Abstract:

Abstract: Power and energy systems are the foundation of human survival and development in modem society. They played an indispensable role in daily production and life. However, significant inefficiencies and wastes found in the process of energy transmission and transformation, such as power generation and energy conversion, could result in increasingly environmental pollution and resource consumption. In order to create a low-carbon energy future, it was a necessary way to optimize the design and operation of energy and power systems aiming at maximization to the economic, environmental and social friendly benefit. After decades of development, Multi- objective heuristics optimization algorithms with characteristics of high flexibility, wide application range and high efficiency, have become crucial tools in the solving various engineering optimization. This paper aimed to systematically reviewing state-of-the-art heuristic based multi-objective optimization algorithms for solving six typical problems in power and energy systems. Comprehensive discussions on the methodologies and a brief insight on future research direction have also been proposed.

Abstract: In the rapid development of integrated energy systems and energy network, power load forecasting played an important role in the economic and safe operation of energy and power systems. The traditional load forecasting modelling methods have been widely used in power systems. However, the simple computational model structure limited by traditional methods could not guarantee the dynamic load prediction accuracy under high randomness and big data background. In recent years, in the context of the continuous upgrading of computing tools and the increasing large-scale of training data volume, the application of deep learning methods in the field of power system load forecasting atrracted extensive attentions. This paper analyzed the applications of various deep learning methods in the field of load forecasting, and revieed the Recurrent Neural Network (RNN) , Long- and Short-Term Memory Network ( LSTM) , Deep Belief Network ( DBN) , and Convolutional Neural Network ( CNN). Compared with the traditional load forecasting method, the deep learning method showed higher prediction accuracy and better robustness to various external influences.

Abstract: The development of electric power system was an important evaluation index of national economic level. With the continuous access of distributed renewable energy and the continuous expansion of power grid scale, the power system was gradually become more and more complicated. The power data with the characteristics of multi-source, heterogeneous, large amounts of typical big data, which puts forward higher requirements for the analysis and processing methods of power problems. Artificial intelligence wass the key to the future science and technology progress. China has to promote the development of Artificial Intelligence to national strategy. The emergence and development of artificial intelligence could provide powerful tools for power system planning and design, simulation, coordination and control, prediction and estimation, diagnosis, and recognition. Power system was one of the main application fields of artificial intelligence in our country. This review began with the policy requirements of artificial intelligence in the world in recent years ； and mainly discussed the main application direction of artificial intelligence in power system, which focused on perception prediction, management control and security maintenance. Then, based on the shortcoming of the application of artificial intelligence itself, the shortcomings of artificial intelligence technology in the field of electric power application wass analyzed. In addition, based on the characteristics of various directions, the problems of artificial intelligence in different directions of electric power application was deeply explored. Finally, suggestions and prospects for the development of intelligent power was put forward in view of the overall development trend of artificial intelligence.

Abstract: With the massive integration of distributed generation and electric vehicles, the problems of power and voltage quality in active distribution network were increasingly shown. Aiming at this issue, the factors affecting the interval control were analyzed in terms of energy storage capacity and power, daily load curve characteristics and unit time firstly. An improved interval control method for energy storage output model was then proposed to solve the problem of multiple charging/discharging operation in one cycle. Considering the correlation of random variables, probabilistic load flow using point estimate method was analyzed to state the influence of distributed generation, electric vehicles and energy storage station on voltage level. Finally, simulation analysis was operated on the improved IEEE-33 node active distribution network system with battery energy storage station. The results showed that the integration of energy storage station could effectively reduce the fluctuation of system power and voltage.

Abstract: The number of faults in distribution network was a direct impact on the operation and maintenance of distribution network and the users power consumption experience. At present, there were few stadies on the prediction of the number of faults in distribution network. To measure the historical dependence of distribution network fault magnitude, the optimal feature subset was selected by using the distance correlation coefficient to investigate the correlation of many meteorological features. Finally, the GRU neural network was trained to predict the fault magnitude of distribution network accurately. The results proved the feasibility of this method.

Abstract: With the massive integration of distributed generation and electric vehicles, the problems of power and voltage quality in active distribution network were increasingly shown. Aiming at this issue, the factors af�1Ffecting the interval control were analyzed in terms of energy storage capacity and power, daily load curve char�1Facteristics and unit time firstly. An improved interval control method for energy storage output model was then proposed to solve the problem of multiple charging/discharging operation in one cycle. Considering the correla�1Ftion of random variables, probabilistic load flow using point estimate method was analyzed to state the influence of distributed generation, electric vehicles and energy storage station on voltage level. Finally, simulation anal�1Fysis was operated on the improved IEEE-33 node active distribution network system with battery energy storage station. The results showed that the integration of energy storage station could effectively reduce the fluctuation of system power and voltage.

