Francesco Conte, Fabio D’Agostino, Federico Silvestro
IMAGE LICENSED BY INGRAM PUBLISHING
Modern ports are expected to play a key role in the transportation chain, being transformed into smart grids and smart energy hubs where electric energy needs predominate. Cold ironing (CI), also known as onshore power supply (OPS) or alternate marine power, is indeed one the most energy-demanding applications. The power demand of a berthed ship depends on its characteristics and may vary from hundreds of kW, for container ships, to tens of MW, in the case of cruise ships.
On one hand, the environmental benefits of CI are today well recognized in terms of reduction of polluting emissions, such as carbon dioxide (CO2), nitrogen oxides (NOx), sulphur oxides (SOx), and particulate matter. National grids’ emission factors are lower than the ones of the onboard generators, which typically employ diesel engines as prime movers. Moreover, during berthing time, the specific fuel oil consumption of generators increases because of the loading being lower than the point of maximum efficiency. As a result, emission factors further increase.
On the other hand, the economic convenience of purchasing energy from the port’s grid is not ensured. The cost of energy (COE) produced using onboard generators is often considerably lower than the one absorbed from national grids. Moreover, the shore connection infrastructure has high investment costs that can result in a further increment of COE at the shore side, in accordance with the specific regulatory and policy frameworks. To fight this issue, several European ports are applying extra taxes to all non-CI users, so that the COE produced on board becomes higher than the one absorbed from the CI.
In this context, the integration of renewable energy sources (RESs) within the port’s jurisdiction represents an appealing opportunity to reduce the COE at the shore side. Furthering this concept, storage systems are also becoming of interest as a means to unlock the potential of RESs, stochastic by nature, introducing a new degree of freedom in power and energy management.
Besides the use of battery energy storage systems (BESSs), in recent years, hydrogen has gained a lot of interest as an effective alternative to store energy. The goal of using hydrogen in a smart port framework is twofold: it can store energy from RESs providing the so-called green hydrogen, and it can be used to supply ships equipped with fuel cells (FCs). Among the technologies available today, in fact, the proton exchange membrane FC (PEMFC) has been recognized as an opportunity for onboard applications. PEMFCs are fueled by pure hydrogen, they work at low temperatures (around 70 °C) with a relatively good efficiency (around 45%), and they do not produce local polluting emissions. Several types of hydrogen storage (HS) exist, such as compressed hydrogen (CH2)-, liquid hydrogen (LH2)-, or metal hydrides (MH)-based systems. The analysis of this aspect is beyond the scope of this work; however, it is possible to conclude that LH2 technology is well suited for onboard applications because of the high energy density, while for terrestrial applications, CH2 and MH storages are usually considered as the common practices.
In view of these considerations, smart ports can play a pivotal role in the world’s decarbonization challenge as natural energy hubs for different types of energy vectors. A proper mix of resources and storage systems provides the background to decarbonize both marine and industrial sectors, strongly connected by commercial ties. However, this new port’s smart grid general framework results are complex, and its design requires the energy management strategy to be properly developed. Therefore, this work focuses on the optimal management strategy for a smart port, organized as a multienergy hub (MEH).
Our general idea of an MEH port is depicted in Figure 1. There are two quays: one equipped with a shore connection (electrical quay) and one equipped with a hydrogen supply system (hydrogen quay). The electrical quay supplies conventional, diesel powered, all-electric ships (AESs) with shore connection capability, which realizes the main electrical load. The hydrogen quay supplies zero-emission ships (ZESs), where onboard diesel generators (DGs) are substituted by PEMFCs or hybrid AESs (HAESs). In both cases, the ships are equipped with FCs and an HS. The hydrogen is produced through an electrolyzer (Ely), installed within the port area, that uses the energy from RESs, so that green hydrogen is realized. A PEMFC system allows the conversion again of hydrogen into electrical energy ashore. The port’s HS, the Ely, and the PEMFC constitute the port’s hydrogen energy storage system (HESS). To provide an even faster flexibility on the electrical side, the employment of a BESS is also assumed. In this general view, other electrical and hydrogen loads can also be considered, such as a recharge station for plug-in electrical vehicles (PEVs), port utility buildings, and a hydrogen fueling station for hydrogen-powered vehicles.
