Abstract: With the increasing penetration of renewable energy in power grids, the response performance of thermal power units in primary frequency regulation under frequency disturbances faces higher requirements. Traditional acquisition and control devices exhibit deficiencies in signal interference suppression and dynamic response accuracy, making it difficult to meet high-precision, real-time frequency regulation demands. To address this issue, an adaptive filtering method based on deep Fourier analysis is proposed, along with a novel energy acquisition and control device. The method achieves high-precision suppression of DC components, positive harmonics, and random spikes in power signals while maintaining dynamic response characteristics through multi-layer Fourier series decomposition, interference identification, and adaptive weight updating. The designed device integrates high-precision frequency and power acquisition modules, an embedded edge computing platform, dual hardware and software redundancy mechanisms, and human-machine interaction functions, ensuring continuity and reliability in signal acquisition and processing. Experimental results show that the energy of second- and higher-order harmonics is reduced by approximately 70% after filtering; the response time decreases from a mean of 1.45s to 0.85s, overshoot amplitude reduces to 6%, and signal homology is significantly enhanced.