This paper proposes a fault diagnosis algorithm for DC distribution network based on improved half-tensor product Bayesian networks. The Bayesian network method based on half-tensor product is improved by introducing the protection and circuit breaker action moment confidence and action state confidence

which improves the fault diagnosis accuracy and enables to correct the calculation results even when the conditional probability is inaccurate. Considering the deep integration of protection and control in DC distribution network

the control quantity reflecting the change of control state and the protection quantity reflecting the protection action are integrated into the half-tensor product Bayesian network

which makes the fault diagnosis results more accurate. The correctness and reliability of the proposed diagnosis method are verified through the analysis of arithmetic cases.