Figure 1. The MEH port general configuration. AES: all-electric ship; HAES: hybrid AES; PEVs: plug-in electrical vehicles.
With this work we discuss the principal issues related to the optimal management of an MEH port. First, the optimization problem is introduced, taking into account all the relevant properties and limitations of each component of the MEH port. Then different control strategies employing the optimization problem are discussed, with particular focus on the management of uncertainties introduced by RESs and loads.
Finally, a study case of optimal management is provided and discussed, for a port where only a subset of the mentioned components is present. The management is realized by a stochastic model predictive control (MPC) algorithm. The literature provides several uses of MPC for smart cities with power-to-gas devices and, in general, for MEHs. The peculiarity of the methodology proposed here is that uncertainties of RESs are considered to maximize the port economical income, by dispatching the compensation of forecast errors to the available storage systems, i.e., BESSs and HESSs.
Thanks to the information and communication technologies, all the components and apparatuses of a port can be monitored, forecasted, and/or controlled in real time. A port designed as an MEH, being composed of heterogeneous components, needs to use all the data from the field to be managed so that all the available resources are optimized in their operation. The real-time operation of an MEH is exploited by an MEH management system (MEHMS), which has the following main targets:
It is worth remarking here that the present discussion is focused on the port operation. Thus, sizing and consequent CAPEX is not considered here. Operating costs to be taken into account in the economical optimization operated by the MEHMS are the OPEX of all the different components and the cost of the electrical energy purchased from the main grid. Being an MEH including RES generation, shore connection, and hydrogen bunkering, the port will also have economical incomes by selling energy to the market and to the AESs berthed at the CI quays and by selling self-produced hydrogen to ZESs or HAESs. For the sake of readability, in the following, we will refer only to ZESs, which actually also indicate HAESs, except when specific differentiation is required.
As is well known, an optimization algorithm is composed of a cost function and a set of constraints. The cost function, which is conventionally minimized, sums all operating costs and subtracts all incomes. The purchase price depends on the agreement between the port administration and the energy retailer, and as for any electricity supply agreement, it can be constant or time-varying. Energy sales tariffs also depend on an agreement, but a time-varying component determined by the energy market clearing price is present. Concerning the tariffs applied by the port to supply AESs and fuel ZESs, they will be established by the port administration. OPEXs depend on the type of devices included in the smart port. They can be constant and independent of the use made of the device or be time-varying and dependent, sometimes nonlinearly, on the way they are used. In this second case, the cost function will include terms that allow the minimization of OPEXs. According to these considerations, the MEHMS cost function is generally time-varying and possibly (depending on the OPEXs) nonlinear.
Constraints are used to model the dynamical behavior and the technical limits of the different components of the MEH port. Beyond the classical distinction between generators and loads, we can distinguish three classes of devices:
Fully and partially controllable components are the ones that make available to the MEHMS the flexibility needed to optimize the port operation, whereas only predictable components introduce uncertainties that require suitable treatment.
Storage systems are fully controllable components. In our idea of an MEH port, they can be BESSs and HESSs, the latter being composed of an Ely, an FC, and an HS. BESSs are usually characterized by the battery capacity, the rated charge/discharge power, and two different charge and discharge efficiencies. HESSs are characterized by the HS capacity, the rated powers, the efficiencies, and the (nonzero) technical minima of the Ely and the FC. Both storage systems are therefore usually modeled by mixed-integer linear constraints. In the case of BESSs, integer variables are required because of the different charge and discharge efficiencies; for HESSs, integer variables are required to model the possible switching off of the Ely and the FC. Given the mentioned technical limits, the MEHMS can establish the power exchange of BESSs and HESSs.
RES power plants, such as wind farms (WFs) and photovoltaic plants (PVs), are partially controllable since the unique control available to the MEHMS is the reduction of their generation from the maximal one, which is determined by wind velocity and solar irradiation (RES curtailment). As mentioned, an objective of the MEHMS is to avoid such a reduction. The data used by the MEHMS about RES power plants usually are time series, provided by suitably developed forecasting algorithms, included in the optimization problem as linear constraints.
