Power data asset evaluation occupies the core position in the digitization of the power industry. It involves the analysis and utilization of a large amount of power data. The key is to process time series data, such as power consumption and power generation. These data have both long-term and short-term patterns, and traditional evaluation methods such as autoregressive models or Gaussian processes may be difficult to fully capture their characteristics, resulting in evaluation bias. In response to this challenge, this paper proposes a new deep learning framework, namely Time-Series Convolutional Memory Efficient Network (TSENet). TSENet uses complex Sophisticated Convolutional Neural Network (SCNN) and Expressway Network (ENet), combining the advantages of Long-and Short-term Time-series Network (LSTNet). It can simultaneously capture short-term local features and long-term global trends in power data, help to deeply mine spatial correlations and local patterns in data, effectively extract fine relationships between variables and optimize information flow. In the evaluation of the complex and rich Solar-Power dataset and Electricity dataset, TSENet achieved significant performance improvements over other state-of-the-art baseline methods.Through the synergistic design of deep convolutional structures and an efficient memory mechanism, it effectively addresses issues such as inadequate modeling of long-term dependencies, insufficient extraction of short-term features, and high prediction volatility, thereby significantly enhancing both the accuracy and robustness of forecasting in power asset evaluation tasks.