lacks an accurate and complete mathematical model, and its signal is weak and contains a large amount of harmonic noise, on the basis of establishing the mathematical model of the sensor’s induction electrode, the finite element method was adopted to analyze the influence laws of parameters such as the thickness of the shielding electrode and the distance between the induction electrode and the shielding electrode on the sensor sensitivity. On this basis, the above parameters were optimized. A signal processing circuit incorporating a pre-integral transformation circuit, a differential amplification circuit, and a bias circuit was investigated, and a completed mathematical model of the input and output of the field mill type electric field sensor was established. An improved harmonic detection method combining fast Fourier transform and back propagation neural network (FFT-BP) was proposed, the learning rate, momentum factor, and excitation function jointly participated in the adjustment of the network, and the iterative search range of the algorithm was limited by the threshold interval, further improving the accuracy and rapidity of the sensor measurement. Experimental results indicate that within the simulated electric field intensity range of 0–20 kV/m in the laboratory, the measurement resolution of this system can reach 18.7 V/m, and the measurement linearity is more than 99%. The designed system is capable of measuring the atmospheric electric field intensity in real time, providing necessary data support for lightning monitoring and early warning.

Due to the difficulty of achieving absolutely precise coaxial installation of measurement components, a coaxial error will occur between the measurement axis of the sensor and that of the electronic compass when conducting the measurement of the airborne atmospheric electric field intensity based on the geographic coordinate system under the airborne mode of unmanned aerial vehicles. This error leads to less satisfactory measurement accuracy of the atmospheric electric field intensity components based on the geographic coordinate system. In this study, the angle between the measurement axis of the sensor and that of the compass is set as the parameter to be identified, and a modified model for three-dimensional electric field orientation decomposition based on geographic coordinates is constructed. In view of the characteristics of the nonlinear equations of this model, an algorithm based on tansig–Nonlinear Dynamic Inertia Weight Adaptive Particle Swarm Optimization (NDIWAPSO) is proposed to solve the modified model, successfully addressing the problem of insufficient measurement accuracy of the electric field intensity components in the geographic coordinate system caused by the coaxial error. The experimental results show that the parameters of the three-dimensional electric field orientation decomposition modified model can be accurately identified by the algorithm proposed in this paper, improving the measurement accuracy of the atmospheric electric field intensity components based on the geographic coordinate system and laying a necessary foundation for lightning warning.