Conditional Multivariate Elliptical Copulas to Model Residential Load Profiles From Smart Meter Data

Abstract

The development of thorough probability models for highly volatile load profiles based on smart meter data is crucial to obtain accurate results when developing grid planning and operational frameworks. This paper proposes a new top-down modeling approach for residential load profiles (RLPs) based on multivariate elliptical copulas that can capture the complex correlation between time steps. This model can be used to generate individual and aggregated daily RLPs to simulate the operation of medium and low voltage distribution networks in flexible time horizons. Additionally, the proposed model can simulate RLPs conditioned to an annual energy consumption and daily weather profiles such as solar irradiance and temperature. The simulated daily profiles accurately capture the seasonal, weekends, and weekdays power consumption trends. Five databases with actual smart meter measurements at different time resolutions have been used for the model's validation. Results show that the proposed model can successfully replicate statistical properties such as autocorrelation of the time series, and load consumption probability densities for different seasons. The proposed model outperforms other multivariate state-of-the-art methods, such as Gaussian Mixture Models, by one order of magnitude in two different distance metrics for probability distributions.

Operation Orchestration of Local Energy Communities through Digital Twin: A Review on suitable Modeling and Simulation Approaches

“…Thanks to the cutting-edge Artificial Intelligence (AI) technologies, behaviors which are difficult to model by physics-based model can be captured by data-driven models or a combination of these methods, using machine learning techniques taking into account historical data. In the model library, the LEC profiles (PV, base-load, flexibility) are generated by leveraging some existing application programming interfaces (APIs) such as Artificial Load Profile Generator (ALPG) [4], Load Profile Generator [5], Residential load profile based on gaussian mixture model [6] or multivariate elliptical copulas model [7]. The model library is developed in Python open source to facilitate replication both in academia and industry.…”

“…3) is also one of the important factors that needs to be considered for setting up a DT platform. Depending on the analytical needs of the stakeholders, the DT platform can be developed to model different levels, ranging from component [8], household/building [7], [9], and complete LEC level [10], to distribution [11], [12], and grid/market interaction level [13]. Each modeling level corresponds to a specific time-scale frame (from microsecond to years or longer).…”

“…To enhance the static load models, proper stochastic modeling of load consumption is required in different types of studies such as modern grid planning, quantification of impact of low carbon technologies, finding secure levels of penetration of PV generation, and LV state estimation as in [7], [40]- [42]. The main purpose of these approaches is to capture the statistical properties of the smart meter measurement dataset to generate the load profiles.…”

“…In [7], the authors proposed a top-down data-driven modeling approach named conditional multivariate elliptical copulas for generating the resident load profiles in quasi-dynamic time frame. The proposed approach unifies the consumption modeling for MV and LV networks, simulating active power consumption scenarios in flexible time horizons for a whole year.…”

“…The profile requirements, besides data from smart meter, important data such as weather condition, cost rate, and assigned data from transmission/distribution system operators (TSO, DSOs) are also considered in DT platform for LEC. [7], [17], [19], [36], [40] [1], [17], [19] Building/LEC selfsufficiency [26], [27], [38] In TROEF project, the architecture of DT platform is designed in the smart grid architecture model (SGAM) [58], including three layers as shown in Fig. 5.…”