Research and Development of a Large-Scale Axial-Flux Generator for Hydrokinetic Power System

Featured Application

A developed large-scale axial-flux generator is used for electricity generation at a run-of-river hydrokinetic power system. The system is to be used in a river with a high volumetric flow rate in Asia.

Abstract

The study demonstrates an application of actual technologies and tools for the development of an axial-flux electricity generator. The specifics of its application—a run-of-river sited power station—predefine some of the design parameters that are close to a wind turbine generator. An extensive study of available solutions is used as a starting point for further concept development. The study aims to provide a viable solution for a large-scale electrical machine. A step-based methodology is defined for concept parameters’ assessment and a feasibility study. It demonstrates the advantages of virtual prototyping when assessing various design parameters such as air gaps, coil thickness, and the number of rotor disks. Several simulations over different virtual prototypes provide sufficient information to elaborate an improved design concept. The major result is a ready-for-detailed design concept, with defined major parameters and studied work behavior for a specific, large structure of an electrical machine. Another important result is the presentation of the application of virtual prototyping in the assessment of large structures, for which physical prototyping is an expensive and time-consuming approach. The application of virtual prototyping at a very early product development stage is an effective way to undertake efficient solutions involving the concept of the product.

1. Introduction

The demand for energy and electricity has been rising at its fastest rate in recent years, driven by a growing population, robust economic growth, intense heatwaves, and the increasing uptake of technologies that run on electricity. The strong increase in global electricity consumption is set to continue into the next years, with growth of around 4%, according to a report by the International Energy Agency (IEA) [1]. Two major challenges, related to the energy demand increases, could be solved: equal energy access to all countries and people, and the need to reduce greenhouse gas emissions and environmental pollution [2]. It is known that renewable energy will play a key role in decarbonizing energy systems in the coming decades. Currently, the most popular energy sources are solar, wind, hydro, geothermal, and biomass energy. As a renewable energy resource, hydropower is one of the most commercially developed. Hydropower plants are currently the most contributory renewable energy source worldwide, and they continue to be installed, especially in emerging countries [3,4]. Hydropower generation could be divided into the next major types: storage (and its variant—pumped storage), run-of-river (including in-stream), and tidal. A run-of-river hydropower plant draws energy for electricity production mainly from the available flow of a river (sometimes called “zero-head” hydropower), sometimes with short-time storage [5]. In-stream solutions are a good alternative, having a lower environmental impact, in high-flow rivers since they use rivers’ high kinetic energy. Generally, hydrokinetic systems convert the energy of moving water in rivers, oceans, or tidal currents into electricity without the use of a dam or barrage associated with conventional hydropower [6]. There are several kinds of hydrokinetic turbines of various sizes and different energy-capture principles. The major classification of hydrokinetic turbines is related to rotating axis positions with respect to the water flow: with a horizontal axis and a vertical axis. One of the constraints for the turbine power application is the low typical river current velocity [7]. The design philosophies of hydrokinetic turbines are similar to those of wind turbines, sharing the same working principles, which consist of converting the hydrokinetic power into mechanical power in the form of rotating blades [8]. Moreover, horizontal axis installations, with a buoyant mechanism, may allow a non-submerged generator to be placed closer to the water surface [9]. These specifics outline the major requirements concerning electricity generator design.

An electricity generator for similar applications (including wind turbines) has also been subjected to active research [10,11,12,13]. Traditionally, three main types of wind turbine generators can be considered for the various wind turbine systems, these being direct current (DC), alternating current (AC) synchronous, and AC asynchronous generators [14]. This is very similar when discussing generators for hydrokinetic power systems. For instance, a 2 MW turbine with a permanent magnet ring generator that measures 16 m in diameter was installed in the Bay of Fundy, Canada, in 2016 [15].

This study presents key points in the design development of an electricity generator for a hydrokinetic power system. In addition its specifics, similar to wind turbine generators, it also involves large dimensions and a high power output. The electrical generator is to be mounted on an in-stream installation of a buoyant type, as is shown in general in Figure 1.