Loads are, in general, partially controllable or only predictable components. Electrical and hydrogen quays are only predictable since their demand of electricity and/or hydrogen must be mandatorily satisfied. Such demands can be known in advance using again suitably designed forecasting algorithms that determine their predictions based on the berth scheduling and the class of the berthing ships. In the case of ZESs, it is possible that the quantity of hydrogen to be supplied will be declared in advance, making available, in this way, an exact prediction to the MEHMS. In any case, as for RES power plants, predictions are provided as time series, included in the optimization problem as linear constraints. It is worth remarking here that to satisfy the electrical demand of AESs is fundamental to minimize emissions. Indeed, if such a demand is not satisfied, AESs will be forced to activate their onboard DGs.
All other loads, such as the port buildings or PEV recharge stations, may be only predictable or partially controllable. For example, buildings can be equipped with some flexible load, such as air heating and cooling systems, whose energy demand can be modulated by the MEHMS. PEV recharging stations can also offer some flexibility to the MEHMS. Especially if they are used to recharge the port’s service vehicles, we can assume that a recharge scheduling, expressed in terms of amount of energy and end-of-recharge time, can be known in advance. In this case, the MEHMS may have the possibility to modulate the recharging power within the prescribed time interval. In any case, load demands are modeled by time series, computed by forecasting algorithms or based on a known scheduling, included in the optimization problem as linear constraints. When some load control is available, it is usually modeled by mixed-integer linear constraints. For example, PEV recharging may offer not a continuous modulation of the recharging power but a multilevel or simply on–off regulation.
Once the MEH port optimization problem is formalized, there are several possibilities to employ it in managing the port operations. First of all, we should consider that within the MEH port, there are dynamics with different time scales. For example, RES generation can significantly vary within 1 h, whereas bunkering of ZESs can last many hours. HS must be managed with considerable forethought (one day ahead at least, depending on the storage capacity) since the generation capacity of Elys is not, at least for the moment, so high and fast to produce in real time the amount of hydrogen supposedly required to refuel one or more ships. Therefore, the most efficient solution is to realize a hierarchical optimization, where a first-level optimization problem is defined along the time horizon of several days (one week at least), with a time granularity of 12–24 h, whereas a second-level optimization problem is defined along the time horizon of one day, with a time granularity of 1 h or less. In this way, the first-level optimization will allow the programming of lower dynamics, whereas the second-level optimization will manage faster dynamics and eventually correct the first-level program. To realize this control architecture, two types of forecasts are also required, long-term for the first-level control and short-term for the second-level control. In this work, we will discuss the results of a possible second-level control algorithm.
In the MEH port optimization, forecasts play a crucial role since, as described before, there are many not fully controllable components. To summarize, potentially, there will be forecasts for RES generation, AES electrical demand, ZES hydrogen demand, and all other electrical loads (e.g., buildings and PEV recharging stations). Regardless of the employed forecasting methods, which are not the focus of this article, forecasts will always introduce errors. Without considering these errors, the optimization results will be not effectively realized. In other words, the programmed minimum of the operating costs (or maximum of the port incomes) will be not obtained.
Managing uncertainties is therefore fundamental in an MEH port. As mentioned before, the fully and partially controllable components of the MEH port, such as storage systems, are the ones that make available to the MEHMS the flexibility needed to optimize the port operation. However, they can be used also to compensate prediction errors and effectively get the programmed cost minimization (or income maximization). To be able to achieve this compensation, they need to allocate a sufficient energy reserve. The sizing of this reserve must be included within the optimization problem based on a proper model of uncertainties. The scientific literature provides many methods to include stochastic data in the optimization. The two approaches mainly adopted are chance-constrained and scenario-based optimization.
In the first case, constraints involving stochastic data are reformulated in probabilistic terms to assure that their violation, which cannot be avoided with certainty, is avoided with a probability higher than a prescribed value, usually set close to 95%–99%. Such a reformulation is carried out based on the probabilistic model adopted to represent the stochastic data. Gaussian probability distribution models, for example, allow preserving the linearity of constraints. Differently, other models lead constraints to become nonlinear, making the optimization problem harder to be solved. In the latter case, stepwise linearization is a solution adopted in many approaches. It is clear that the model choice actually depends on the nature of the stochastic data. Even if it is the easiest to be used, the Gaussian model is not always the best choice to represent RES generation and electrical load. In any case, some approximation should be accepted, and a consequent tradeoff between accuracy and optimization solvability should be accepted.
