Power Load Forecasting via 2D Time-Series Image Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current power load prediction methods based on user tags are limited by fixed tags that fail to reflect the latest user conditions, resulting in reduced prediction accuracy.
Innovation Solution
A power load data prediction method that converts historical one-dimensional time series data into two-dimensional image data, using a trained neural network model for prediction, which includes a feature extraction sub-model and a time series prediction sub-model, and preprocesses data to supplement missing and extreme values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If power load prediction is based on user tags and growth rate, then the prediction method is simple to implement, but the prediction accuracy is limited due to fixed tags that cannot reflect the latest user conditions
Solution Approach 1:
The patent transforms one-dimensional time series power load data into two-dimensional image data by mapping time points to a coordinate system and encoding values as pixel intensities. This dimensional transformation enables the use of image processing techniques and deep learning models (such as CNNs) that can capture spatial patterns and temporal correlations more effectively, thereby improving prediction accuracy while maintaining implementation feasibility through standardized image processing pipelines
2Ease of operation
If traditional one-dimensional time series analysis is used, then the data processing is straightforward, but important data characteristics are lost
Solution Approach 1:
By converting one-dimensional time series into two-dimensional images where the x-axis represents time periods, the y-axis represents time points within periods, and pixel intensity encodes power load values, the method preserves temporal sequences while adding spatial structure. This enables the extraction of additional features such as temporal patterns, periodicity, and anomaly detection that are difficult to capture in one-dimensional analysis
Solution Approach 2:
The patent uses pixel intensity (analogous to color/brightness variations) to encode different power load values. Different values correspond to different pixel intensities, creating a visual representation where variations in power load are captured as intensity changes. This allows the preservation of data characteristics through visual encoding that can be processed by image-based algorithms
Data Source
AI summary
A power load data prediction method and device, and a storage medium are disclosed. In an embodiment, the he method comprises: acquiring historical power load data of a one-dimensional time sequence, the historical power load data including values of corresponding time points; mapping the values of corresponding time points to a coordinate system in which a horizontal axis is a set time period, and a vertical axis is time points within the time period, and performing marking at each mapping point by using predetermined pixel values corresponding to the values to obtain a mapping image, wherein different values correspond to different pixel values; and inputting the pixel values of the mapping image to a trained data prediction model, and acquiring a power load data prediction value output by the data prediction model. The method and device and the storage medium can improve the prediction accuracy of the power load data.