The floating nature of the system allows it to be easily anchored and unanchored in different water bodies, such as canals and rivers. This mobility enables the system to be deployed where needed, offering greater accessibility to remote areas or locations with varying water-flow conditions [16,17]. This floating hydropower plant is intended to use a water wheel, with a relatively low rotational speed, that requires a gearbox [18]. Water wheels are actually the focus of various research studies [19,20,21,22,23], including large-scale designs such as the presented one [24]. However, the specifics of large-scale electricity generators for such applications are not widely examined. The application of this solution to a river stream requires certain parameters of the river stream, such as a water-flow velocity minimum of 2 m/s. Such conditions exist in rivers with average to high discharge volumes, some of them in areas with dominant tropical monsoon climates. The water-flow velocity has a nonlinear effect on the return on investment in this type of system, and artificial parallel channels could also be searched for to improve this parameter. Another option is to use tidal conditions to increase efficiency. Additional studies, mostly related to large-scale water-wheel concept development, would involve detailed feasibility analyses.

The concept of the developed electricity generator is of the axial-flux permanent magnet (AFPM) type. Axial-flux permanent magnet generators are always less widely known than radial flux generators because of their drawbacks, which include manufacturing issues and difficulties maintaining air-gap uniformity [25]. However, an axial flux-type machine applies to the low-speed, high-power operation of a direct drive energy system [26]—such as wind or run-of-river systems [27]. The research and development of the current AFPM generator is based on the virtual prototyping of electromagnetic processes, as it is already a common practice in product development [28,29,30,31]. Virtual prototyping offers the significant advantage of being able to explore certain products without the necessity of building their physical replication. This is very important in the case of heavy industrial product development as large-scale electricity generators. This type of prototype (or digital mock-up) allows for performing variations of different parameters in a search for better performance and even optimization [32,33,34,35]. The evaluation of these conceptual design variants is achieved using electromagnetic simulations via the application of numerical analysis techniques.

2. Materials and Methods

Design parameters’ assessment and improvement is a step-based process in which an initial conceptual design is evaluated. The next step is to perform a study of various parameters to reach an improved version of the initial design. Finally, a feasibility study is to be performed in order to include additional key points such as manufacturability. This section describes the actions performed during this improvement process, and the results are presented and commented on in the next chapter.

2.1. Virtual Prototype of Preliminary Design

The examined AFPM generator is based on an initially predefined geometry, with the concept of a stacked rotor that forms the magnetic circuit in the coreless generator and overall dimensions, as is shown in Figure 2. This initial conceptual design is rated for 20 MW at 13 rpm, having a 13.5 m overall diameter and 4 rotor stages. It has 48 permanent NdFeB magnets installed on each rotor at an air gap of 140 mm to the stator coils. Rotor magnets are embedded in a disk hole made of non-magnetic material, while the stator coils are embedded in the disk and made of insulating material to avoid generating an eddy current. The magnet polarity direction is aligned with that of the neighboring rotor, and it generates the magnetic field between the gaps of the rotors. The coreless coil is arranged in a ring shape in the stator, and the stator itself is positioned in the gap of the rotor to generate electromotive force by receiving the alternating magnetic field.

This prototype is used to perform three-dimensional magnetic field analyses. The applied boundary conditions are the current and rotational speed of the system. The electromotive force was derived at different generated powers from the magnetic flux quantity that interlinks the coil. Additionally, the losses were estimated, and the major parameter of the design could be calculated—the efficiency. The efficiency (η_G) of a generator is the ratio of the electrical power produced via the generator (P_G) to the mechanical supplied to the generator (P_M).

$${\eta }_G={\displaystyle \frac{P_G}{P_M}}$$

(1)

The generated power (P_G) could be calculated directly from the simulation results, but it could also be presented as the difference between the input power (P_M) and the power loss (P_L).

$$P_G=P_M-P_L$$

(2)