Abstract: With the rapid development of interconnected power systems and the large-scale intervention of re�1Fnewable energy, power systems were characterized by interconnectedness and multiple sources. In this circum�1Fstance ,the study in this paper began with the actual needs of multi-source interconnected power system. First�1Fly, a four-area multi-source interconnected power system with the participation of fire, water, gas and renew�1Fable energy was established and each area contained more than one form of power generation. Then, aiming at the randomness and fluctuation of renewable energy generation process, the load frequency was controlled by the grey wolf optimization ( GWO) algorithm based PID controller to stabilize the frequency fluctuation quickly. Finally, the simulation results showed that the proposed controller could achieve good control effect and show strong robustness under different disturbances, regardless of the frequency deviation of the intercon�1Fnected system or the switching power of the tie-line.

Abstract: This study focused on an optimization method that combined simultaneously the reliability and the efficiency of radial power distribution systems ( RDS) , minimized active energy losses, through a process of network reconfiguration. The study based on the failure analysis on network branches, with a special concern on the protection system response to faults and the service restoration procedures, in the emergency state. A non-sequential Monte Carlo simulation based on the branch reliability was used to evaluate reliability of the network configurations. Due to a large number of possible configurations and the need of an effcient search, the optimization was made through an improved genetic algorithm (IGA) . In a first step, the method analyed the RDS considering the absence of investment, and in a second step, the possibility of placing a limited number of new tie-switches in certain branches, according to the definitions made by a decision maker. The effectiveness of the proposed methodology was demonstrated through the analysis of a 69 bus RDS and by comparison against other reported methodologies.

Abstract: In order to provide voltage sag compensation in distribution network, a model to optimize the alloca�1Ftion of distributed energy storage system ( DESS) and a control strategy incorporating DESS and dynamic volt�1Fage restorer (DVR) was formulated and solved have been formulated and solved. In this paper, a double-layer DESS allocation model based on minimize the installation cost of DESS, voltage sag detection equipment and voltage sag of sensitive load was formulated to find optimal configuration of DESS. Then, the minimization cost of DVR, DESS and maximization of voltage for sensitive loads were achieved by joint compensation control model combining DESS and DVR. Moreover, the particle swarm optimization algorithm with random mutatio was employed to seek optimal solution of the proposed model. This approach was tested on the IEEE 33 bus system integrated with DESS and sensitive load. The results revealed that the optimal allocation model could re�1Fduce storage capacity, and dispatch model could successfully meet the demands when considering voltage sag and further reduce the investment of the compensation equipment.

Abstract: To deal with the uncertainty of intermittent energy in the island microgrid model, based on the uncertain cost of adjustable wind power, this paper enemined the different control characteristics of battery energy storage system and diesel generator set, on the premise of ensuring stable power supply of the system, the optimized charge and discharge of energy storage and diesel unit output. And it introduced the adjustable load and other related constraints with the system cost and pollutant emission as the goal, and established the energy optimization scheduling model of the island microgrid. The related constraints such as adjustable load were introduced to establish an energy optimization scheduling model for the island microgrid. On this basis, the improved particle swarm optimization algorithm was used to solve the model. According to the comparative analysis of the micro-grid system optimization models under three different scenarios, the influence mechanism of nine different scheduling interval coefficients on the scheduling results was further analyzed. The effectiveness and feasibility of the wind turbine uncertain cost optimization model and optimization method were verified by case study.

Abstract: Considering the multi-scale characteristics of various scenes for the fire detection, in this paper, we propose a fire detection algorithm based on multi-resolution convolutional neural network. This algorithm leverages the BN_Inception network as the basic structure. Different coarse and fine resolution neural networks complementarily learn the multi-scale visual features of the fire in complex scenes, while paying attention to the background environment, local targets and overall layout of the scene. We also construct a fire dataset covering most of natural scenes, and test our method in this dataset. The experiment proves that the proposed method can achieve better detection results that other methods and can be effectively applied in the real world

Abstract: In order to get the excellent accuracy for port logistic demand forecasting, a combination model based on the BP and RBF neural network was utilized to forecast the logistic demand of Shantou port in this paper. According to the nonlinear change of logistic demand, the BP neural network and RBF neural network were used to establish the single forecasting sub-model separately. And then, the sub-models were combined through the magnitude of the forecasting error to forecast the logistic demand. The simulation was performed by using MATLAB software. Experiment results showed that the combination model could achieve considerably better predictive performances than the single model of BP or RBF neural network. It could reduce the mean absolute percentage error and root mean square error in the logistic demand of Shantou port. These results indicated that forecast combination could improve the precision of the single neural network model for port logistic demand forecasting, and could help the decision maker in relevant port sector make proper decisions.

Abstract: In this paper, the algorithm principles of Gaussian mixture model, HOG+SVM classifier and Haar+ Adaboost classifier were exploved. A pedestrian detection algorithm based on foreground extraction and pattern recognition was proposed. The background modelling was executed by using Gaussian mixture model and then the moving object was entracted by using foreground modeling methods. The pedestrian detection hased on the moving objects and face recognition on the objects were execnted. By this, the misjudgment problems was solved based on background modeling methods and efficiency problems based on statistical learning methods. The experimental results showed that the new algorithm could greatly reduce the missed detection rate compared to using the pattern recognition algorithm alone, and it performed well in terms of running speed and detection rate.

Abstract: