An Adaptive Machine Learning Framework for Behind-the-Meter LoadPV Disaggregation
ABSTARCT :
A significant amount of distributed photovoltaic (PV) generation is invisible to distribution system operators since it is behind the meter on customer premises and not directly monitored by the utility. The generation essentially adds an unknown varying negative demand to the system, which causes additional uncertainty in determining the total load. This uncertainty directly impacts system reliability, cold load pickup, load behavior modeling, and hence cost of operation. This paper proposes an adaptive machine learning framework to disaggregate PV generation and load from the net measurement. The framework estimates PV generation and load based on measurements/data collected from smart transformers, smart meters, other sensors, and weather stations. The proposed framework core idea is to transform the data such that: a) the machine learning model can effectively utilize the time dependency of measurements, and b) the measurements are transformed into a lower-dimensional space to reduce complexity while maintaining accuracy. The transformed measurements are then used to train the machine learning models for load/PV dis-aggregation. Machine learning models investigated include linear regression (LR), decision tree (DT), random forest (RF), and Multilayer Perceptron (MLP). Several test/ training split scenarios, including 90%-10% Split, One-Month-Out, One-Season-Out, and Panel-Independent Split, are presented to provide a thorough evaluation of the proposed framework. Results show that the proposed framework can estimate PV generation with high accuracy using low-complexity methods, and random forest is found to provide superior performance compared to the other ML models investigated.
EXISTING SYSTEM :
? This level of error is comparable and, in some cases lower than the error found in existing load disaggregation techniques for different load types.
? Therefore, the model developed here can be accurately used for the purpose of power disaggregation of identical loads.
? The minimum amount of energy saved from a three-hour DR event was found to be 27.5%, which is comparable and, in some cases, greater than savings found from using existing DR algorithms for a DR event of the same duration of time.
? The existing model could be replaced with a physical building model, using standards defined in the ASHRAE (American Society of Heating, Refrigerating and Air-Conditioning Engineers) handbook.
DISADVANTAGE :
? This invisibility, along with the stochastic nature of solar power, can cause new problems for utilities, such as inaccurate load forecasting and estimation, inefficient service restoration, and sub-optimal network expansion decisions.
? One solution to this problem is to monitor each single rooftop PV generation by installing extra metering devices.
? Based on similarity graphs obtained from demand/solar power data, the clustering process is transformed into a graph partitioning problem, which cuts a graph into multiple smaller sections by removing edges.
? The weight updating process is cast as a repeated game model with vector payoff, in which two components are defined: a player and a set of experts, which in our problem correspond to the disaggregator and the candidate exemplars, respectively.
PROPOSED SYSTEM :
• Autonomous control of multiple lighting systems during a DR event using Reinforcement Learning (RL) has been proposed, taking into account the occupants illuminance comfort.
• The problem is modelled as a Markov Decision Process, and an optimization approach based on RL is proposed.
• Another Q-learning approach that controls the illuminance level by adjusting lights and blinds in a smart building has been developed.
• An automated DR algorithm has been proposed and tested where in an interruptible load scheme, buildings participate as an aggregated load.
• A Stackelberg Game-based Demand Response (SGDR) algorithm is proposed.
ADVANTAGE :
? The objective of the coefficient optimization process is to minimize the disaggregation residuals.
? The reason for this better performance is that the RGVP-based method can identify the candidate exemplars that are highly correlated with the BTM real load/solar powers.
? These coefficients determine the optimal disaggregation of the measured net demand for customers in SN.
? Parametric models along with weather information have been used to estimate solar generation.
? These updated weights are then used to generate the composite native demand and solar generation exemplars for the next time point.
? These calibrated parameters are then used to perform BTM PV generation disaggregation.
|