Renewable energy power generation technology is becoming increasingly mature, and its share in the energy supply of building complexes is increasing year by year. However, renewable energy is random and volatile, and user loads are adaptive and diversified, posing new challenges to the operation of energy systems of building complex (ESBC). To address the above issues, this paper proposes a game-based operation strategy for energy systems of building complexes that takes into account source-load uncertainty and integrated demand response. First, based on the energy consumption characteristics of multiple types of building group users, an energy cascade utilization and user diversified load model is constructed. Then, user participation in the integrated demand response mechanism and comfort requirements are analyzed. Next, a distributed robust optimization method is used to address source load uncertainty issues. Third, we comprehensively analyze the interests of integrated energy operators, multi-energy suppliers, and building group users to establish a two-layer master-slave game model for the ESBC. Finally, we employed an improved genetic algorithm (GA) combined with CPLEX to solve the model. Using a northern industrial park as a case study, we achieved a 15.59% reduction in carbon emissions, a 21.44% decrease in carbon emission costs, and a 2.13% improvement in user comfort. This demonstrates that the proposed strategy enhances the environmental sustainability of energy system operations while fully addressing the interests of multiple stakeholders.

Uniform indoor lighting is essential for enhancing visual comfort, reducing glare and shadows, and improving occupant productivity. However, achieving consistent illumination across multiple zones is challenging due to the dynamic interplay between daylight and artificial lighting, as well as the mutual influence among distributed lighting units. This study proposes a novel multi-zone uniform lighting control strategy based on a modular and scalable distributed framework. The system integrates both daylight and electric lighting control: a fuzzy logic controller regulates venetian blinds in response to changing daylight conditions, while a deterministic policy iteration algorithm, built upon an improved Extreme Learning Machine (ELM), manages artificial lighting. To enhance the generalization and adaptability of the ELM, the Sparrow Search Algorithm (SSA) is employed for hyperparameter optimization. A prototype system incorporating multiple luminaires and shading devices was developed and tested under real-world dynamic lighting conditions. Experimental results demonstrate that the proposed method effectively maintains uniform illumination across multiple areas, with a maximum relative error of less than 10% between actual and target illumination levels. The system exhibits strong real-time performance, stability, and adaptability to environmental changes. This research provides a practical, learning-based control solution suitable for complex indoor environments, and offers additional insights into the design of intelligent lighting systems for modern buildings.