Hybrid online model based multi seasonal decompose for short-term electricity load forecasting using ARIMA and online RNN

Sliding limited penetrable visibility graph for establishing complex network from time series

This study proposes a novel network modeling approach, called sliding window limited penetrable visibility graph (SLPVG), for transforming time series into networks. SLPVG takes into account the dynamic nature of time series, which is often affected by noise disturbances, and the fact that most nodes are not directly connected to distant nodes. By analyzing the degree distribution of different types of time series, SLPVG accurately captures the dynamic characteristics of time series with low computational complexity. In this study, the authors apply SLPVG for the first time to diagnose compensation capacitor faults in jointless track circuits. By combining the fault characteristics of compensation capacitors with network topological indicators, the authors find that the betweenness centrality reflects the fault status of the compensation capacitors clearly and accurately. Experimental results demonstrate that the proposed model achieves a high accuracy rate of 99.1% in identifying compensation capacitor faults. The SLPVG model provides a simple and efficient tool for studying the dynamics of long time series and offers a new perspective for diagnosing compensation capacitor faults in jointless track circuits. It holds practical significance in advancing related research fields.

Optimization strategy of power purchase and sale for electricity retailers in a two-tier market

Against the backdrop of the gradual advancement of China's electricity market reform, the number of Power Trading Companies in China has been increasing year by year, and as of October 2022, the number has reached more than 10,000. As an important hub connecting the electricity market and users, electricity retailers face double risks from downstream user load fluctuations and electricity market price fluctuations. Therefore, a reasonable power purchase and sale strategy is very important for an electricity retailer. In this study, a block bidding mechanism is adopted to optimize the clearing of the medium-to long-term market and a DA-RBF neural network is established for spot electricity price forecasting model based on numerical feature similarity to improve the accuracy of electricity price forecasting. Furthermore, the model considers the differences in user demand responses and investigates the optimal power purchase and sale strategy, guided by differentiated time-of-use electricity pricing. The case study analysis demonstrated that the proposed power purchase and sale optimization strategy yields favorable results, improving profitability and enhancing the stability of the power system.

An AQI decomposition ensemble model based on SSA-LSTM using improved AMSSA-VMD decomposition reconstruction technique

Air quality index (AQI) is a key index for monitoring air pollution and can be used as guide for ensuring good public health. Accurate AQI prediction allows timely control and management of air pollution. In this study, a new integrated learning model was constructed to predict AQI. A smart reverse learning approach based on AMSSA was utilized to increase the diversity of populations, and an improved AMSSA (IAMSSA) was established. The optimum parameters with penalty factor α and mode number K of VMD were obtained using IAMSSA. The IAMSSA-VMD was used to decompose nonlinear and non-stationary AQI information series into several regular and smooth sub-sequences. The Sparrow Search Algorithm (SSA) was used to determine the optimum LSTM parameters. The results showed that: (1) IAMSSA exhibits faster convergence and higher accuracy and stability using simulation experiments compared with seven conventional optimization algorithms in 12 test functions. (2) IAMSSA-VMD was used to decompose the original air quality data results in multiple uncoupled intrinsic mode function (IMF) components and one residual (RES). An SSA-LSTM model was built for each IMF and one RES component, which effectively extracted the predicted values. (3) LSTM, SSA-LSTM, VMD-LSTM, VMD-SSA-LSTM, AMSSA-VMD-SSA-LSTM, and IAMSSA-VMD-SSA-LSTM models were used for prediction of AQI based on data from three cities (Chengdu, Guangzhou, and Shenyang). IAMSSA-VMD-SSA-LSTM exhibited the optimal prediction performance with MAE, RMSE, MAPE, and R² of 3.692, 4.909, 6.241, and 0.981, respectively. (4) Generalization outcomes revealed that the IAMSSA-VMD-SSA-LSTM model had optimal generalization ability. In summary, the decomposition ensemble model proposed in this study has higher prediction accuracy, improved fitting effect and generalization ability compared with other models. These properties indicate the superiority of the decomposition ensemble model and provides a theoretical and technical basis for prediction of air pollution and ecosystem restoration.

Anomaly Detection in Automatic Meter Intelligence System Using Positive Unlabeled Learning and Multiple Symbolic Aggregate Approximation

With the development of automatic electrical devices in smart grids, the data generated by time and transmitted are vast and thus impossible to control consumption by humans. The problem of abnormal detection in power consumption is crucial in monitoring and controlling smart grids. This article proposes the detection of electrical meter anomalies by detecting abnormal patterns and learning unlabeled data. Furthermore, a framework for big data and machine learning-based anomaly detection framework are introduced. The experimental results show that the time series anomaly detection for electric meters has better results in accuracy and time than the expert alternatives.