The energy-saving dispatch is becoming one of the most important smart grid components in China, which asks for both generation and demand side resources’ participation. In smart electricity grid, the uncertainties related with variable renewable energy outputs on the generation side, as well as the deviations between the scheduled demand response (DR) resources and their actual availabilities on the demand side should be simultaneously taken care of. In this paper, a new energy-saving dispatch problem is studied considering the uncertainties from both generation and demand sides. A bi-level optimization model is established to address the generation and demand side’s interactions considering economic dispatch of thermal units and load control schemes of end users in the new energy-saving dispatch problem. The upper-level minimizes the electric transmission company’s power generation and carbon emission costs while maintaining the system reliability, and the lower-level minimizes the expected total load dispatch cost paid by the electric distribution company to DR providers and the demand side emission cost. An iterative algorithm is adopted to solve the proposed bi-level model. In addition, improved NSGA-II method is used to solve the lower-level multi-objective optimization problem for seeking optimal compromise solution. Numerical case studies on the practical energy-saving dispatch problem of Guangdong Province in China illustrate the effectiveness of the proposed approach.