In this article, an augmented complex Kalman filter (ACKF) is proposed for distribution networks as a noniterative forecasting-aided state estimator. Although most of the distribution state estimator (DSE) algorithms deal with real and imaginary parts of distribution networks' states independently, the proposed algorithm in this article considers the states as complex values. In the case of real-time DSE and the presence of a large number of customer loads in the system, employing DSEs in one single estimation layer is not computationally efficient. Consequently, our proposed method performs in several estimation layers hierarchically as a multilayer DSE based on an ACKF (DSE-MACKF). In the proposed method, a distribution network can be divided into one main area and several subareas. The aggregated loads in each subarea act like a big customer load in the main area. Customer load aggregation results in lower variability and higher spatial-temporal correlation. This increases the accuracy of the estimated states in the main estimation layer. Additionally, the proposed method is formulated to include unbalanced loads in low-voltage radial distribution networks. This approach is applied to two real distribution networks for comparison and evaluation. The effectiveness of the proposed method is discussed using several criteria such as computational time, standard deviation, and maximum and average voltage error. The performance of the proposed method has been assessed against the weighted least square estimation method, while the computational time and the average voltage error have been decreased from 14 to 2 s and 1.13% to 0.23%, respectively.