The input mechanical power (P_M) could be presented by the torque (T) and rotational speed (n) of the rotor as follows:

$$P_M={\displaystyle \frac{\pi \ast n}{30}}\ast T$$

(3)

Thus, the efficiency (η_G) could be shown as a function of the torque (T), rotational speed (n), and power loss as follows (P_L):

$${\eta }_G=1-{\displaystyle \frac{30}{\pi \ast n\ast T}}\ast P_L$$

(4)

Generally, the simulation is set by adjusting the input parameter for the current (I) in the coils at a constant speed. Obtained values for the torque (T) and power loss (P_L) are further used to estimate the generator efficiency (η_G).

2.2. Study of Design Parameters

Numerical models and simulations, or virtual prototyping, are relatively cheaper and faster when compared to physical ones, especially for large-scale structures. Despite design simplicity, there are different parameters that could be examined using virtual prototyping techniques that could form a large domain of values for varied parameters. Based on experience, a group of several design parameters are marked to be studied:

• Air gap—this is probably the most important geometry parameter, and its minimum is searched for to improve the design performance. It is limited by technology requirements specific to such a large-scale product.
• Coil thickness—this parameter is related to the coil current density, and it influences the electromagnetic performance of the examined generator.
• Number of rotor disks—this parameter of the concept is related to the cost-effectiveness of the design, and it is limited by the required capacity of the generator.
• Planned design variants and simulations are shown as a scheme in Figure 3 to present the variety of models that are to be prepared. This plan presumes three stages:
• Stage I—a general review of parameter sensitivity. It includes three different virtual prototypes with gradually added changes. The first model is used to assess air-gap decreases’ influence over system performance, while the other two models introduce the rest of the design parameters.
• Stage II—it is a combination of air-gap and coil-thickness variants to achieve an improved design solution.
• Stage III is dedicated to the evaluation of the final design and a comparison with the preliminary concept.

2.3. Feasibility Study

This final step is needed to summarize information concerning the developed concept for large-scale electricity generators. It focuses on technical aspects such as key design parameters, efficiency, manufacturability, and the overall mechanical solution of the product.

3. Results

3.1. Preliminary Design Simulation Results

Dynamic simulations were performed using the virtual prototype, which is shown in Figure 2b. The simulation model presents complete generator functionality, including the rotor’s relative motion to the stator, using a model of the entire structure. Three-dimensional low-frequency electromagnetic transient simulation software (ANSYS Maxwell v.19) was used; it solves electromagnetic field problems by solving Maxwell’s equations in a finite region of space with time-varying or moving electrical sources. It uses the differential equations for Faraday (5), Gauss (6), and Ampere (7) laws, involving the electric field strength, E, the magnetic flux density, B, the magnetic field strength, H, and the electric current density, J, together with the constitutive material equations (B = f (H) for the magnetic field).

$$\nabla \times E=-{\displaystyle \frac{\partial B}{\partial t}}$$

(5)

$$\nabla \times B=0$$

(6)

$$\nabla \times H=\mathrm{J}$$

(7)

The numerical solution of such equations is based on T-Ω formulation (a single connected solution domain) in which Ω is the nodal-based magnetic scalar potential, defined in the entire solution domain, and T is the edge-based electrical vector potential, defined only in the conducting eddy-current region.

Three simulations are run for the model at various input mechanical power values—10 MW, 14 MW, and 18 MW. These simulations were performed using a workstation with actual configuration (corresponding to 2× Intel Xeon Gold 5318Y @ 2.1GHz 24 cores CPU and 1TB DDR4 RAM) that resulted in an adequate computational time. The simulation results show the losses and allow for calculating the generator efficiency. The software (ANSYS Maxwell v.19) calculates the core loss (eddy current, calculated using the Steinmetz formula, and hysteresis losses in a transient solution) and conductor loss (solid or stranded conductor). The conductor loss is dominant because of the type of electrical machine examined—coreless, and the conductor loss is determined from the current density and the material conductivity (volume-integral value).

