Search

In this paper, a new control architecture of virtual power plants (VPP) is proposed for the frequency regulation (FR) in main grid. Firstly, to address the power prediction of aggregation nodes (AN) with spatio-temporal correlation under VPP, a novel combined model based on graph convolutional networks and bi-directional long and short-term memory is established. Further, considering the effect of prediction uncertainty, a power shortage measurement method based on interval estimation is proposed to determine the effective power of AN. Secondly, a dispatch-based control strategy is introduced to enable VPP to respond to regulation commands from the main grid. Additionally, the influence of transmission line power loss on FR is considered, which can be eliminated by applying the developed power compensation design scheme based on proportional allocation. Finally, numerical simulations illustrate that the proposed method enhances prediction performance by 68.65%, taking mean absolute error as an example, while ensuring the accuracy and reliability of power prediction. Also, the adopted control strategy can effectively utilize VPP to provide FR services for the main grid. Moreover, the application results in a real scene are better than those of benchmark methods, which also demonstrates that the proposed approach has a high practical application potential in power system. • Predict node power using a combined model based on the GCN-BiLSTM. • Propose a method for power shortage measurement for the forecast uncertainty. • Design a power compensation scheme for eliminating the effect of network loss. • The combined method enhances the prediction performance than existing methods. • The frequency problem of system improves through utilizing virtual power plants. [ABSTRACT FROM AUTHOR]

The rapid growth of electric vehicles (EVs) imposes great challenges to the flexible management and economic dispatch of virtual power plants. To overcome these obstacles, this paper proposes a bi-level economic dispatch strategy based on EVs aggregation. First, an aggregation method based on improved artificial bee colony Kmeans clustering algorithm is presented to aggregate the EVs with different dynamic characteristics in the upper level. Aiming at the issue that multi-parameter weights and aggregation effects are difficult to determine in the clustering process, information entropy and Silhouette metric are proposed to enhance the accuracy of clustering. Furthermore, pre-dispatch is conducted based on aggregation information and electricity market prices. Next, in the lower level, the charging demands information is utilized to formulate specific dispatch strategies for EVs users, and the pre-dispatch deviation in the upper level is modified. Finally, the simulation results show that the proposed aggregation method improves the clustering effect by 22.4% and guarantees the quality of aggregation and the flexibility of system management. Besides, the proposed strategy can reduce the total system operation cost and the computation time by 5.83% and 55.95% respectively and ensure the economy of the system operation and the computational efficiency of the dispatch. • Aggregate electric vehicles using a method based on improved Kmeans algorithm. • Formulate an allocation model considering the charging needs of users. • Construct a bi-level optimal dispatch model for developing effective dispatch plans. • The aggregation method enhances the aggregation quality than existing methods. • The flexibility of system dispatch improves through aggregating electric vehicles. [ABSTRACT FROM AUTHOR]

• Analyze the correlation between power load data and relevant search engine indices. • Take search engine indices as input variables to forecast electricity load. • Use Hodrick-Prescott filter to decompose the time series of search engine indices. • Propose a LSTM model based on quantum-behaved particle swarm optimization algorithm. Accurate electricity load forecast plays an important role in the operation of a power system. Many factors influence the electricity load data such as air temperature, humidity and holidays, and they are taken as the explanatory variables in load forecasting cases traditionally. Search engine indices, a variable may be related to load data which has been never considered before in load forecasting, is discussed and utilized to increase the accuracy of power load prediction in this paper. Spearman's correlation coefficients and Granger test results verify the correlation between Google Trends (GT) and electricity load data. A methodology for processing GT time series with Hodrick-Prescott filter is proposed. To forecast electricity load with an adaptive network model in such a novel situation, we propose a long short-term memory neural network model based on quantum particle swarm algorithm. The performance of load forecast for Long Island region taking GT and weather data as input variables is compared with that taking only weather data as input variables, which shows that the introduction of GT improves short-term forecasting effectiveness significantly. [ABSTRACT FROM AUTHOR]