Electricity generation presents distinctive modeling challenges due to the absence of storage, instantaneous demandsupply balancing requirements, and heterogeneous generation technologies with different cost structures. This paper addresses these challenges by incorporating a base--peak load structure into large-scale computable general equilibrium (CGE) models, offering a middle ground between detailed energy system models and multisectoral global economic frameworks. We first develop a transparent toy model inspired by \textcite{bachmann2022} to demonstrate that first-order approximations of cascading effects, following Hulten's theorem, are inadequate when shocks are large and elasticities of substitution are low. Building on the theoretical insights, we embed a base--peak structure into the MIRAGE CGE model, treating electricity as a Leontief production function between base load generation (coal, nuclear, hydro, and part of renewables) and peak load generation (gas, oil, and peak renewables). This refinement captures the merit order dispatch mechanism and bottleneck effects when peak generation is constrained. We apply the enhanced model to assess the 2022 Russian gas shock in Germany and the European Union. Our results demonstrate that the base--peak structure more accurately reproduces observed macroeconomic impacts compared to standard electricity representations, with significantly larger GDP and welfare losses particularly affecting energyintensive industrial sectors, and less possible substitution from variable renewable energies. Theoretically, we show that third-order effects become important under conditions we explicitly identify, complementing recent findings on shock amplification in production networks. For policy, our findings highlight two key levers for responding to energy shocks: supply flexibility through storage and grid interconnection, and demand smoothing through dynamic pricing and interruptible contracts. The paper contributes methodologically by demonstrating how simplified yet realistic electricity representations can be integrated into global CGE frameworks without sacrificing the broader economic feedback mechanisms essential for policy analysis.