Real-time monitoring of satellite status and anomaly detection are conducive to ensuring the safe and stable operation of satellites. Clustering analysis has proved effective in many engineering anomaly detection problems. However, the quality of clustering is highly sensitive to parameters, and there is no convenient parameter selection method for now. To realize the adaptive selection of clustering parameters, this paper treats parameter adjustment as a single-objective optimization problem and introduces intelligent optimization algorithm to solve it. Accordingly, the UMOEAsII_BIRCH algorithm is proposed by combining BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies) and UMOEAs-II (United Multi-operator Evolutionary Algorithms-II). Based on the real data of a space science satellite, the telemetry data containing point anomalies are simulated and used for anomaly detection effect test after preprocessing. The traditional clustering algorithms K-Means, MeanShift, OPTICS, DBSCAN, BIRCH and the algorithm proposed in this paper are selected for experiments to compare the accuracy, recall, F1-score and false positive rate of anomaly detection. The results demonstrate that the UMOEAsII_BIRCH algorithm outperforms other algorithms in anomaly detection, and the F1-score can reach 0.861017. Meanwhile, compared with grid search, the proposed algorithm requires less manual intervention and achieves automatic selection of optimal clustering parameters, in line with the improvement expectations.