In 2022, the domestic sector accounted for 38% of UK electricity demand at 109 TWh, exhibiting a larger percentage share of electricity demand than the industrial sector. With the push towards net-zero goals, grid stress is only going to be exacerbated by the electrification of heat and EVs. Community energy schemes aim to directly address growing concerns surrounding peak grid demand stress by maximising the usage of on-site micro-generation and simultaneously decarbonising the community. In the case of utilising community PV alongside energy storage technologies, understanding site-specific future PV generation is a key challenge in implementing an energy management strategy. This work outlines a linear regression model that relates the measured on-site global solar irradiance and community PV array generation. The model is used to generate a 5-day PV generation forecast using forecast solar irradiance data via a weather API whilst using observational data to continually retrain the model to improve specificity and accuracy. This data will then be used to optimise a community energy management system in determining the available on-site generation and how this interacts with community electricity demand loads, storage and grid imports/exports. The model has demonstrated an average forecasting accuracy of 88% ± 9%, thus providing an effective forecast tool for on-site PV generation. Over time, the linear regression fit has changed by over 36% when compared with the initial training set fit, demonstrating a clear improvement in model performance over a 5-month period. This flexibility is an important characteristic in making the model specific to the community site and resilient to change. Further work will involve improving forecast resolution and implementing this as part of a wider community energy management system.