To address the problems of the traditional grey wolf optimizer (GWO), such as being prone to trapping in local optima and slow convergence speed when dealing with the high-dimensional, nonlinear and strongly coupled characteristics in the load optimal dispatch of combined heat and power (CHP) systems, this study proposes a chaotic multi-layer grey wolf optimizer (CML-GWO). The core innovation of the proposed algorithm lies in two aspects: first, chaotic mapping is introduced to initialize the population, which effectively improves the uniformity of initial search and prevents the algorithm from falling into local optima at the early stage; second, a hierarchical guidance mechanism is integrated to balance the global exploration and local exploitation capabilities, thereby solving the slow convergence problem caused by the capability imbalance in the traditional GWO. Before applying it to the CHP system load optimal dispatch, the performance of CML-GWO is verified through the CEC2017 test function set. The results show that compared with the traditional GWO, the CML-GWO exhibits better robustness and optimization accuracy in complex multi-modal and composite function scenarios, which lays a solid foundation for its engineering application. For practical verification, four units of a thermal power plant are taken as the research object, and two multi-objective scenarios and two regulation modes are designed. The multi-objective scenarios include “minimum coal consumption-maximum renewable energy accommodation” and “maximum profit-maximum renewable energy accommodation”, while the regulation modes are “practical constraints” and “free whole-plant load distribution”. The verification results demonstrate that under the practical constraint scenario, compared with the traditional GWO, the average hourly coal consumption of the CML-GWO is reduced by more than 2 tons, the average hourly profit is increased by more than 14 yuan, the renewable energy accommodation capacity is increased by more than 20 MW, and all load deviations meet the requirements of safe operation. Under the free load distribution scenario of the whole plant, the optimization potential of the algorithm is fully released: the daily coal saving reaches 23.8 tons or the daily income increases by 6515 yuan, and the renewable energy accommodation capacity is improved by 0.74%~0.85%. Overall, this study realizes the multi-objective collaborative optimization of economic, energy and environmental benefits of the CHP system. The comprehensive performance of the CML-GWO in both numerical tests and engineering applications fully verifies its significant engineering practical value, providing a new effective optimization method for the load optimal dispatch of CHP systems under the background of high-proportion new energy integration.