With the rapid development of smart grids and renewable energy

the power Internet of Things(IoT)faces increasing challenges in edge device access

data processing

and platform-level management and control.Load forecasting

as a key technology

directly affects system scheduling and operational efficiency.This study proposes a hybrid load prediction model that integrates a convolutional neural network(CNN)and a long short-term memory network(LSTM)for power load forecasting of edge terminals.The model leverages the local feature extraction capability of CNNs and the temporal dependency modeling strength of LSTMs.Experimental results obtained on a public power consumption dataset show that the proposed CNN-LSTM model outperforms standalone multi-layer perceptron(MLP)

CNN

and LSTM models

achieving superior performance in terms of mean squared error(MSE)and mean absolute error(MAE).Further analysis indicates that temperature

humidity

and temporal features have a significant impact on load variations.The proposed approach provides effective support for optimizing edge node access strategies

improving resource allocation efficiency

and enabling data-driven management and control decisions

and demonstrates potential for extension to broader intelligent analytics scenarios in IoT systems.