A non-linear cubic Bezier-functional expansion-based adaptive filter (CB-FEBAF) has been designed for achieving shunt compensation in this article. The algorithm is developed for mitigating current-based power quality problems such as harmonics in supply current, reactive VAR compensation, active power compensation, power factor improvement, load balancing, and so on. The CB-FEBAF is developed using Bezier curve expansion of the input signal. The functional expansion-based real time and online training converges fast and shows fast response over offline techniques such as neural network and neuro-fuzzy-based algorithms. The designed CB-FEBAF controller is trained online using gradient descent least mean square algorithm to extract the fundamental component of the load current. Moreover, the feedforward active power term corresponding to PV power contribution has been added to the developed controller. This helps to overcome the challenges in the integration of renewable energy-based distribution systems. The proposed controller is compared with non-adaptive synchronous reference frame theory, backpropagation neural network, and legendre-based functional neural network. Hardware results prove the multifunctional capabilities of the developed approach.