Modern power-grid information systems require efficient and intelligent use of heterogeneous computing resources, including CPU, GPU, memory, and network bandwidth. To address this, we propose a method for quantitative evaluation and coordinated scheduling of heterogeneous computing resources. The framework consists of three modules: First, a resource sensing and abstraction module continuously collects multidimensional resource usage data and maps them into normalized indices to reflect CPU load, memory pressure, GPU intensity, and network demand, forming a unified resource status view. Second, a scheduling and allocation module evaluates each task’s resource demand and priority level, and calculates a feasibility index to determine schedulability. A multi-objective strategy generates allocation plans balancing task urgency, system load, and resource efficiency. Third, a monitoring and feedback module evaluates execution using indicators such as task completion rate and latency, enabling dynamic policy adjustment. This approach enables fine-grained, adaptive scheduling of heterogeneous resources in power-grid computing scenarios, improving resource utilization and service responsiveness.