Influenced by many complex factors, it is very difficult to obtain high-performance industrial power load forecasting. The industrial power load forecasting is deeply studied by fusing some machine-learning methods for industrial enterprise power consumers. As a result, a novel power load forecasting method is proposed by taking into account the variation of load characteristics in different regions, industries, and production patterns. First, through the improved K-means clustering analysis, the historical load data are classified as the production patterns to which they belong. Then, the prediction algorithm combining reinforcement learning with particle swarm optimization and the least-squares support vector machine is proposed. Finally, the improved algorithm in this article is used for short-term load forecasting separately by the load data in different patterns after the above processing. The forecasting method in this article is based on data driven with real datasets. The results of the simulation experiment show that the improved prediction algorithm can distinguish the changes in different production patterns and identify the load characteristics of different regions and industries with high prediction accuracy, which has practical application value.

Distributed generation (DG) is gaining importance as electrical energy demand increases. DG is used to decrease power losses, operating costs, and improve voltage stability. Most DG resources have less environmental impact. In a particular region, the sizing and location of DG resources significantly affect the planned DG integrated distribution network (DN). The voltage profiles of the DN will change or even become excessively increased. An enormous DG active power, inserted into an improper node of the distribution network, may bring a larger current greater than the conductor’s maximum value, resulting in an overcurrent distribution network. Therefore, DG sizing and DG location optimization is required for a systematic DG operation to fully exploit distributed energy and achieve mutual energy harmony across existing distribution networks, which creates an economically viable, secure, stable, and dependable power distribution system. DG needs to access the location and capacity for rational planning. The objective function of this paper is to minimize the sum of investment cost, operation cost, and line loss cost utilizing DG access. The probabilistic power flow calculation technique based on the two-point estimation method is chosen for this paper’s load flow computation. The location and size of the DG distribution network are determined using a genetic algorithm in a MATLAB environment. For the optimum solution, the actual power load is estimated using historical data. The proposed system is based on the China distribution system, and the currency is used in Yuan. After DG access, active and reactive power losses are reduced by 53% and 26%, respectively. The line operating cost and the total annual cost are decreased by 53.7% and 12%, respectively.

A popular explicit analytic Borowy 2C PV module model is proposed for power generation prediction. The maximum power point and the open-circuit point which are calculated in this model cannot be equal to the data given by manufacturers under standard test condition (STC). The derivation of this model has never been mentioned in any literatures. The parameter forms of 2C model in this paper are more simplified, and the model is decomposed into a STC sub-model and an incremental sub-model. The STC model is derived successfully from an ideal single-diode circuit model. Relative error estimations are developed to do the conformity error measurements. The analysis results showed that though the biases at those critical points are very small, the conformity will depend on both of the two ratio values I_m/I_sc and V_m/V_oc, which can be used to verify whether 2C model is applicable for the PV module produced by a particular manufacturer.

The distributed control in networked microgrids (NMGs) significantly improves grid-edge voltage performance in distribution systems with high penetration of distributed energy resources (DERs). Even though a prior asynchronous distributed leader-follower control (ADLFC) method can withstand certain levels of communication delay and failure, the peer-to-peer communication of the distributed control method is still vulnerable to cyber-attacks, such as the continuous communication failure (CCF) event led by Denial of Service (DoS) attack. To address this issue, we propose an improved ADLFC method with a cybersecurity enhancement strategy, which can guarantee both the solution convergence and voltage control performance. When the peer-to-peer communication is disconnected under attack of the CCF, the proposed strategy can first detect the CCF event, then utilize redundant channels and backup computations to ensure seamless data transfer via redundant paths and task takeover by backup resources. We also build a DoS attack model to generate the CCF by considering the probability of each DoS attack request, as well as the resource consumption and recovery process of attack targets. The simulation results show that the proposed network security reinforcement strategy can ensure that the voltage deviation remains within a controllable range.

Temporal action localization is a fundamental video understanding task. Meanwhile, due to the complex video background, the varied duration and amplitude of the actions, it is also a considerable challenge. Currently, offline class-semantics representation is the mainstream input of this task since untrimmed videos occupy a large memory, high-quality untrimmed videos and annotations are difficult to access. Because these representations only focus on the class-semantics information, they are sub-optimal for the temporal action localization tasks. At the same time, the exploration of localization-semantics representation is very few due to the high resource consumption. Therefore, it is necessary to improve the detection capability of class-semantics representation directly. As an exploration, we propose the ForeBack module to enhance class-semantics features’ locating performance by augmenting the distinction modeling between foreground and background clips. This module could also eliminate part of the noise of inference probability sequences. Furthermore, we use phased training to learn and use the ForeBack module more effectively. Finally, we reveal the effectiveness of our approach by conduct experiments on THUMOS-14 and the mAP at tIoU@0.5 is improved from 38.8% (BMN action detection baseline) to 47.1%.