Most studies focus on reserve optimization by finding the optimal reserve allocated during power outages. The paper presented a different perspective in addressing the reserve allocation by forecasting the outage capacity to be used as reserve margin correction to minimum mandated reserve allocation by policy, considering the generation sector can exercise market power to force outage. The outage capacity forecasting model cleaves into encoding the outage capacity data into images using Gramian Angular Field (GAF), predicting the outage capacity from the set of images using a Long Short-Term Memory (LSTM) network, and using the forecasted outage capacity to correct the reserve margin. The image-based LSTM yields training and validation loss between 0.17 and 0.21, while the time series yields \(10^3\) . The image-based LSTM model yield testing performance of MAE \(\approx \) 0.94, MSE \(\approx \) 1.00, MAPE \(\le 10^{-3}\) , and RMSE \(\approx \) 1.00. Experimental results show our approach performs better in imaging than in time series. The reserve margin correction yield from the forecasted outage capacity substantially improves the reserve allocation to address peak demand during outage events in the generation sector. The severity of an outage juxtaposed a decrease in the RMI of the electricity market and an increase in the capacity of an outage, wherein \(RMI \approx 0\) as the reserve is low and capacity on outage is high.