In the scenario-based optimization, always based on a stochastic model, a set of possible realizations (scenarios) of the stochastic data are generated, and the constraints involving them are required to be satisfied for each of these realizations. This allows the adoption of any probability model but augments the number of constraints and the number of variables of the optimization problem. In general, the accuracy in representing the stochastic data is higher as there are more generated scenarios. In this case, the tradeoff to be established is between accuracy and computational burden.
Asking which solution between chance-constrained and scenario-based optimizations is the best has no answer, and in general, stochastic optimization applied to energy systems is today an open research item. In the following, we will provide the results of MEHMS adopting chance constraint optimization.
The correct evaluation of the CI load represents an essential task for both the design and management of the smart port. Typically, the estimation of the load power demand is addressed during the sizing stage of the shore connection infrastructure, while the term “forecast†refers to the real-time management of the resources during the port’s operating life.
Despite this conceptual distinction, both estimation and forecast rely on the ships’ characteristics and marine traffic scheduling. The power demand of a berthed ship ranges from hundreds of kW, for container ships, to tens of MW, in the case of cruise ships. However, the strict relation between the expected load demand and the marine traffic scheduling makes these loads potentially predictable, a feature which is essential for the definition of energy management strategies oriented to renewables integration. It is worth noting that the estimation of CI power is critical, not only within a smart port framework but also for defining reinforcing actions, required to face the expected increase of the port’s power demand. Indeed, feasibility studies and cost–benefit analyses are strongly dependent on an accurate prediction of power and energy demands of the port.
Two main approaches can be today considered as state-of-the-art practices for the estimation of the OPS load. One is based on the analysis of the monitoring, reporting, and verification (MRV) data, introduced by European Regulation 2015/757. This regulation defines the rules for the monitoring, reporting, and verification of CO2 emissions from ships arriving at, within, or departing from ports under the jurisdiction of any European member state. The European Maritime Safety Agency provides the information system to support the regulation, so that a database is available, and if the arrival and departure times of ships are known, then it is possible to obtain the actual power demand of the ships during the berthing time.
When the MRV data are not available, such as in ports outside the European jurisdiction, an alternative approach is given by the U.S. Environmental Protection Agency. The “Shore Power Technology Assessment at U.S. Ports†report, and the related tool, are focused on estimating the environmental benefits of the shore power when a vessel is docked and connected to the port’s grid. To achieve this evaluation, the power demand during berthing time is calculated as a percentage of the total engine power, through the inclusion of factors depending on the type of ship. A remarkable aspect to be considered when using this approach is that some ships can experience more than one operating mode. For example, cargo ships can experience the cargo handling and the idle operating modes, with the first being more demanding because of the handling equipment, such as pumps and cranes, that needs to be supplied in addition to the hotel load.
For both of the two approaches, the berthing time, i.e., the arrival and departure time of the ship, represents critical information. It is usually extracted from the ship positioning database and from the automatic identification system. Load forecast can be considered as a day-by-day refinement of the load estimation, obtained through the inclusion of the effective marine traffic scheduling.
Similar procedures can be followed to estimate the hydrogen demand of a hydrogen-powered ship equipped with FCs and HS. The same amount of energy previously discussed has to be provided through the hydrogen vector, instead of using CI. Therefore, electrical energy can be converted into hydrogen mass considering the lower heating value of hydrogen, equal to 33.33 kWh/kg. The mass flow, in kilograms per hour, can be easily derived by looking at the expected power profile. In addition to the power required to sustain the ship’s operation during berthing time, the onboard HS also has to be refilled. In conclusion, the total amount of hydrogen flowing into the ship, ZES or HAES, is defined by two contributions: the one required for port’s operations and the one dedicated to the HS bunkering. The last contribution depends on the size of the storage and on the docking time.
To offer an example of the potentialities of a port as an MEH, we consider a hypothetical scenario located in the area of the Genova harbor, in Italy. Figure 2 shows a schematic representation of this study case, whereas Table 1 reports the principal parameters.
Figure 2. Scheme of the study case MEH port.
Table 1. Study case port parameters.
