LOAD FORECASTING FROM INDIVIDUAL CUSTOMER TO SYSTEM LEVEL BASED ON PRICE
The present invention relates to system and method for providing near real-time DR events and price signals to the customer end-points to optimally manage the available DR resources. The system utilizes bottom up load forecasting for accurate individualized forecasts for customer loads in the presence of dynamic pricing signals. For better efficiency and reliability of grid operation the system utilizes advanced machine learning and robust optimization techniques for real-time and “personalized” DR-offer dispatch.
1 . A method for individualized forecast of customer load in presence of dynamic pricing signals comprising:
recording the customer's participation history in different demand response events at each customer locations;
segmenting the demand response specific data in a plurality of related time series;
building a self-calibrated model for each customer using the time series;
taking feedback from the time series to predict the changes in customer load profile;
forecasting load usage and load shed as well as error distribution associated with forecast using machine learning and data mining techniques.
2 . The method of claim 1 wherein the demand response event specific data includes demand response resources data, its type, its locations, characteristics such as response time, ramp time, utility meter data, user specific data, time series data, seasonality data, price index data, notification time requirement, number of events in a particular period of time and number of consecutive event, user preference to participate in the event, price index and other regression based data.
3 . The method of claim 1 wherein demand response specific data is segmented on the basis of seasonality, time of occurrence, price index, temperature and other regression parameters.
4 . The method of claim 1 wherein the segmenting techniques used for segmenting the demand response event specific data includes K-mean and fuzzy K-means algorithm.
5 . The method of claim 1 wherein pricing signals are variable on current conditions and advanced notice requirements associated with a demand response event.
6 . The method of claim 1 wherein the forecasting of load is performed as a function of time of day, weather and price signal.
7 . The method of claim 1 wherein the self-calibrated model will be able to forecast shed capacity, ramp time and rebound effect for the customer.
8 . The method of claim 1 wherein the pricing signals include cost, reliability, loading order, preference, GHG etc.
9 . The method of claim 1 wherein the feedback is provided through machine learning techniques.
10 . The method of claim 1 wherein the participation history is collected through advanced metering infrastructures and sensors installed on the grid distribution.
11 . The method of claim 1 wherein the machine learning algorithm includes ARIMAX, KNN, SVM or Artificial Neural Network or a combination thereof.
12 . A method for individualized forecast of customer load in presence of dynamic price signals comprising:
Collecting a periodic electricity usage data at each customer level;
aggregating the electricity usage data at transformer, feeder and sub-station level;
creating a customer profile for electric load usage with a function of price elasticity, the said price elasticity function is estimated using a machine learning technique;
segmenting the customer electricity usage data in time series using clustering techniques;
forecasting the electricity load usage for each customer and the aggregated load usage at feeders, transformers and substation level.
13 . The method of claim 12 wherein the dynamic price signals include price based DR for load forecasting.
14 . The method of claim 12 wherein the pricing signals include cost, reliability, loading order, preference, GHG etc.
15 . The method of claim 12 wherein the participation history signifies history of participation in past event, strategy for reducing participation in high price event, notification time requirements.
16 . The method of claim 12 wherein the profile for individual customer is generated on the basis of electric usage at the end-level.
17 . The method of claim 12 wherein the machine learning techniques include ARIMAX, KNN, SVM or Artificial NeuralNetwork or a combination thereof.
18 . The method of claim 12 wherein the clustering techniques are used to segment the usage data in similar time series on the basis of seasonality, time of occurrence, price index, temperature and other variables.
19 . The method of claim 12 wherein the segmenting techniques used for segmenting the demand response event specific data includes K-means and fuzzy K-means methods.
20 . The method of claim 12 wherein the aggregated power load is calculated as the sum of forecast of individual customer.