Buildings account for substantial global energy consumption, with heating, ventilation, and air conditioning (HVAC) systems as major contributors. We study the setpoint schedule optimization of HVAC systems that minimize both energy costs and occupant discomfort. Since building performance simulation (BPS) tools provide high-fidelity models of building dynamics, integrating simulation with optimization is expected to obtain an effective schedule for building energy management. Consequently, many simulation-based optimization methods that integrate BPS into optimization processes are proposed. However, these methods still face challenges due to non-analytical system dynamics, computational complexity, and the lack of theoretical convergence guarantees. To address these challenges, a Lagrangian relaxation-based simulation optimization (LRSO) method is developed in this paper. A dynamic linear surrogate model iteratively refines itself with simulation outputs, balancing tractability and accuracy. Within Lagrangian relaxation framework, the problem is decomposed into simulation and optimization subproblems, which can be solved in a coordinated and decomposed way. The surrogate subgradient method further ensures the convergence. Experimental results demonstrate its superior performance in minimizing energy cost and occupant discomfort across all test scenarios, with computational times suitable for real-time scheduling.

With the popularity of electric vehicles (EVs) and renewable energy sources (RES), the flexibility of charging and discharging of EVs and the intermittency of RES have brought challenges to building operations. Considering the mobility of EVs as commuting tools between buildings and the uncertainty of RES, it is of great practical significance to coordinate multiple buildings with RES and EVs on the premise of meeting the state of energy (SOE) requirements of the future trip. We formulate this coordination problem as a stochastic centralized mixed integer linear programming problem. A polyhedral convex set is constructed to describe the SOE uncertainty of EVs. New nonanticipative constraints (NCs) are derived through forward recursion based on constructed scenarios to guarantee the all-scenario-feasibility (ASF) and nonanticipativity of the decision. A Lagrangian relaxation-based decentralized all-scenario-feasible (LR-DASF) algorithm is developed to solve the centralized optimization problem in a decomposition and coordination way. In this method, the optimal ASF solution can be obtained with a fast convergence rate by updating Lagrangian multipliers without solving all subproblems with NCs. The performance of the LR-DASF algorithm is verified by numerical results, which shows that the algorithm can guarantee the ASF of the solution, as well as promote computational efficiency. Note to Practitioners—EVs as energy storage devices bring energy exchanges between buildings accompanying the mobility of EVs which is an opportunity to improve the energy efficiency of multiple buildings. However, as commuting tools, the SOE of EVs must be guaranteed to be larger than the trip requirement over the randomness of RES generation. Furthermore, solving the coordinated optimization problem of multiple buildings with RES and EVs still faces computational complexity challenge due to the spatio-temperal coupling between EVs and buildings, which may lead to costly computational effort in the premise of guaranteeing the feasibility and nonanticipativity of the decision over the uncertainties in practice. Therefore, in order to overcome the above challenges, an LR-DASF algorithm is developed in this paper to solve the coordinated optimization problem of multiple buildings with RES and EVs. Based on the algorithm, for the system operator, it updates and broadcasts the Lagrangian multipliers information to the local coordinators of buildings. For each building, the local coordinator can make ASF decisions based on new NCs with the information obtained from the system operator independently to guarantee the SOE requirement. The method developed in this paper can make faster optimal decisions without perceivable degradation in accuracy and guarantee the SOE requirement of EVs simultaneously, to meet the requirements of feasibility and computational efficiency of decision making in practice. It is conducive to the future application of LR-DASF in the coordinated optimization of buildings and EVs at the city or regional scale.