The different components have been sized using the commercial software Hybrid Optimization of Multiple Energy Resources (HOMER). The design was achieved by minimizing the levelized COE. The yearly electrical demand of ships moored at the electrical quay is defined based on approach proposed in the previous section, which provides an estimation of the total electrical load of roll-on/roll-off passenger (Ro-Pax) ships moored at a quay in the Genova harbor in 2019. The same procedure is followed to establish the yearly demand of hydrogen from HAESs using the equivalence between hydrogen and electrical energy (1 MWh = 30 kg) and under the hypothesis that they are equipped with onboard PEMFCs. Hybrid ships are assumed to belong to the macrocategories of small chemical and oil tankers. The hydrogen mass flow to the HAES is kept constant, so that the difference between the supply and the consumption goes into the HS. It is also assumed that only one Ro-Pax ship and one HAES at a time can be moored at each quay.
Wind and PV power plants are sized based on the wind speed and solar irradiation data in the area of Genova, collected from the NASA Prediction of Worldwide Energy Resource database. To set the maximal rate of the PV power plant, the availability of an area of 123,800 m2 is assumed, estimated by the Authority of the Genova port system. Efficiencies and potential sizes of the BESS, FC, Ely, and HESS come from datasheets of commercial devices.
Table 2 reports the energy prices adopted in the simulations. They all refer to the first semester of 2021. The energy purchase price is set based on the data shared by the Authority of the Genova port system. The energy selling prices, which vary with hour and day of the week, are set based on an average of the Italian market prices along the considered period. The tariff applied to AESs for CI is set 3% higher than the energy purchase price, in order to make it sustainable for the port when RES generation is low. Finally, the price of hydrogen, whose value is hard to estimate for the future, is set equal to the one of shore connection.
Table 2. Simulation scenario: port energy prices.
We propose the results obtained in four weeks from 1–7 February, 1–7 April, 1–7 June, and 1–7 October 2019. Figures 3 and 4 report the forecasted and actual profiles of the RES generation and of the shore connection AESs load, respectively. Notice that all power profiles have a time granularity of 1 h, actually representing energy exchanges. RES profiles have been collected from the NASA Prediction Of Worldwide Energy Resources (NASA POWER) database. AES load forecasts are computed as constant values associated with the estimated energy required by the AESs scheduled to berth at the electrical quay. According to an analysis on real data, the actual profile of a Ro-Pax ship deviates from this constant value estimate according to a normal distribution with a 3% standard deviation. Therefore, the real profiles in Figure 4 were generated based on such a hypothesis.
Figure 3. Forecasted and actual profiles of RES generation. Week of 1–7 October.
Figure 4. Forecasted and actual profile of AESs load. Week of 1–7 October.
The port operation is managed by an MEHMS based on a chance-constrained MPC. MPC is a discrete-time optimal control technique that consists of solving an optimization problem along a time interval from the current time to a given future time (time horizon), thus obtaining an optimal control trajectory; applying only the first element of this trajectory to the system; and repeating the procedure at the next time step. In this case, the MEH port optimization problem is formulated along a time horizon of 12 h with a time granularity of 1 h. Forecasts of RES and AES demand are modeled by Gaussian models. The compensation of forecasting errors is entrusted to the BESS, the Ely, and the FC. One the objectives of the MEHMS is therefore to establish the contributions to the compensation that are required from these three devices. This is realized by defining three compensation participation weights that establish the percentage of contribution required from each device. Introducing such weights, the linear chance constraints obtained adopting the Gaussian model for RES and AES forecasts become nonlinear. Thus, piecewise linearization is employed, finally obtaining a linear mixed-integer optimization problem. All details can be found in Conte et al.
Figures 5–7 report the simulation results obtained in the October week. Figure 5 shows how the MEHMS dynamically establishes the contributions of the BESS, FC, and Ely to the compensation of the RES and AESs prediction errors. The effects of the decisions of the MEHMS can be observed in Figure 6 for the BESS and in Figure 7 for the FC and Ely, whose power profiles deviate from the forecasted one, precisely for compensating the RES and AES prediction errors. In these two figures, we can also note that the BESS state of charge (SoC) and the level of hydrogen (LoH) in the storage are correctly managed, being limited within the prescribed upper and lower bounds (10% to 90% for SoC and 0% to 100% for LoH), and that the technical minima of FC and Ely are only violated by switching them off.
