To address the impact of extreme situations on the power and energy balance of the novel power system, a risk assessment method for power and energy balance of the novel power system based on integrating machine learning is proposed. An operation model of the novel power system is established, which includes the output model of new energy sources, the energy storage model, the generator set and the power flow model of transmission lines. Considering extreme situations mainly dominated by extreme weather, the extreme weather scenarios are defined, and the power and energy imbalance indicators of the novel power system under extreme situations are proposed. Anintegrating machine learning optimization (ILO) method is proposed to solve the power and energy imbalance. The two-stage ILO method includes an imitative learning (IL) stage for accurate strategy initialization and a reinforcement learning (RL) stage for efficient fine-tuning. The simulation results based on the PJM5-bus system and the improved IEEE118-bus system show that the proposed method can quantitatively define the power and energy imbalance, and improve the efficiency and stability of the risk assessment of power and energy balance.