Substation Rate Forecasting for Anomalous Grid Events
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current forecast models for electrical grids do not accurately account for anomalous electrical rate increases due to outage events and natural disasters, leading to low accuracy and underestimation of locational marginal pricing volatility, and lack a comprehensive analysis of both normal and unusual grid operations.
Innovation Solution
A hybrid approach combining statistical anomaly detection with domain-specific rules is used to filter anomalous data, followed by training machine learning models to forecast short-term and long-term electrical rate deviations and fluctuations, incorporating a stochastic model to predict anomalous events and normal operations, and selecting the model with the highest composite performance score for integrated forecasting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current forecast models are used for electrical grids, then the forecasting process is simple, but the accuracy of electrical rate predictions is low and anomalous events are not accounted for
Solution Approach 1:
The forecast model is segmented into multiple specialized models: a stochastic model for anomalous events and machine learning models for normal operations. This segmentation allows each model to specialize in specific types of electrical rate fluctuations, improving overall prediction accuracy while managing complexity through modular architecture.
Solution Approach 2:
The system changes parameters by filtering historical data to separate anomalous from normal operations, then using different modeling approaches for each parameter type. This parameter-based differentiation enables accurate forecasting of both anomalous electrical rate increases and normal fluctuations.
2Reliability
If comprehensive analysis of both normal and unusual grid operations is performed, then the forecast accuracy improves, but the computational complexity and data processing requirements increase
Solution Approach 1:
The system extracts and separates anomalous data from historical records using filtering techniques, isolating unusual events from normal operations. This extraction allows the stochastic model to focus specifically on anomalous patterns while machine learning models handle normal operations, improving reliability without requiring a single overly complex model.
Solution Approach 2:
A data filtering and preprocessing system acts as an intermediary between raw historical data and the forecasting models. This intermediary layer cleans and organizes data, separating anomalous from normal operations, which simplifies the task for downstream models and improves overall system reliability.
3Measurement precision
If statistical anomaly detection with domain-specific rules is used to filter data, then the quality of training data improves, but the data processing time and computational resources increase
Solution Approach 1:
The system performs preliminary data filtering and anomaly detection on historical data before training the forecasting models. By pre-processing and cleaning data in advance, the system improves training data quality while reducing the computational burden during model training and deployment phases.
Data Source
AI summary
A computer-implemented method is provided. A processor set selects a number of nodes. The processor set filters anomalous data from historical data for each node in the number of nodes to generate a refined historical data. The processor set trains a number of computational models using the refined historical data to generate a number of machine learning models for forecasting electrical rate deviation and fluctuation for the number of nodes. The processor set assigns a composite performance score to each machine learning model based on performance for each machine learning model. The processor set computes anomalous electrical rate fluctuation based on the anomalous data for each node using a stochastic model. The processor set generates a forecast comprising anomalous electrical rate fluctuation, and electrical rate fluctuation and deviation for the number of nodes for a period of time.