Abstract Commissioning is an effective means to reduce the energy consumption of cooling systems in buildings. However, owing to the uncertainty of the load, the deviation between the true value of the load and the predicted value may cause a mismatch between the cooling load and the cooling capacity of the commissioning strategy, resulting in low robustness of the commissioning strategy. Therefore, a low-cost cooling system commissioning strategy, which can effectively quantify the load uncertainty and ensure the robustness of the strategy, is proposed in this study. A Quantile Regression Neural Network (QRNN) model is established to obtain the uncertainty range of the cooling load of a building in the form of a probability distribution. The low-cost commissioning method under each working condition with different partial load rates is obtained using an optimisation algorithm. An entropy–weight Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) multicriteria decision-making method is applied to optimise the commissioning strategy for the highest guarantee rate and lowest energy consumption. Through the case study of an existing office building located in Inner Mongolia, northern China, it was concluded that the proposed commissioning strategy can reduce the building’s average daily energy consumption by 7.75%. The results indicate that compared with the deterministic commissioning strategy, this robust commissioning strategy can achieve a higher guarantee rate under various load demands. In particular, on days with high load demand, a high guaranteed rate and low energy consumption can be achieved simultaneously.