Smart meter based profiling for load forecasting and demand side management in smart grids

Khan, Zafar Ali ORCID: 0000-0003-3149-6865 (2019). Smart meter based profiling for load forecasting and demand side management in smart grids. University of Birmingham. Ph.D.

Preview

Khan2019PhD.pdf
Text - Accepted Version
Available under License Creative Commons Attribution.
Download (2MB) | Preview

Abstract

The smart grid incorporates an integrated system of smart meters and communication networks that enable two-way communication between utilities and consumers. The granular information from smart meters can be used to improve the load forecast and influence consumer’s energy consumption patterns through demand side management (DSM). However, for localized studies of power system, using a large quantity of smart meter data having high level of noise preclude the use of computationally intensive techniques. Reduction of smart meter data to extract the load profiles and smoother load profiles at lower aggregation level (individual consumer or small groups of consumers) are highly desirable for use in linear techniques for power system studies. Therefore, this thesis addresses the challenges of smart meter data size, complexity, variability and volatility for efficient use in load forecasting and DSM.

This thesis presents a novel clustering-based approach for analysis of smart meter data, aimed at more accurate and detailed load profiling, reduced profile complexity and improved load forecast accuracy and DSM solutions. The approach uses an innovative clustering algorithm to reduce the data size by proposing new cluster validity indices. The extremely volatile profiles having high levels of noise and complexity are linearized using Taylor series linearization process to alleviate the non-linearity and complexity of profiles. Finally, particle swarm optimization is applied for energy optimization in linearized profiles. The approach is demonstrated on Irish smart meter dataset and simulated PV data, to achieve improved load forecast accuracy using artificial neural network and improved DSM solutions using linear optimization with reduced computational burden.

Investigations suggest that proposed clustering algorithm can produce clusters with high intra-cluster pattern similarity as a result of the introduction of new stopping criteria specifically tailored for load forecasting applications. A comparison between the proposed alternative profiles and raw profiles further suggests that the alternative profiles guide the underlying energy consumption with reduced complexity making them computationally efficient. Use of the alternative profiles suggests that the load forecasting accuracy can potentially be higher compared to raw profiles. The alternative profiles in combination with the novel cluster selection approach provide higher peak reduction by shifting the load from peak hours to off-peak hours and higher monetary benefits for the participating consumers. The proposed clustering algorithm and the alternative profiles represent an advancement of the conventional load profiling approach, benefiting system operators through more accurate forecasting and efficient DSM. The novel mathematical framework suggested in this thesis provides an advancement to the new knowledge in the area of smart metering and smart power grids.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):