Battery swapping station (BSS) can provide fast battery swapping and flexible battery charging in off-peak hours, it is thus beneficial for electric vehicles (EVs) and power grid in terms of battery life extension and power grid regulation. However, this increases the charging scheduling complexity in BSS since the batteries are not necessarily required to be charged immediately as they arrived. This problem becomes challenging in the presence of nonlinear battery charging characteristics and demand/supply uncertainties. Since it is difficult for the traditional learning-based methods to deal with the constraints caused by nonlinear charging characteristics, low sampling efficiency and unstable training issues can occur. In order to solve these issues, we present a novel deep reinforcement learning (DRL) approach. In contrast to the traditional approaches where the battery charging characteristic is simplified to a constant-current or constant-power process, we propose an equivalent circuit model (ECM) to capture the nonlinear charging characteristics. In ECM, the battery’s open-circuit voltage (OCV) is a function of its state of charge (SoC), as a result, the upper bound of the charging/discharging power of battery is influenced by its SoC. Then we construct a constrained Markov decision process (CMDP) model and propose a Beta distribution-based DRL approach with a continuous action mask (AM) to improve the sampling efficiency and consistency of the training process. Numerical experiments show that our new approach can provide better results in terms of operation cost and quality of service (QoS) in comparison with other state-of-the-art DRL methods.

Renewable energy sources and electric vehicles provide an effective way to reduce the energy cost of an enterprise microgrid. However, the uncertainties of renewable energy sources and the time coupling characteristic of electric vehicles bring great challenges of non-anticipativity and feasibility for supply-demand coordination. To satisfy the non-anticipativity, we develop a supply-demand coordination optimal model using pre-scheduling method with virtual re-scheduling. In this model, the current decision only depends on the current and past realizations of random variables. Furthermore, we enhance the model with time-coupled robust constraints to guarantee the feasibility of the strategy under all possible realizations of the random variables. These time-coupled robust constraints bring high computational complexity to solve this model. So, we develop the method of combining forward recursion and backward recursion to decouple these time-coupled robust constraints in time. In this way, the coordination model is transformed to a mixed integer linear programming (MILP) model which can be efficiently solved. Finally, numerical test based on a real case is analysed and the results show that the energy cost of the enterprise is about 136129$\$ $if the flexible load is about 20% and load shifting and generators rescheduling can reduce the energy cost more than 6%. Note to Practitioners—This study is encouraged by the challenging problem caused by the multi-distributed energy introduced into an enterprise microgrid. In enterprises, as the large-area flat workshop roof assists in convenience for photovoltaics’ development and the EVs are widely used, the issue to best utilize renewable energy and EVs shows vital significance in reducing the energy cost. However, there exist the following three main challenges: (1) the non-anticipativity of the model, (2) the solution’s feasibility under all possible realizations, and (3) the effectiveness of the solution method. For the concerns of non-anticipativity, we develop the model using a pre-scheduling model with virtual re-scheduling in which the current decision only depends on the current and past realizations of random variables. To handle the second challenge, an ideal of scenario model with robust constraints is developed considering both feasibility and economy. In order to solve the model with robust constraints, the all-scenario-feasible method and a combination of the forward recursion and backward recursion method are used to deal with time-independent and the time-coupled robust constraints, respectively. The numeric results demonstrate that load shifting and generators rescheduling can reduce the energy cost more than 6%, and using the method with the forward and backward recursion process can reduce the energy cost more than 9%.

