A data science team at an energy analytics company is leveraging a neural network to enhance the precision of their predictive model, which forecasts electricity consumption based on diverse environmental and user interaction variables. Initially, the model was underperforming, prompting the team to increase its complexity by adding more layers in an attempt to capture intricate patterns within the data. However, post-modification, the team observes that the model's training accuracy struggles to converge, indicating a deterioration in learning efficiency. What is the most probable reason for this observed behavior in the neural network's performance?