Due to the uncertainties associated with renewable energy's stochastic fluctuations and the unique seasonal characteristics of rural loads

the operational scenarios of rural distribution networks with high penetration of distributed photovoltaics and biomass energy have become multidimensional. The complex coupling relationships between generation and load significantly increase the difficulty of prediction. To address this issue

this paper proposes a method for uncertainty net load forecasting based on multi-scenario clustering and an interpretable Temporal Fusion Transformers (TFT) network. Firstly

we implement scenario dimensionality reduction based on load fluctuation patterns and establish normalized metrics that comprehensively characterize the daily supply-demand state of the distribution network by combining the characteristics of photovoltaic and biogas power generation

enabling unsupervised clustering of multiple forecasting scenarios. Subsequently

we use multi-temporal dimension information along with the scenario clustering results as inputs for the interpretable model

mitigating the influence of irrelevant variables. Quantile loss is employed to obtain intervals for net load forecasting. Finally

the model's interpretability is utilized to quantify the contribution of various inputs to the output. Experimental results demonstrate that the proposed model achieves higher deterministic forecasting accuracy compared to conventional models

while also producing narrower prediction intervals during periods of frequent net load fluctuations. This capability provides valuable data support for the operational planning of distribution networks.