An improved energy management system framework for solar energy integration.

Renewable energy sources like wind and solar play a crucial role in decarbonizing energy supply, but their variable and intermittent nature lead to reliability and stability issues. One way of sustainably integrating these energy sources into the grid is through an energy management system. The study reported in this thesis gives a comprehensive definition of an integrated energy management system and creates a novel framework that identifies energy forecasting, demand-side management, and supply-side management, as crucial components for grid balancing. In addition, this research looks particularly at solar integration, and how the integrated energy management system offers a unique combination of solar energy forecasting, time-of-use tariffs, direct load control demand response, and generator control, in increasing penetration levels of solar energy. The significance of this research is that the proposed system presents a viable, sustainable, and cheaper way of increasing PV usage and thereby grid penetration by prioritising efficient use of available PV supply before calling up additional supply. To validate the proposed integrated energy management system, this research looks to understand the functions of each individual component and how their interconnectedness creates a novel management system. Firstly, this research develops a three-step solar forecasting approach that uses low-level data fusion to combine weather variables from both an on-site and a local weather station to improve solar energy forecasting. The forecasting model response is historic PV generation, and the predictors are weather variables with moderate to strong positive correlations to solar radiation. Data obtained is preprocessed using Low-level Data Fusion, Pearson Correlation Coefficient analysis, Rescaling method, and List-wise Deletion method. This approach is then tested on a 1MW utility scale solar plant, resulting in a 6% and 13% prediction accuracy improvement when compared to solely using data from an on-site, and local weather stations respectively. This approach is also validated for three residential rooftop solar systems (8 kW, 10.5 kW and 15 kW), achieving root mean square error values of 0.0984, 0.1425, and 0.0885 respectively. The resulting low root mean square error values, a measure of the predicted PV to actual PV generation, proves that the model can be adopted for different PV plant sizes and is suitable for any customer across the distributed generation spectrum. To further improve the accuracy of the model, other preprocessing techniques are investigated and applied. The study shows that the combination of Low-level Data Fusion, Linear Interpolation, filling outliers, data smoothing, Rescaling method, moderate to strong PV correlation of weather parameters using Pearson Correlation Coefficient, day/time/month decomposition, seasonal decomposition, Principal Component Analysis, and holdout validation, increases the accuracy of the model by 75%. The ability of direct load control to manage energy consumption is validated in a case study by using Connected Power’s unique smart sockets and Lumen radio’s Mira Mesh Radio Frequency wireless network. Small plug-in loads were connected to ten smart sockets located in a robotics laboratory and a café, resulting in reduced energy consumption by 44% and 72% respectively when compared to the baseline without direct load control. Finally, the integrated energy management system framework is validated by testing its capacity to increase PV usage for an off-grid residential house with a PV/diesel generator power source. A decision-based algorithm is created that adjusts PV supply forecast errors, initiates direct load control responses to reduce excess load during periods of low PV supply, and/or increase power supply by calling up a diesel generator. In addition, this is combined with the proposed three-step solar energy forecasting approach and a programmable load schedule based on time-of-use criteria. The effects of customer behaviour are also analysed by using a 14% override rate, with 80% preconditioning and 20% rebounding. The hybrid PV/diesel generator power source with the proposed integrated energy management system is compared against two configurations: a baseline configuration that uses a solely diesel generator source, and a hybrid PV/diesel generator power source. Results show that the integrated energy management system reduced the lifetime expenditure costs and CO2 emissions by 44% and 46% respectively when compared to the baseline configuration, and by 8% and 9% in the hybrid photovoltaic/diesel generator, while also increasing the PV usage from this configuration by over 113%. This research also addresses opportunities and limitations of the proposed system and lays the foundation for future research using other intermittent renewable energy sources such as wind.