In an order-driven financial market, the price of a financial asset is discovered through the interaction of orders - requests to buy or sell at a particular price - that are posted to the public limit order book (LOB). Therefore, LOB data is extremely valuable for modelling market dynamics. However, LOB data is not freely accessible, which poses a challenge to market participants and researchers wishing to exploit this information. Fortunately, trades and quotes (TAQ) data - orders arriving at the top of the LOB, and trades executing in the market - are more readily available. In this paper, we present the LOB recreation model, a first attempt from a deep learning perspective to recreate the top five price levels of the LOB for small-tick stocks using only TAQ data. Volumes of orders sitting deep in the LOB are predicted by combining outputs from: (1) a history compiler that uses a Gated Recurrent Unit (GRU) module to selectively compile prediction relevant quote history; (2) a market events simulator, which uses an Ordinary Differential Equation Recurrent Neural Network (ODE-RNN) to simulate the accumulation of net order arrivals; and (3) a weighting scheme to adaptively combine the predictions generated by (1) and (2). By the paradigm of transfer learning, the core encoder trained on one stock can be fine-tuned to enable application to other financial assets of the same class with much lower demand on additional data. Comprehensive experiments conducted on two real world intraday LOB datasets demonstrate that the proposed model can efficiently recreate the LOB with high accuracy using only TAQ data as input.
The explicit model predictive control (MPC) can solve the piecewise control laws offline to save online implementation burden. However, many offline control laws have to be stored to adapt the operating point variation, the correct control law needs to be searched, and the control parameter needs to be calculated. The large storage and computational burdens make the explicit MPC difficult to be applied to the scenarios with high switching and control frequencies. To solve these problems, this article proposes to utilize a backpropagation neural network (BPNN) to fit the input-output relationship of the offline control laws under different operating points. It not only guarantees the control performance but also reduces the storage and computational burden. Such a BPNN method directly calculates the control parameter in a parallel way and thus eliminates serial evaluation of the searching process. Simulation results are provided and compared with the state-of-the-art controls to show the effectiveness of the proposed method. Experimental results demonstrate that a BPNN with 49 parameters can fit more than 10 000 offline control laws, and its implementation can be completed within three clock cycles by field-programmable gate array (FPGA) or application-specific integrated circuit (ASIC), so the 1-MHz switching and control frequency can be achieved with 4-MHz clock frequency.
Hardware design and test of a cryogenic boost chopper under 77K are presented. Instead of using one transistor and one diode in conventional chopper, the new cryogenic boost chopper is integrated with two metal-oxide semiconductor field-effect transistors (MOSFETs). An experimental setup is built with one main circuit board, one drive circuit board and one capacitor board. The test results show that the on-state resistance of one cryogenic MOSFET unit is reduced from about $0.65 \mathrm {m}\Omega $ at 300 K to about $0.33 \mathrm {m}\Omega $ at 77 K. The tested operating efficiency is rated at about 96.6% under 1 kHz PWM control and about 95.3% under 5 kHz PWM control. Therefore, the proposed cryogenic chopper can be well expected to high-efficiency power conversion.
Compared to the conventional gate driver for the insulated gate bipolar transistor (IGBT), the active gate driver (AGD) not only reduces the switching delay but also suppresses voltage or current overshoot, however, at the expense of increasing the switching losses unavoidably. In this article, a self-adaptive active gate driver (SAGD) is proposed to further optimize IGBT switching performance, which is particularly suitable for applications where the load current varies over time or the dc bus voltage is changed. Since in such applications, AGD with fixed overshoot suppression ability will cause unnecessary switching loss, especially when the load current or the dc bus voltage is low. In contrast, with neither A/D nor D/A converters applied, the proposed SAGD indirectly detects the load current and dc bus voltage as time quantities, based on status monitoring, then self-adaptively chooses the optimal gate resistance increment online during the specific voltage or current rising phase. In this way, the current or voltage with overshoot is suppressed to be inside the safe operation area of IGBT while the switching loss is minimized, thus a better tradeoff between overshoots and switching losses is achieved. Finally, the feasibility of the proposed SAGD is verified by experiments.
A new diagnostic method for the dc-dc converters is proposed in this paper. The shape of voltage across the magnetic component is used as the diagnostic criterion. The Fast Fourier Transform (FFT) is utilized to extract the features of the waveform and the neural network (NN) is applied to realize the state classification. The voltage sensor is needless because the required voltage signatures can be obtained easily by adding a winding in the magnetic component, or fixing a magnetic near field probe near the magnetic component. The diagnostic system is isolated from the power stage naturally; all the A/D conversion, FFT and NN are realized in a single DSP chip TMS320F2812; using the inner A/D channels of the DSP, up to sixteen dc-dc converters can be monitored synchronously. For the illustrative purpose, the diagnostic process of one A/D channel will be described in this paper and the phase shift full bridge (PSFB) converter is chosen as the diagnostic object. Based on the discussion, the proposed method can be easily extended to the other types of dc-dc converters.
We demonstrate a robust, simple, and compact all polarization-maintaining (PM) fiber laser source with a repetition rate of 79 MHz for broadband coherent anti-Stokes Raman scattering (CARS) spectroscopy based on impulsive excitation and narrowband probing. The careful dispersion management during the generation of pump pulses ensures efficient impulsive excitation, which is verified to cover an ultrabroad bandwidth exceeding 4000 cm–1. The employment of PM fibers enables the laser source to withstand external disturbances. This turn-key configuration can potentially simplify the implementation of many applications of CARS, such as spectroscopic histopathology, weaponized endospore detection, and precise thermometry of gases.
The performance of switching devices has a great impact on the operating frequency and loss of power electronic circuits. Therefore, the working characteristics of the device are needed to be tested by double pulse test. However, the traditional double pulse test circuit has some disadvantages, such as high requirements for power drive capability, high cost and poor safety. Therefore, referring to the working principle of the Marx impulse voltage generator, this paper designs a double pulse test circuit platform with low charging requirements, low cost and good security. This paper verifies the reliability of the test platform by testing a GaN switching devices.
This paper designs a type of methane gas inspection tour and control system based on single chip microprocessor (SCM) technology and wireless communication technology, etc. This system fulfills methane gas concentration collection by LXY-3 methane gas sensor. SCM MSP430F149 analyzes and processes data as central core controller. Wireless transceiver nRF401 realizes data communication between field monitor and inspection tour equipment. And main controller is connected with inspection tour equipment by MAX232 interface. This paper describes system structure, parts of hardware and software design in detail. Experiments show that system monitor error is less than plusmn0.3% to methane gas concentration between 0 and 5%, which performs well to coal mine methane gas concentration monitor and alarm design demand as expected.
The applications of the modular multilevel converter (MMC) have been expanded to medium-voltage fields in recent years, and the diagnosis of submodule faults is a crucial issue for the system reliability. The fault diagnosis of a MMC is usually realized by software methods for cost-effective purpose. However, most of the software methods are based on complicated models and require two operational steps (i.e., fault detection step and fault localization step), and thus is time-consuming. To address this issue, this paper proposes a fast diagnosis method where each submodule is associated with a first-order sliding-mode observer which can detect and locate the submodule fault simultaneously. The diagnosis process of the submodules is both simple and independent to each other, thus can be implemented in parallel with Field Programmable Gate Array (FPGA). Both the open-circuit and short-circuit faults can be diagnosed, and no extra sampling cost is required. The proposed diagnosis method, including the first-order observer and the fast logic-based criterion, is verified by the experiment, and the results show good robustness under varied operation conditions.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)