Power Load Forecasting Using a Refined LSTM

Tang, Dedong; Li, Chen; Ji, Xiaohui; Chen, Zhenyu; Di, Fangchun

doi:10.1145/3318299.3318353

Cited by 10 publications

(3 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By comparing the proposed model with LSTM model and AM-LSTM model, as well as various combination prediction models proposed in literature [6], literature [8] and literature [9] as well as VMD_AM_LSTM model without parameter optimization, Forecast the future resource load for the next 60 seconds based on the data from the previous 30 minutes [20]. Figure 4 illustrates the experimental outcomes of load prediction for various models, and the statistical results of prediction error indicators for each model can be seen in Table 2 (average values of 30 experiments are taken).…”

Section: Experimental Results and Analysismentioning

confidence: 99%

“…The results show that the hyperparameter optimization algorithm can boost the forecasting accuracy of the model to a large extent. The forecast outcomes of the Stacked-LSTM model presented in literature [8] and the Refined-LSTM model presented in literature [9] demonstrate that the forecasting accuracy of the Stacked-LSTM and Refined-LSTM is greater than that of the lone LSTM, indicating that the forecasting accuracy of the combined model is higher than that of the single model. Literature [10] assigns weight coefficients to the prediction results of different models such as ARIMA and nonlinear neural network by computing the Mean squared error(MSE) value on the training set, models with low MSE values are assigned larger weight coefficients, and models with high MSE values are assigned smaller weight coefficients or 0.…”

Section: Introductionmentioning

confidence: 98%

See 1 more Smart Citation

Cloud computing resource load prediction based on improved VMD and attention mechanism

Zhang,

Fan,

Miao

et al. 2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Dynamic adjustment of resource supply according to users’ resource load is one of the important technologies to achieve efficient management of cloud computing resources. In order to accurately obtain users’ demand for resource load in the future, based on quantum particle swarm optimization (QPSO), a prediction model QVMD_AM_LSTM was proposed to optimize variational mode decomposition (VMD) and to add attention mechanism to AM_LSTM. A comparative experiment was conducted on the open-source dataset cluster-trace-v2018 from Alibaba Cloud. The outcomes show that compared with LSTM, AM-LSTM, GRU-LSTM, Refined-LSTM, Stacked-LSTM and other existing prediction models, the mean square error of the QVMD_AM_LSTM model proposed in this article decreases by 8-14, and the correlation coefficient rises by 6%-11%. QVMD_AM_LSTM model has higher prediction accuracy.

show abstract

Section: Experimental Results and Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 98%

Cloud computing resource load prediction based on improved VMD and attention mechanism

Zhang,

Fan,

Miao

et al. 2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

show abstract

“…Figure 4 represents the mathematical structure of the recurrent neural network. Thus, the below mathematical formulas ( 2) and ( 3) allow to calculate ht and yt in a recurrent way [25,26]. ℎ𝑡=𝑔h(𝑊𝑖 * 𝑥𝑡+𝑊𝑅 * ℎ𝑡−1+𝑏ℎ)…”

Section: Classification Partmentioning

confidence: 99%

A Hybrid Approach for Web Pages Classification

Hioual¹,

Hemam²,

Hioual³

et al. 2022

ISI

View full text Add to dashboard Cite

Currently, the internet is growing at an exponential rate and can cover just some required data. However, the immense amount of web pages makes the discovery of the target data more difficult for the user. Therefore, an efficient method to classify this huge amount of data is essential where web pages can be exploited to their full potential. In this paper, we propose an approach to classify Web pages based on their textual content. This approach is based on an unsupervised statistical technique (TF-IDF) for keyword extraction (textual content) combined with a supervised machine learning approach, namely recurrent neural networks.

show abstract