Amid escalating environmental pressures and dynamic socio-economic changes, current research on ecological security often needs to provide comprehensive predictive models that effectively incorporate multivariate relationships and long-term trends. This study aims to fill this gap by employing the Pressure-State-Response(PSR) framework, utilizing 18 indicators derived from meteorological, remote sensing, soil, terrain, and socio-economic datasets to evaluate the ecological security of the West Liao River Basin from 2010 to 2021. A transformer-based artificial intelligence model was developed to predict time-series indicators from 2022 to 2070, enhancing the accuracy of trend, seasonality, multi-scale, and multivariate relationship predictions. Our findings reveal that the Ecological Security Index(ESI) remained within the “Generally secure” category, exhibiting a slightly declining trend with values ranging between 0.478 and 0.499. Key obstacle factors identified include the proportion of the non-agricultural population, power of agricultural machinery, effective irrigated area, GDP per capita, and the proportion of cultivated land to land area. Compared to state-of-the-art models such as Informer, LightTS, TimesNet, and Dlinear, our model demonstrates significant improvements in Mean Absolute Error(MAE) of 1.04%, 4.09%, 3.22%, and 4.54%, respectively. This research provides critical insights into the region’s management and enhancement of ecological security.