The results obtained from all three simulations are briefly presented in Figure 4. Torque variations over time are shown, demonstrating some fluctuations, partially due to the specifics of the design and partially due to simulation accuracy (a nonlinear model of a rotating electromagnetic field). Magnetic field intensity is also presented—in Figure 4b—to visualize its distribution in the structure during the rotation of the moving parts. The major output of the simulation is the averaged value of the torque (used to adjust simulation parameters to the required input mechanical power values) and the losses, which are indicators of the efficiency, as well as further requirements for the thermal management of the entire system.

A summary of three performed simulations is shown in Figure 5. The graphs present the changes in power loss and, respectively, the efficiency by changing the input mechanical power. This is an essential indicator for generator design, and it was used in further considerations.

Several conclusions could be identified, based on the reported results:

• The examined preliminary design reaches a 2.4 A/m² current density per winding at a 10 MW input mechanical power. This case also shows the best efficiency—75%.
• The 14 MW input mechanical power reflects in nearly a 3.5 A/m² current density per winding. The efficiency for this case is much lower—66%.
• The design shows a high rate of loss in windings, and it is not efficient.
• Several reasons for this low efficiency were found after detailed results’ analysis.
• A large air gap between the magnet and the winding—140 mm. A similar design [27] has an air gap of 40 mm.
• A large distance between two stages—480 mm. Again, a similar design shows a gap between magnets of 140 mm.
• Further design improvements had to be performed in the next step.

3.2. Results for the General Review of Parameters’ Sensitivity

Next, the results from the first stage are presented, which presents the opportunity to check the importance of predefined parameters—the air gap, the coil thickness, and the number of rotor disks. Firstly, it is evident that the air gap needed revision, and it was decreased for the first simulation from 140 mm to 50 mm (simulation model I.1). Next, the coil thickness was decreased from 200 mm to 120 mm while maintaining the air gap of 50 mm (simulation model I.2), and the last simulation model I.3 has an additionally decreased number of rotor disks, from 5 to 3. The obtained results are summarized in Figure 6.

The comments for this stage are as follows:

• A significant improvement in efficiency is achieved directly in design variant I.1, where just the distance between magnets and coils is decreased from 140 mm to 50 mm.
• A further improvement is shown for the rest of the examined variants that led to the best variant, I.3, which also has approximately half as many components and is expected to be cheaper. Its efficiency reaches 90%.
• Further simulations are to be performed using the best chosen variant (as the performance of the twice-decreased power loss)—I.3.

3.3. Results for the Design Variants’ Exploration

This stage is related to a more detailed review of two parameters’ influence—the air gap and the coil thickness—as it was planned in the scheme shown in Figure 3. Half of the models (II.1, II.2, and II.3) have an air gap of 50 mm, and the aim was to explore coil thickness’s influence mainly by varying it to 200 mm, 250 mm, and 330 mm. Another three models (II.4, II.5, and II.6) explore the same thickness variations but at an increased air gap of 80 mm. The design model uses three rotor disks, and all simulations are run at an 18 MW mechanical power input.

The results from these six separate simulations are presented similarly—by comparing power losses and efficiencies calculated for each variant. This comparison is shown in Figure 7.

The results from this stage are commented upon as follows:

• Increasing the coil thickness leads to a definitive negative effect. A minimal thickness of the coil should be searched for in the next stage to improve the design performance.
• An increased air gap of 80 mm has less if an effect, but it is also negative. A further decrease in the air gap is limited by design and manufacturing constraints related to the maintained reliability during operation.
• Further simulations are to be performed using a design with a minimal coil thickness and a minimal air gap.

