Data center, as an important carrier of digital infrastructure, has been growing rapidly in the number and scale, which results in increasing system energy consumption and operating costs. The integrated energy system (IES) is a complex system which takes the power system as the core and covers the production, supply and marketing of energy such as electricity, gas, cooling and heat. Its optimized operation can realize comprehensive management and economic dispatch of multiple energy sources. Using IES to replace the original energy supply system is an effective solution to reduce the operating cost of the data center. At present, the typical architecture of IES has not been clearly defined, and there is little research on the optimization operation of IES based on data center. Firstly, this paper introduces the typical physical framework of IES and data center-based IES. Secondly, it analyzes the physical model of the coupled operating equipment and the application of energy hub and its extended model in the system. On this basis, this paper summarizes and sorts out the current status of IES and the optimized operation of IES for data centers, and proposes an operation method based on data centers, including multiple energy input, coupling and output. Finally, this paper looks forward to the key development directions and potential research contents of energy uncertainty and load forecasting, energy dynamic modeling, optimized scheduling strategies, and benefit evaluation in IES based on data center, with a view to provide reference for future related project landing, simulation research and development.

At present, the coverage of measuring equipment in distribution network is low, so only part of the nodes’ load data can be collected in real time. This situation makes it impossible to use the optimization based on power flow calculation in the real-time reactive power optimization of distribution network. Considering the above situation, this paper proposes a data-driven reactive power optimization method based on partial real-time visible distribution network. According to the historical operation data, the optimal power flow is used to generate the reactive power optimization strategy offline, and the mapping between the real-time measured node load data and the reactive power optimization strategy is established by training the neural network to realize the real-time reactive power optimization of the partial real-time visible distribution network. Finally, in the modified IEEE 33-bus system, the proposed method is compared with the 9-zone diagram method to verify the effectiveness of the proposed method.

Load portrait modeling of power user is an important user-oriented method to create differentiated labels by mining the load characteristics in power consumption data. Most of the existing research focuses on the study of portrait methods, but lacks comprehensive load characteristic label system. This paper proposes a general method of load characteristic analysis based on data-driven. The load characteristic label system is constructed from power consumption regularity, smoothness, load control capability and epidemic impact, which are most concerned by dispatching department. Firstly, the typical load curve is extracted from massive actual load data by using fuzzy C-means clustering algorithm. Considering the power consumption characteristics of each industry from above four aspects, a comprehensive load characteristic label system and the load characteristic portrait models of different power users are established. Secondly, the load characteristic label is refined and every definition and calculation method of corresponding index is given. Furthermore, the index boundary is determined by fuzzy clustering algorithm, and the smoothness label is scored by entropy weight method. Finally, the data of typical users in different industries are analyzed from an example, and universal index boundaries are given, which provide a new idea for load modeling of users in various industries.

Multi-park integrated energy system can significantly improve the operation economy by complementing each other with multiple energy sources. However, the complex interactions between parks and multi-energy coupling decisions can bring challenging problems such as large decision space and difficult convergence of algorithms to the energy management of multi-park integrated energy system. To solve the above problems, an energy management method based on modified deep Q network (MDQN) algorithm for multi-park integrated energy systems is proposed. Firstly, the external meteorological data and historical interactive power data independent of the park are used to construct a long short-term memory (LSTM) deep network-based external interactive environmental equivalence model for each park integrated energy system, which reduces the computational complexity of the reinforcement learning reward function. Secondly, an improved DQN algorithm based on k-first sampling strategy is proposed to replace the greedy strategy with k-first sampling strategy to overcome the inefficiency of exploration in large-scale action spaces. Finally, the results are validated in an algorithm containing three integrated energy systems in the park, and show that the MDQN algorithm has better convergence and stability compared with the original DQN algorithm, while it can improve the economic efficiency of the park by 29.16%.

As renewable energy generations represented by wind and photovoltaic power connect to the distribution network (DN) through multiple micro-grids (MMGs), the uncertainty will bring challenges to the reliability and economy of the operation of the DN and MMGs. In response to this, this paper proposes a data-driven robust dispatch method for DN and MMGs considering correlation between wind and solar output. Firstly, a distributed dispatch method is adopted to establish the dispatch model of the DN and a two-stage dispatch model of the MMGs, with the tie-line power as the coupling parameter of the two. Aiming at the uncertainty of renewable energy output as well as the temporal and spatial correlations, the wind-solar output ellipsoid uncertain set is constructed based on the data-driven algorithm, thereby establishing the two-stage data-driven robust dispatch model of the micro-grid. Finally, an improved column and constraint generation algorithm based on extreme scenarios is proposed to solve the robust dispatch problem of the micro-grid, and the analytical target cascading method is used to solve the overall dispatch problem of DN and MMGs. The simulation results show that the proposed method can capture the spatial-temporal correlation between wind and solar, improve the economy of dispatch while ensuring the robustness of the DN and MMGs dispatch, and has good convergence.

Under the dual carbon target, it is imperative to develop integrated energy system planning and research, and the flexible loading, unloading and configuration of the user-side integrated energy module is one of the emerging research objects. In this paper, a two-stage integrated energy planning model is established with the introduction of multiple hybrid energy storages and carbon capture devices, which achieves the complementary use of energy and carbon emission recovery and utilization in the system. To deal with new energy output such as photovoltaic and wind power, and user load uncertainty, extreme scenario is proposed. According to ellipsoid set data driven method of robust optimization in uncertainty, the paper accurately describes the correlation between variables and improves the conservation of traditional robust optimization result. Compared with the traditional method with mass probability calculation, a simpler ellipsoid endpoint extraction method is used to obtain extreme scenes. In this paper, the steps of column and constraint generation (CCG) solution are improved by using the advantage of ellipsoidal extreme scenarios to avoid the complicated duality processing of sub-problems. Finally, by example simulation and comparison with the traditional interval uncertain set robust optimization method, it is proved that method proposed in this paper has advantages in reducing economic cost, energy saving and reducing carbon emissions.