Short‐term load forecasting method based on deep neural network with sample weights

Abstract

Summary The reform of power market has presented new challenges to short‐term load forecasting (STLF), and the accuracy of forecast results is of great significance to the orderly and efficient operation of the market. To improve the accuracy of load forecasting, an STLF method based on deep neural network (DNN) with sample weights is proposed. By filtering samples and assigning corresponding weights to different training samples, this method effectively improves the forecasting accuracy of the DNN model. Two important steps of this method are as follows: (a) calculating the distance between the samples and the forecast day to measure the similarity between them, and the training samples are selected based on the distance; (b) after selecting training samples, corresponding weights can be assigned to training samples according to the distance between the training samples and the forecast day. This allows DNN to focus on the key samples. Finally, we carried out simulation analysis by using the actual load data of Guangdong Power Grid in January 2017. The results show that the proposed method can effectively improve the accuracy and reliability of load forecasting.

A CNN‐BiLSTM based deep learning model for mid‐term solar radiation prediction

“…The LSTM based short term solar radiation prediction is proposed with solar radiation and clearness index of cloud as the input for the location of Atlanta, New York, and Hawaii. 20 However, in this work, hybrid DL models are not compared, and only standard errors are compared as a benchmark for comparison of proposed LSTM based prediction. The CNN and LSTM based hybrid DL model is proposed for solar power forecasting with roughly estimated weather data.…”

“…The DL models are studied for different regions and with a different configuration for solar radiation and power forecast. The LSTM based short term solar radiation prediction is proposed with solar radiation and clearness index of cloud as the input for the location of Atlanta, New York, and Hawaii 20 . However, in this work, hybrid DL models are not compared, and only standard errors are compared as a benchmark for comparison of proposed LSTM based prediction.…”

An attention‐based CNN‐LSTM‐BiLSTM model for short‐term electric load forecasting in integrated energy system

“…These energy conversion characteristics are difficult to summarize using traditional artificial feature extraction methods. With the increasing amount of data in power grid and the rise of artificial intelligence technology, machine learning and deep learning have been widely used in short‐term load forecasting of the power system, such as support vector machine regression (SVR), 14 artificial neural networks (ANN), 15‐17 etc. The use of artificial intelligence technology can significantly improve the forecasting accuracy.…”

NARX: Contribution ‐ factor‐based short‐term multinodal load forecasting for smart grid

“…1 This huge amount of realtime data collected from AMI opens the way for unprecedented opportunities such as demand-side management, load forecasting, resource allocation, maintenance scheduling, system planning and operation, etc. 2 However, smart metering generates data at such a huge rate and volume that it surpasses the analyzing ability of many conventional systems. As a result, new techniques based on big data analysis 3 and machine learning are used to analyze the collected data.…”

“…Various techniques have been discussed in the literature for STLF of a single node or substation [2][3][4][5][6] but a few deal also deal with multinodal problems. [7][8][9] In power system applications, we need to forecast the load at multiple nodes simultaneously for the distribution network.…”