3.4. Results for the Final Conceptual Design Assessment

The last stage of the design parameters’ exploration used a developed virtual prototype and numerical techniques to determine values from the electromagnetics dynamic modeling. The results obtained from previous stages were combined with conclusions derived from the results study, and a new design was developed with the following main parameters: number of poles, 48; number of rotor disks, 3; magnet length [mm], 1800; magnet width [mm], 400; magnet thickness [mm], 50; coil width [mm], 350; coil thickness [mm], 60; and air gap [mm], 40. Other important specifics are as follows: four rows of magnets are used—i.e., the middle wall has two rows of magnets, one on each side; the coils are larger than magnets so as to avoid a magnet overlapping with tangential sides of the coil.

The performed simulations fully correspond to the previous simulations (rotational speed of 13 rpm and simulation model setup), and they were run for three load cases of the following mechanical input power: 10 MW, 14 MW, and 18 MW. The results are summarized and shown in Figure 8, together with preliminary variants’ results, for a better review and comparison.

Several observations could be outlined, based on the results obtained via the simulations of the final design variant:

• The preliminary defined design has insufficient efficiency, reaching ≈75% at a 10 MW mechanical power input;
• The best design variant, III, reaches ≈96% efficiency at a 10 MW mechanical power input;
• The final design variant also shows consistent efficiency, even at a mechanical power input increased up to 18 MW, reaching ≈92% efficiency;
• A further feasibility study needs to examine the manufacturability and other technical specifics of the developed conceptual design variant.

3.5. Feasibility Study

3.5.1. Axial Generator Efficiency

Axial-flux permanent magnet (AFPM) generators have many unique features. As permanent magnets, they are usually more efficient, as field excitation losses are eliminated, reducing rotor losses significantly. Their generator efficiency is thus greatly improved, and a higher power density is achieved. Axial-flux construction has less core material, so it has a high torque-to-weight ratio. Also, AFPM machines have thin magnets, so they are smaller than their radial-flux counterparts. The noise and vibration they produce are less than those of conventional machines.

Their air gaps are planar and easily adjustable. Also, the direction of the main air gap can be varied, so the derivation of various discrete topologies is possible. These benefits give AFPM generators advantages over conventional machines in various applications.

Efficiency is the major point of generator choice and design. It reflects not only in thermal behavior (power losses are heat-dissipated) but also in the overall price performance of a product. Generally, all available studies show ≈97% efficiency for this type of generator.

Efficiency also depends on the applied load (mechanical power input), as, with the increase in the current density in the windings, the power output increases, though slower than the increase in the losses. This shows the need for an examination of designs at various loads’ mechanical power input.

Initially, the preliminary defined design was examined in detail, and after several iterations, an improved design is proposed. The major reasons for obtaining improved efficiency are listed below in

Table 1. Improvements and their influence to reach better efficiency.

Parameter	Preliminary Defined Design	Improved Final Design
Ratio of number of magnets to number of windings rows	Five rows of magnets against four rows of windings, which leads to sharing a magnet between two coils	Four rows of magnets against two rows of windings, which leads to a pair of magnets for each winding
Radial size of the winding against the radial size of the magnet	Almost equal radial sizes of the magnet and the winding, which causes the magnetic field to influence tangential winding branches and decrease efficiency	The windings are larger than the magnets, and their tangential branches are not influenced by the magnetic field
Air gap size	140 mm air gap size decreases design efficiency	The air gap decreased to 40 mm
Coil thickness	200 mm thickness of the coil in the axial direction has a high negative effect on the efficiency	The coil thickness is decreased down to 60 mm, leading to a decreased overall axial dimension (flat design)
Coil cross section	Its big cross section (40,000 mm2) is based on a high winding thickness, which has a high negative effect on the efficiency	A sufficient cross section (21,000 mm2) is the result of decreased thickness but an increased width of 350 mm
Magnet width	400 mm magnet width against 200 mm winding width causes a magnet to overlap neighbor winding radial branch	A 400 mm magnet width fits the 350 mm winding width of branches

.

