Evaluation of PSO-Tuned DBSCAN and Local Hash-Linked Recording of Synthetic Power Quality Disturbances in Electric Vehicle Charging and Electromobility Systems

Electric vehicle charging motivates the joint consideration of disturbance analysis and traceable event records. This study evaluates particle-swarm-tuned density clustering coupled to a local hash-linked recording procedure. The primary search maximizes silhouette without event labels, using duration and magnitude from synthetic grid-side and electromobility populations of 325 and 132 events. Thirty independent populations and matched-budget random and grid searches quantify agreement, rejection and variability. The primary method attains mean adjusted Rand indices of 0.477 and 0.267, with retained fractions of 0.591 and 0.313, respectively. Complete synthetic waveform-feature ablations reduce agreement under the same search settings. A separate experiment on 47 eligible measured grid events also finds no mean improvement from adding a band-limited harmonic descriptor. Held-out comparisons favor the tested support-vector, random-forest and graph neural classifiers over the specified nearest-core extension of density clustering. Canonical event payloads connect fitted clusters to stored records. Alterations without rehashing are detected locally, whereas consistent rewriting and tail deletion require an independently trusted terminal digest and record count. These results identify the limits of silhouette-only optimization and distinguish local record consistency from sensor authenticity, distributed consensus and operational charging-event recognition.