Green and sustainable development of Internet data centers (IDCs) has attracted more attention in both academia and industry. Full utilization of renewable energy sources is widely known as the most effective way to supply electrical and thermal energy while reducing carbon emission. However, the integration of renewable energy into IDCs is still challenging due to the mismatch between uncertain renewable supply and time-varying demand requirements, and high requirement of operation reliability against IDC failures. Therefore, in this paper a hydrogen-water-based energy (HWBE) system is developed and its integrated planning-and-operation problem is formulated as a mixed-integer linear programming problem to determine the optimal capacity of energy facilities in the HWBE system with considering IDC operation reliability. A hybrid physics-based and data-driven method is developed to accurately capture the electrical and thermal energy consumption characteristics and their coupling which are the basis for the optimal planning of the HWBE system. Furthermore, a Benders decomposition-based reliability improvement algorithm is developed to enhance the operation reliability, which decomposes the problem into the planning problem with normal operation as the master problem and the operation problem with IDC failure as the subproblem. The reliability can be enhanced using the solution obtained by the master problem with the feasibility cut obtained from the subproblem. Numerical results show that the developed HWBE system is energy-efficient with low carbon emission, since the power usage efficiency of IDCs could be as low as 1.09 and the carbon emission could be reduced by 74.9% as compared by the electricity-driven IDC energy system.Note to Practitioners—This paper focuses on the integrated planning-and-operation optimization of an HWBE system for the application in IDCs. We improve the energy consumption model of IDCs based on a hybrid physics-based and data-driven method, which can describe the interaction between the dynamic thermal process and electricity consumption of IDCs. In this way, both the high accuracy of the physics-based model and the lower computational effort of the data-driven method could be simultaneously achieved in the energy consumption model. Furthermore, in practice, the optimal planning problem of IDCs is necessary to take into account the operation reliability against data center failures, since the capital expenditure of the backup energy devices is generally significant. This means that a trade-off between the solution accuracy of the planning problem and the computational complexity caused by the operation problem should be considered. Therefore, we develop a Benders decomposition-based reliability improvement algorithm to address the trade-off mentioned above. This technique can integrate the feasibility cut obtained from the operation problem with IDC failure into the planning problem, in order to improve the operation reliability against the supply-demand mismatching and IDC failures while reducing the capital cost, as compared to the system designed by the conventional redundancy standard. Numerical results show the effectiveness of the developed method which can make full use of renewable energy sources and support the green and sustainable development of IDCs.

The penetration of plug-in electric vehicles (PEVs) has greatly increased over the past few years. By using vehicle-to-grid (V2G) technology, PEVs can be used as “mobile batteries” in a microgrid. Here, we aim to coordinate the V2G dispatch with traditional energy management in a corporate energy system (CES). To do so, a two-stage robust optimization (RO) model is built with respect to uncertainties in the CES, e.g., photovoltaic (PV) power. Particularly, relationships between the working time schedule and PEVs are investigated and analyzed for the first time, and a novel PEV aggregator model, i.e., shift-working V2G, is presented. The shift-working V2G model provides beneficial characteristics, like weakened randomness and stable storage capacity. A quantitative method to evaluate the V2G capacity is then presented. An analytical solution methodology is also proposed, which can equivalently convert the robust “min-max-min” model to a single-level mixed-integer linear programming (MILP) model. Case studies are conducted for an iron and steel company in Shanghai, China, with almost 40 000 PEVs. The results show that V2G integration can significantly improve the load-tracking ability of CES and help reduce the energy cost, although the V2G cost is considered. The computational efficiency is also improved compared with the existing methods. Note to Practitioners-This article is motivated by the problem of using the battery storage capacities of plug-in electric vehicles (PEVs) in a corporate energy system (CES) such as an iron and steel plant, aiming at minimizing the energy cost. In existing studies, behaviors of PEVs (e.g., arriving or leaving time) have usually been elusive since they are directly decided by drivers, and thus, it is difficult to determine how much storage capacity PEVs can provide. However, in a CES where a shift-work regulation is implemented, employees should arrive or leave punctually during each shift, and the same is true of their PEVs. Based on this fact, PEVs within one shift could show weakened randomness and stable storage capacities. In this article, the influences of the shift-work regulation on the PEVs are fully analyzed, and the shift-working V2G provides an easy way to integrate PEVs into the CES. In practice, the shift-working V2G model can be applied to any energy system that also implements a shift-work regulation, such as a fire station.

In this paper, we present a system to detect and count the number of vehicles in traffic surveillance videos based on Fast Region-based Convolutional Network (Fast R-CNN). Fast R-CNN is a state-of-the-art object detection network, which takes an entire image and a set of object proposals as input, produces bounding-box positions with probability estimates over object classes as output. First, we fine-tune a pre-trained Fast R-CNN net with images captured from traffic videos for accuracy improvement. Second, we define a series of rules of bounding boxes screening for vehicle counting. The proposed system takes around 3 seconds per image to count vehicles on a GTX970 GPU, and then records the corresponding number of vehicles into a database for traffic flow analysis. Experimental results demonstrated that the proposed system can provide significant improvements on the detection accuracy. In addition, experiments on challenging videos with occlusions or full of vehicles show that the proposed system works effectively.