Cyberattacks against power systems increasingly exploit industrial communication and control dependencies arising from distributed energy resources, remote management, and market-based incentives. Although intrusion detection and resilience mechanisms are actively researched, progress is constrained by a lack of public, well-documented datasets that align network telemetry with physical measurements and control context in realistic operational settings. This paper presents Data4Cyber, a suite of labeled datasets recorded in a cyberphysical testbed emulating a distribution-grid management system for self-consumption optimization with photovoltaic generation, battery storage, controllable loads, and substation/feeder metering. The release combines synchronized (i) time-series electrical measurements, (ii) controller-relevant context including price incentives, and (iii) packet captures of OT communication (Modbus/TCP and MQTT). It covers one benign baseline (S0) and six attack scenarios (S1–S6), including protocol-semantic Modbus manipulation, man-in-the-middle false-data injection, and MQTT supply-chain-style price manipulation. Data4Cyber is provided in machine-learning-friendly formats with fixed time alignment (default: 1 s), phase-resolved attack annotations, and standalone metadata describing scenario context, experiment setup, and testbed infrastructure.