Figure 5. Simulation results: compensation participation weights. Week of 1–7 October. af: FC participation weight; ael: Ely participation weight; ab: BESS participation weight.
Figure 6. Simulation results: BESS (a) power exchange and (b) SoC. Week of 1–7 October.
Figure 7. Simulation results: (a) HAES hydrogen load, power profiles of (b) FC and (c) Ely, and (d) level of hydrogen (LoH) profile. Week of 1–7 October.
Concerning the results obtained in all four simulated weeks, it is worth remarking that no RES curtailment is performed, and the use of onboard DGs is always avoided. Moreover, both shore connection and hydrogen loads are always fully satisfied. Figure 8 summarizes the results in terms of total energy exchanges. For each week, the left bar reports the energy imported from the main grid and the one generated by RESs, FCs, and BESSs, whereas the right bar depicts how this energy is used to satisfy the shore connection demand, to recharge the battery and the HSs, and to sell green energy to the market.
Figure 8. Simulation results: total energy exchanges.
Table 3 reports the results obtained in economic terms. Here we can observe that in all four weeks, the port realizes an economical gain, which is mainly due to shore connection. This occurs because the load estimated for the electrical quay is significantly higher than the one estimated for the hydrogen quay since Ro-Pax ships are definitely more energy-consuming than chemical and oil tankers.
Table 3. Simulation results: costs and incomes.
Smart ports can play a pivotal role in the world’s decarbonization challenge, being natural energy hubs for different types of energy vectors. The principal issues to be taken into account to optimally manage the operation of an MEH port have been analyzed, considering the properties of each potential component of the port and the different tasks to be carried out, such as CI and hydrogen bunkering. Particular attention has been devoted to the possible strategies for dealing with the uncertainties introduced by forecast errors. Finally, a specific optimal management algorithm based on a chance-constrained MPC has been applied to a study case MEH port performing CI and hydrogen bunkering. The control approach takes into account the uncertainty of RES generation exploiting the two storage systems installed in the port: a BESS and a HESS composed of an Ely, an FC, and an HS. The obtained results show that thanks to an optimal management system, an MEH port can optimally work, by simultaneously maximizing the port economical income and avoiding the use of onboard DGs and of any RES curtailment.
F. D’Agostino, G. P. Schiapparelli, S. Dallas, D. Spathis, V. Georgiou, and J. Prousalidis, “On estimating the port power demands for cold ironing applications,†in Proc. IEEE Electric Ship Technol. Symp. (ESTS), 2021, pp. 1–5, doi: 10.1109/ESTS49166.2021.9512359.
F. Conte, F. D’Agostino, D. Kaza, S. Massucco, G. Natrella, and F. Silvestro, “Optimal management of a smart port with shore-connection and hydrogen supplying by stochastic model predictive control,†in Proc. IEEE Power Energy Soc. General Meeting (PESGM), 2022, pp. 1–5, doi: 10.1109/PESGM48719.2022.9916817.
M. Banaei, M. Rafiei, J. Boudjadar, and M. H. Khooban, “A comparative analysis of optimal operation scenarios in hybrid emission-free ferry ships,†IEEE Trans. Transport. Electrific., vol. 6, no. 1, pp. 318–333, Mar. 2020, doi: 10.1109/TTE.2020.2970674.
J. Kumar, L. Kumpulainen, and K. Kauhaniemi, “Technical design aspects of harbour area grid for shore to ship power: State of the art and future solutions,†Int. J. Elect. Power Energy Syst., vol. 104, pp. 840–852, Jan. 2019, doi: 10.1016/j.ijepes.2018.07.051.
Francesco Conte (f.conte@unicampus.it) is with the Unit of Innovation, Entrepreneurship & Sustainability, Department of Engineering, Campus Bio-Medico University of Rome, 00128 Rome, Italy.
Fabio D’Agostino (fabio.dagostino@unige.it) is with the Department of Electrical, Electronic, Telecommunication Engineering and Naval Architecture, University of Genova, 16145 Genova, Italy.
Federico Silvestro (federico.silvestro@unige.it) is with the Department of Electrical, Electronic, Telecommunication Engineering and Naval Architecture, University of Genova, 16145 Genova, Italy.
Digital Object Identifier 10.1109/MELE.2022.3232981
2325-5897/23©2023IEEE