Abstract: With the continuous decline in the share of synchronous units in receiving-end power grids with high-proportion renewable-energy, the inherent frequency-regulation capability of power systems is continuously weakened, leading to increasingly severe frequency-security risks. To investigate the evolution laws of frequency-regulation capability and frequency security region (FSR) for receiving-end power grids, this paper proposes a joint analysis method based on small-sample time-series data for the frequency-regulation capability and FSR evolution of high-proportion renewable-energy receiving-end power grids. Firstly, a multi-machine aggregated frequency response model for the receiving-end power grid is established. The system equivalent inertia and primary frequency-regulation coefficient are extracted as key parameters to quantify the system’s frequency-regulation capability. Analytical expressions of frequency stability indices under anticipated disturbances are derived. On this basis, critical parameter sets are determined subject to frequency security constraints, and the FSR boundary of the receiving-end power grid under anticipated disturbances is delineated. Secondly, a fractional-order multivariate grey model (F-MGM) is adopted to jointly predict the equivalent inertia and primary frequency-regulation coefficient of the power grid for future years, and its prediction accuracy is evaluated. Meanwhile, the multilayer perceptron (MLP) is combined with the Koopman operator to learn the evolution law of security-region boundaries from historical boundary data, so as to realize the interannual time-series prediction of the FSR boundary. Furthermore, the future frequency security margin can be quantified by comparing the predicted frequency-regulation-capability parameters with the security-region boundary. Finally, the proposed method is verified using the actual historical data of the East China Power Grid from 2021 to 2026 and its planning data for 2027—2028, and the future frequency-regulation capability and FSR of the power grid are deduced. The results demonstrate that F-MGM outperforms conventional models in the prediction of frequency-regulation-capability parameter sets. For FSR boundary prediction, the MLP-Koopman method can capture the long-term evolution characteristics of FSR boundaries more accurately than the pure MLP approach. In addition, this method reveals the physical connotation underlying FSR evolution from the perspective of system dynamics: under identical conditions, the required frequency-regulation and inertia-support capabilities of the East China Power Grid rise continuously over time, and the system’s frequency-security operation margin exhibits year-by-year attenuation characteristics.