Long term climate variability increases the uncertainty of renewable energy output fluctuations in future power systems with high renewable penetration

leading to risks of sustained supply demand imbalances.Addressing the current insufficiency in grid design practices where climate factors are rarely incorporated

this study first develops a long short term memory (LSTM) neural network model to capture the response characteristics of renewable energy to meteorological conditions

utilizing historical generation data and high resolution weather datasets.On this basis

a mathematical optimization model tailored for regional hybrid alternating current (AC)/direct current (DC) grid design is constructed to evaluate transmission expansion strategies over multi year time horizons.Furthermore

climate scenario data from the Coupled Model Intercomparison Project Phase 6 (CMIP6) are employed to test the methodology through case studiesThe results show that under low emission scenarios

the demand for new transmission corridors to maintain supply demand balance is relatively limited.In contrast

high emission scenarios require more intra and inter provincial transmission expansion to alleviate renewable energy shortfalls during certain critical periods.