Increased single-phase electric vehicle (EV) charging often causes phase unbalance in low-voltage networks (LVNs). This paper presents a dynamic assessment and mitigation framework that uses a linear model predictive control (MPC) approach to mitigate phase unbalance. EV data from Leeds and two representative UK low-voltage feeders were used for testing, comprising an urban feeder and a rural feeder. EV charging scenarios were constructed at 15-minute resolution under different EV scenarios.

In the urban feeder, most scenarios remained within the voltage unbalance factor (VUF) limit of less than 2\% and the voltage constraint of $\pm10\%$, showing that the feeder was relatively robust under the tested charging conditions. However, under fixed phase-skewed EV placement, the maximum VUF reached 2.1263\%, with 63 intervals exceeding the 2\% VUF limit. After applying the control framework, the maximum VUF dropped to 1.9500\%, and all phase-unbalance violations were resolved.

In contrast, the rural feeder violated the phase-unbalance limit under the strongly skewed scenario. Before control was applied, the maximum VUF reached 4.0201\%, with 217 intervals exceeding the 2\% VUF limit. After applying the control framework, the maximum VUF dropped to 1.9498\%, and all phase-unbalance violations were resolved. These results demonstrate that the proposed device-agnostic MPC can dynamically mitigate EV-induced phase unbalance and determine the corrective capability required as network conditions vary over time.