Electrical Thermal Load Forecasting with Trend-Segmented ELM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electrical system thermal load predictions are inaccurate due to reliance on single weather factors, neglecting factors affecting load changes in different time sections of the same load day.
Innovation Solution
A method using dynamic segmentation and an extreme learning machine (ELM) algorithm to predict thermal load trends in the next 24 hours, involving data preprocessing, trend similarity calculation, and model training with a similarity sequence matrix.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single weather factor is used to select similar trends for thermal load prediction, then the prediction method is simple, but the prediction accuracy deteriorates because factors affecting load changes in different time sections are not considered
Solution Approach 1:
The patent divides the daily thermal load curve into multiple time sections (e.g., peak load periods, valley load periods, intermediate periods) based on extreme points and slope changes. Each time section is analyzed separately with its own similarity comparison, allowing different weather factors to be considered for different periods. This segmentation resolves the contradiction by increasing prediction accuracy through differentiated analysis while maintaining reasonable complexity through systematic processing.
Solution Approach 2:
The patent dynamically selects weather factors and similarity thresholds based on the specific time section being analyzed. Different weather factors (temperature, humidity, wind speed, etc.) are weighted or selected differently for different time periods. This dynamic adaptation allows the system to capture the varying influences of weather on thermal load throughout the day, improving accuracy without requiring a completely complex static model.
2Device complexity
If historical data from all time sections is treated uniformly, then the data processing is simple, but important change trend information in different time sections is lost
Solution Approach 1:
The patent segments the historical thermal load data into multiple time sections corresponding to different operational periods (peak, valley, intermediate). Each segment is processed independently to identify and preserve the characteristic change trends specific to that period. This prevents the loss of important temporal patterns that would occur if all data were treated uniformly, while maintaining manageable complexity through structured segmentation.
Solution Approach 2:
The patent applies different processing qualities and similarity criteria to different time sections. Each time section receives processing tailored to its specific characteristics, with appropriate similarity thresholds and weather factor selections. This local quality approach ensures that important change trend information is preserved in each segment without requiring overly complex uniform processing across all data.
3Productivity
If N is set to a small value for calculating data daily reference line, then the calculation is faster, but the prediction accuracy deteriorates due to insufficient historical data coverage
Solution Approach 1:
The patent dynamically determines the value of N (number of historical days to average) based on the specific time section and meteorological conditions. For time sections with stable patterns, a smaller N may suffice for faster processing. For sections with variable conditions requiring more historical context, a larger N is applied. This dynamic adjustment resolves the contradiction by optimizing the balance between calculation speed and accuracy for each specific case.
Solution Approach 2:
The patent changes the parameter N (historical data coverage) based on different time sections and prediction requirements. By adjusting this parameter dynamically, the system can increase historical data coverage when accuracy is prioritized and reduce it when speed is needed, resolving the contradiction between productivity and measurement precision through parameter optimization.
Data Source
Figure 1
Figure 2
AI summary
A method and device for predicting the thermal load of an electrical system. The method comprises: S1: pre-processing historical daily data of the thermal load of an electrical system; S2: acquiring a data daily reference line according to the pre-processed historical daily data; S3: dividing the acquired data daily reference line into multiple time sections; S4: screening the historical daily data, and calculating a trend similarity value of the screened historical daily data and the data daily reference line within each divided time section respectively; S5: choosing historical daily data corresponding to a trend similarity value greater than a preset reference value to form a similarity sequence matrix; S6: inputting the similarity sequence matrix into an extreme learning machine (ELM) for training, acquiring a prediction model, and predicting the thermal load of the electrical system. The method effectively retains important change trend information in a time sequence of a thermal load by using a time sequence representation method based on trend segmentation, thereby being capable of more accurately predicting the change trend of the thermal load.