### Abstract:

Accurate load models are required for the computation of load flows in MV distribution networks. Modern microprocessors in recent times enable researchers to sample and log domestic loads. The findings show that they are stochastic in nature and are best described by a beta probability distribution. In rural areas two different load types may be present. Such loads are domestic and pump loads, the latter may be modeled as constant P - Q loads. An analytical tool for computing voltage regulation on MV distribution networks for rural areas feeding the mentioned loads is therefore required. The statistical evaluation of the consumer voltages requires a description of load currents at the time of the system maximum demand. To obtain overall consumer voltages are any specified risk for the two types of the loads, the principle of superposition id adopted. The recent work deals with conventional 22kV three-phase distribution connected networks as used by ESKOM, South Africa. As the result of the connected load, MV networks can experience poor voltage regulation. To solve the problem of voltage regulation, voltage regulators are employed. The voltage regulators considered are step-voltage regulators, capacitors and USE (Universal Semiconductor Electrification (devices. USE devices can compensate for the voltage drops of up to 35 percent along the MV distribution networks, thus the criteria for the application for the USE devices is also investigated. The load currents are treated as signals when assessing the cost of distribution system over a period of time due to power losses. The individual load current signal is modeled by its man and standard deviation. The analytical work for developing general expressions of the total real and total imaginary components of branch voltage drops and line power losses in single and three-phase networks without branches are presented. To deal with beta-distributed currents on MV distribution networks, new scaling factors are evaluated at each node. These new scaling factors are derived from the distribution transformer turns ratio and the deterministic component of the statistically distributed load currents treated as constant real power loads. In the case of an individual load current signal, the transformation ratio is evaluated from the distribution transformer turns ratio and the average value of the signal treated as constant real power load. The evaluation of the consumer voltage percentile values can be accurately evaluated up to 35 percent voltage drop. This is possibly by the application of the expanded Taylor series, using the first three terms. The coefficients of these three terms were obtained using a search engine imbedded in the probabilistic load flow. The general expressions for evaluating the overall consumer voltages due to statistical and non-statistical loads currents are also given. These non-statistical currents may be due to constant P- Q loads, line capacitance and the modeling of voltage regulators. The Newton-Raphson algorithm is applied to perform a deterministic load flow on single-phase networks. A backward and forward sweep algorithm is applied to perform a deterministic load flow on single and three-phase systems. A new procedure for modelling step-voltage regulators in three-phase systems, is outlined. A new procedure for modeling step=voltage regulators in three-phase connected networks is outlined. Specifying a transformation ratio of 1.1 and 1.15 respectively, identifies the open-delta of closed-delta configuration for three-phase networks.