3.5.2. Manufacturability and Mechanical Solution

The generator has an outer diameter of nearly 14 m, and its manufacturing needs to use technologies typical of large structures. Several points could be raised at this phase of the product development process:

• Magnet production and mounting: NdFeB magnets are manufactured by arranging the crystal orientation of NdFeB magnet powder in a particular direction using an external magnetic field, applying mechanical pressure to the mold, and sintering it through powder metallurgy. In line with this process, the electromagnet used to apply the magnetic field is incorporated into the press. Due to the limitations of the electromagnet’s performance, the maximum size of a magnet manufactured in one press is approx. 100 mm². Existing studies on similar structures [27] use production technology for the magnet by assembling magnet blocks measuring 100 mm × 100 mm × 50 mm. These blocks are bonded to an iron plate, which is then bolted to the rotor disk, as NdFeB magnets are fragile, and threaded holes cannot be made in them (refer to Figure 9). Magnetic blocks are bedded and bonded to a back plate before magnetizing. Then, the unit is completed by one magnetizing process using the superconductive magnetizing equipment. Another problem is related to the mounting of magnets in the structure. It is expected to be difficult to control the unit due to the strong magnetic attractive force toward the rotor, especially when the unit is near the rotor.

• Windings’ assembly: The improved generator design has two rows of coils. Each row consists of 2× 36 coils placed over a 10 mm-thick aluminum plate—each side of the plate has 36 coils. Other two-sided aluminum plates are mounted to improve the cooling of the module. This is shown in general in Figure 10. Another design specific of the winding is that it is flat—a 30 mm thickness allows for using a single coil row in its thickness, which facilitates winding manufacturing.

4. Discussion

Existing experience, presented in current research studies, and numerical simulations (electromagnetic analyses using a virtual prototype) results and data were collected and studied in this work. The initially proposed design was improved, reaching ≈96% efficiency. The improved design also decreased the number of magnets and coils and decreased production costs, too. Additionally, a technical risk assessment focuses further design development on the next major directions:

• Cooling system (high-fluid flow solution will probably be needed);
• Production of magnets—special attention needs to be paid to the magnets’ assembly and mounting;
• Coils’ production and mounting on aluminum plates (disks).

This study was based on the virtual prototyping of the electromagnetic behavior of a large-scale AFPM generator, and it shows the good feasibility of the examined conceptual design. Further research is needed to go into detail about specific parameters and deepen the understanding of the performance of the generator, as well as improve its virtual prototype. Building and testing a physically scaled-down prototype is an important step in further design development.

5. Conclusions

The main purpose of this paper has been to present the application of virtual prototyping in the assessment of large structures, for which physical prototyping is an expensive and time-consuming approach. It has also presented the strengths of virtual prototyping in the evaluation of design parameters and their influence on searched-for performance. Finally, but not least, it has presented a good sample of the applied scientific approach to a real engineering problematics area that is also very real—green energy harvesting. Next, conclusions are outlined that specifically resulted from the presented study:

• A large-scale, axial-flux, permanent magnet generator concept was developed with a specific application for a hydrokinetic run-of-river system—a min 10 MW power plant.
• The examined final concept also showed consistent efficiency, even at a mechanical power input increased up to 18 MW.
• The final design concept variant showed ≈96% efficiency at 10 MW in mechanical power input and ≈92%—at 18 MW. These values are indicative, and they are a basis for the further development of a detailed design.
• A feasibility study was performed that showed a good perspective on obtaining a viable design solution, marking some critical points. It serves as a good reference point for further design development that will include the specifics of the system and its application.
• A structured approach to assessing electrical machines was demonstrated, using 15 simulations over 9 virtual models (mock-ups), in which the physical model was a challenging and overly expensive alternative. Important design parameters for axial electrical machines were assessed—air gaps, coil thickness, and the number of rotor disks—through dynamic simulations of the developed virtual prototypes.
• Further, the ability of numerical simulations to evaluate relatively easy physical parameters like an electromagnetic field, power loss, torque, etc. is an advantage that enables an engineering analysis of the concept and facilitates its further development. This approach was also demonstrated in this study, presenting its applicability in similar research studies.