A rank-oriented set-up for the compensation algorithm

A procedure is given that solves modified network equations much more efficiently if the modification description suffers from rank deficiency. The key to the method lies in the application of a low-order product factorization to the modification matrix. Consequently, the dimension of the reduced system is determined by the rank of the modification and, thus, the computational process can be speeded up considerably. Test results are given for networks with up to 944 nodes. Two algorithms for the low-order produce factorization are summarized.<>