Using AI to Monitor Robust Operation of the Grid

 Using Complex-Valued Neural Networks (CVNNs) to forecast different parameter in the grid (such as solar irritation, load, etc) with high penetration level of uncertain renewable energies.

The Challenge

High penetration of renewable energy poses a significant challenge in operation of power system. Higher penetration of PV increases the concern about the potential impacts on the transmission system. Based on the uncertainty in the power grid caused by renewable energy generations, forecasting the parameters are really important to minimize the costs and power outage and increase the grid’s reliability.

Artificial neural networks (ANNs) have received considerable attention in academia for solving problems related to power systems. ANNs can single out the correlation pattern from input datasets to target datasets. They use the pattern that is identified to estimate the output for a new set of inputs. Back- propagation (BP) method is the workhorse for ANNs’ learning of network structures, which is used extensively for Short Term Load Forecasting (STLF), because of its extraordinary mapping capabilities.

 

The Solution

Complex-Valued Neural Networks (CVNNs) prove their abilities for the identification of nonlinear systems. They can outperform their real counterparts in many ways. CVNNs give also the possibility of the simultaneous modeling and forecasting. The main motivation to use CVNNs is due to faster convergence, reduction in learning parameters, and ability to learn two dimension motion of signal. The CVNNs are simply the generalization of the real valued neural networks in the complex valued domain, where all the parameters including weights, biases, inputs, and outputs could be complex variables. In one of the projects in REDLab, the goal is to find a CVNN model which can predict the solar irradiation (hourly, daily, monthly, etc.) using the time series technique or with other meteorological variables, providing that this model should be trained before. In another project, we will present different methods for load forecasting. Reliable forecasting of daily load is required to effectively utilize the battery energy storage system.

Related publications: Article 1, Article 2, Article 3            

 

Featured Photos

Hybrid Forecast And Load Commitment (HFALC) unit interaction with Energy Management System (EMS)

Top: Flowchart of series-parallel load forecasting method, down: Series-parallel method for 20 min look ahead forecast accompanied with actual load data

BESS connection to the distribution grid system

(Top: the map of OUR real case study (MAUI ISLAND, HAWAII), down: the electrical circuit of the BESS in the distribution grid)

Architecture of CVNN including nodes, layers, and weights (Inputs: historical data, outputs: forecasting data)

CVNN forecasting method along with the actual load data (top: without filter, down: with Kalman filter)

Want to learn more about our AI methods to monitor the robust operation of the grid?