Static VAR compensators (SVC) can improve some of the power quality (PQ) indices such as voltage flicker, unbalances, and power factors in power systems if properly controlled. Different nonlinear and time-varying loads generate these parameters. One of the great nonlinear loads is the electric arc furnace (EAF), which causes waveform distortions, voltage unbalances, and fluctuations. In this paper, a new digital and prediction-based control for thyristor switched capacitors (TSCs) was proposed to compensate an actual steel industrial plant. The digital control was based on generating adequate synchronous pulses, calculating required suitable susceptance, and measuring correct reactive power. The proposed TSC improved the PQ by means of the feedback from voltage and power factor. A predictive method based on time series and recursive least square with dynamic learning factor was then applied to estimate EAF reactive power in the future so that the TSC performance was enhanced by reducing the natural time delay. A band-pass harmonic filter was designed to compensate the load current harmonics and to protect the capacitor banks in the TSC. Finally, the suggested TSC was implemented in a steel industrial plant as an actual power system with 2 EAFs. The modeling and experimental results indicate the effectiveness of the proposed TSC.