Least Mean Square-Based Fuzzy c-Means Clustering for Load Recognition of Induction Heating

Currently, with the development of induction heating (IH) and artificial intelligence (AI), the domestic induction cooker (DIC) has become smarter and more reliable. DIC should be able to recognize the material and size of cookware, as well as be susceptible to users' misbehavior. This article utilizes AI techniques to recognize the cookware items and protect the users and systems from hazards. First, a laboratory prototype of DIC is used to collect the output voltage and current for different cookware items under different switching frequencies and temperatures. After that, the Gray Wolf Optimizer (GWO) algorithm is applied to estimate the equivalent resistance and inductance of all coil-cookware loads. Then, the fuzzy c-means (FCM) clustering algorithm and the least mean square (LMS) algorithm are combined and applied for clustering the unlabeled data. This article suggested four clusters, including non-ferromagnetic (NF) cookware, small-sized ferromagnetic (SF) cookware, medium-sized ferromagnetic (MF) cookware, and no cookware (NC) cluster. Finally, the proposed AI method is applied to the load recognition of a commercial DIC. By comparing the centers of the clusters with the estimated equivalent resistance and inductance of the coil-cookware load, the size and material of the cookware can be found.