scispace - formally typeset
Search or ask a question
Journal ArticleDOI

A Stochastic Digital Implementation of a Neural Network Controller for Small Wind Turbine Systems

TL;DR: This letter presents a reconfigurable hardware implementation of feed-forward neural networks using stochastic techniques to approximate the nonlinear sigmoid activation functions with reduced digital logic resources on an FPGA device with high fault tolerance capability.
Abstract: This letter presents a reconfigurable hardware implementation of feed-forward neural networks using stochastic techniques. The design is based on the stochastic computation theory to approximate the nonlinear sigmoid activation functions with reduced digital logic resources. The large parallel neural network structure is then implemented on a reconfigurable field-programmable gate array (FPGA) device with high fault tolerance capability. The method is applied to a neural-network based wind-speed sensorless control of a small wind turbine system. The experimental results confirmed the validity of the developed stochastic FPGA implementation. The general design method can be extended to include other power electronics applications with different feed-forward neural network structures
Citations
More filters
Journal ArticleDOI
TL;DR: This article presents a comprehensive overview of the hardware realizations of artificial neural network models, known as hardware neural networks (HNN), appearing in academic studies as prototypes as well as in commercial use.

638 citations


Cites methods from "A Stochastic Digital Implementation..."

  • ...…developed as software, there are specific applications such as streaming video compression, which demand high volume adaptive real-time processing and learning of large datasets in reasonable time and necessitate the use of energy-efficient ANN hardware with truly parallel processing capabilities....

    [...]

Journal ArticleDOI
TL;DR: A stochastic NN structure is proposed in this paper for an FPGA implementation of a feedforward NN to estimate the feedback signals in an induction motor drive and significantly reduces the number of logic gates required for the proposed NN estimator.
Abstract: This paper applies stochastic theory to the design and implementation of field-oriented control of an induction motor drive using a single field-programmable gate array (FPGA) device and integrated neural network (NN) algorithms. Normally, NNs are characterized as heavily parallel calculation algorithms that employ enormous computational resources and are less useful for economical digital hardware implementations. A stochastic NN structure is proposed in this paper for an FPGA implementation of a feedforward NN to estimate the feedback signals in an induction motor drive. The stochastic arithmetic simplifies the computational elements of the NN and significantly reduces the number of logic gates required for the proposed NN estimator. A new stochastic proportional-integral speed controller is also developed with antiwindup functionality. Compared with conventional digital controls for motor drives, the proposed stochastic-based algorithm enhances the arithmetic operations of the FPGA, saves digital resources, and permits the NN algorithms and classical control algorithms to be easily interfaced and implemented on a single low-complexity, inexpensive FPGA. The algorithm has been realized using a single FPGA XC3S400 from Xilinx, Inc. A hardware-in-the-loop (HIL) test platform using a Real Time Digital Simulator is built in the laboratory. The HIL experimental results are provided to verify the proposed FPGA controller.

123 citations


Cites background from "A Stochastic Digital Implementation..."

  • ...However, the approximation error can be reduced with an increase of the FPGA’s clock rate [47], [ 48 ], [49]....

    [...]

Journal ArticleDOI
TL;DR: In this paper, a model predictive control strategy for the ac-dc-ac converter of wind system is derived and implemented to capture the maximum wind energy as well as provide desired reactive power.
Abstract: This paper presents the operation and controller design of a microgrid consisting of a direct drive wind generator and a battery storage system A model predictive control strategy for the ac-dc-ac converter of wind system is derived and implemented to capture the maximum wind energy as well as provide desired reactive power A novel supervisory controller is presented and employed to coordinate the operation of wind farm and battery system in the microgrid for grid-connected and islanded operations The proposed coordinated controller can mitigate both active and reactive power disturbances that are caused by the intermittency of wind speed and load change Moreover, the control strategy ensures the maximum power extraction capability of wind turbine while regulating the point of common coupling bus voltage within acceptable range in both grid-connected and islanded operations The designed concept is verified through various simulation studies in EMTDC/PSCAD, and the results are presented and discussed

81 citations

Journal ArticleDOI
TL;DR: This paper provides a rigorous mathematical treatment of stochastic implementation of complex functions such as exponentiation and tanh implemented using linear FSMs, and presents two new functions, an absolute value function and exponentiation based on anabsolute value, motivated by specific applications.
Abstract: Most digital systems operate on a positional representation of data, such as binary radix. An alternative is to operate on random bit streams where the signal value is encoded by the probability of obtaining a one versus a zero. This representation is much less compact than binary radix. However, complex operations can be performed with very simple logic. Furthermore, since the representation is uniform, with all bits weighted equally, it is highly tolerant of soft errors (i.e., bit flips). Both combinational and sequential constructs have been proposed for operating on stochastic bit streams. Prior work has shown that combinational logic can implement multiplication and scaled addition effectively while linear finite-state machines (FSMs) can implement complex functions such as exponentiation and tanh effectively. Prior work on stochastic computation has largely been validated empirically.This paper provides a rigorous mathematical treatment of stochastic implementation of complex functions such as exponentiation and tanh implemented using linear FSMs. It presents two new functions, an absolute value function and exponentiation based on an absolute value, motivated by specific applications. Experimental results show that the linear FSM-based constructs for these functions have smaller area-delay products than the corresponding deterministic constructs. They also are much more tolerant of soft errors.

67 citations

Proceedings ArticleDOI
09 Mar 2015
TL;DR: This paper proposes a novel radial basis function artificial neural network using stochastic computing elements, which greatly reduces the required hardware and can be expanded to larger scale networks for complex tasks with simple hardware architectures.
Abstract: Hardware implementations of artificial neural networks typically require significant amounts of hardware resources. This paper proposes a novel radial basis function artificial neural network using stochastic computing elements, which greatly reduces the required hardware. The Gaussian function used for the radial basis function is implemented with a two-dimensional finite state machine. The norm between the input data and the center point is optimized using simple logic gates. Results from two pattern recognition case studies, the standard Iris flower and the MICR font benchmarks, show that the difference of the average mean squared error between the proposed stochastic network and the corresponding traditional deterministic network is only 1.3% when the stochastic stream length is 10kbits. The accuracy of the recognition rate varies depending on the stream length, which gives the designer tremendous flexibility to tradeoff speed, power, and accuracy. From the FPGA implementation results, the hardware resource requirement of the proposed stochastic hidden neuron is only a few percent of the hardware requirement of the corresponding deterministic hidden neuron. The proposed stochastic network can be expanded to larger scale networks for complex tasks with simple hardware architectures.

57 citations


Cites background from "A Stochastic Digital Implementation..."

  • ...Brown and Card [7] proposed a number of stochastic computational units that significantly boosted performance, including a linear finite state machine for exponentiation operations....

    [...]

References
More filters
Book ChapterDOI
01 Jan 1969
TL;DR: The invention of the stored-program digital computer during the second world war made it possible to replace the lower-level mental processes of man by electronic data-processing in machines, but the authors lack the "steam engine" or "digital computer" which will provide the necessary technology for learning and pattern recognition by machines.
Abstract: The invention of the steam engine in the late eighteenth century made it possible to replace the muscle-power of men and animals by the motive power of machines. The invention of the stored-program digital computer during the second world war made it possible to replace the lower-level mental processes of man, such as arithmetic computation and information storage, by electronic data-processing in machines. We are now coming to the stage where it is reasonable to contemplate replacing some of the higher mental processes of man, such as the ability to recognize patterns and to learn, with similar capabilities in machines. However, we lack the “steam engine” or “digital computer” which will provide the necessary technology for learning and pattern recognition by machines.

668 citations


"A Stochastic Digital Implementation..." refers methods in this paper

  • ...Stochastic arithmetic was first introduced in the field of information and computing in the 1960s [13] as a new method aimed at solving complex computations with simple hardware....

    [...]

Journal ArticleDOI
TL;DR: The primary contribution of this paper is in introducing several state machine-based computational elements for performing sigmoid nonlinearity mappings, linear gain, and exponentiation functions, and describing an efficient method for the generation of, and conversion between, stochastic and deterministic binary signals.
Abstract: This paper examines a number of stochastic computational elements employed in artificial neural networks, several of which are introduced for the first time, together with an analysis of their operation. We briefly include multiplication, squaring, addition, subtraction, and division circuits in both unipolar and bipolar formats, the principles of which are well-known, at least for unipolar signals. We have introduced several modifications to improve the speed of the division operation. The primary contribution of this paper, however, is in introducing several state machine-based computational elements for performing sigmoid nonlinearity mappings, linear gain, and exponentiation functions. We also describe an efficient method for the generation of, and conversion between, stochastic and deterministic binary signals. The validity of the present approach is demonstrated in a companion paper through a sample application, the recognition of noisy optical characters using soft competitive learning. Network generalization capabilities of the stochastic network maintain a squared error within 10 percent of that of a floating-point implementation for a wide range of noise levels. While the accuracy of stochastic computation may not compare favorably with more conventional binary radix-based computation, the low circuit area, power, and speed characteristics may, in certain situations, make them attractive for VLSI implementation of artificial neural networks.

497 citations


"A Stochastic Digital Implementation..." refers background in this paper

  • ...The fundamental principles and processes of the stochastic arithmetic [14], [15] are summarized as follows....

    [...]

Journal ArticleDOI
Hui Li1, Michael Steurer1, K.L. Shi1, S. Woodruff1, Da Zhang1 
TL;DR: From the dynamic test results presented, it is concluded that the proposed system shows great potential for the development of a unified wind energy design, test, and research platform.
Abstract: Traditionally, offline modeling and simulation has been the tool of choice for improving wind energy system control strategies and their utility system integration. This paper exploits how a newly established real-time hardware-in-the-loop (HIL) test facility, which is designed for testing all-electric ship propulsion systems, can be utilized for wind energy research. The test site uses two 2.5-MW/220-rpm dynamometers and a 5-MW variable voltage and frequency converter to emulate a realistic dynamic environment, both mechanically and electrically. The facility is controlled by a digital real-time electric power system simulator that is capable of simulating electrical networks and control systems of substantial complexity, typically with a 50-mus time step. Substantial input/output allows the feedback of measured quantities into the simulation. A 15-kW mock-up motor-generator set is used to demonstrate some critical aspects of the concept including the implementation of a proposed neural-network-based sensorless maximum wind energy capture control. From the dynamic test results presented, it is concluded that the proposed system shows great potential for the development of a unified wind energy design, test, and research platform

208 citations

Proceedings ArticleDOI
02 Oct 1994
TL;DR: The application of neural networks for estimation of feedback signals in induction motor drive systems is explored and the neural network estimator has the advantages of faster execution speed, harmonic ripple immunity and fault tolerance characteristics compared to the DSP-based estimator.
Abstract: Neural networks are showing promise for application in power electronics and motion control systems. So far, they have been applied for a few cases, mainly in the control of converters and drives, but their application in estimation is practically new. The purpose of this paper is to demonstrate that such a technology can be applied for estimation of feedback signals in an induction motor drive with some distinct advantages when compared to DSP based implementation. A feedforward neural network receives the machine terminal signals at the input and calculates flux, torque, and unit vectors (cos /spl theta//sub e/ and sin /spl theta//sub e/) at the output which are then used in the control of a direct vector-controlled drive system. The three-layer network has been trained extensively by Neural Works Professional II/Plus program to emulate the DSP-based computational characteristics. The performance of the estimator is good and is comparable to that of DSP-based estimation. The system has been operated in the wide torque and speed regions independently with a DSP-based estimator and a neural network-based estimator, and are shown to have comparable performance. The neural network estimator has the advantages of faster execution speed, harmonic ripple immunity, and fault tolerance characteristics compared to DSP-based estimator. >

190 citations

Proceedings ArticleDOI
03 Oct 1999
TL;DR: In this paper, a neural network based implementation of space vector modulation of a voltage-fed inverter has been proposed that fully covers the undermodulation and overmodulation regions linearly extending operation smoothly up to square wave.
Abstract: A neural network based implementation of space vector modulation of a voltage-fed inverter has been proposed in this paper that fully covers the undermodulation and overmodulation regions linearly extending operation smoothly up to square wave. The neural network has the advantage of very fast implementation of SVM algorithm that can increase the converter switching frequency, particularly when a dedicated ASIC chip is used in the modulator. Two ANN-based SVM techniques have been validated: an indirect method with the help of a timer that generates the PWM waveforms from the command voltage vector at the input, and a direct method that synthesizes waveforms directly without any timer. The indirect method has been fully implemented and extensively evaluated in a volts/Hz controlled 5 hp, 60 Hz, 230 V induction motor drive. The performances of the drive with ANN-based SVM are excellent. The scheme can be easily extended to vector-controlled drive. The direct method, although has a simpler topology, needs very large training data and training time.

125 citations