使用物理信息的神经网络的离散功率电子设备的剩余有用寿命估计.
Zhonghai Lu1, Chao Guo2, Mingrui Liu2
1Department of Electrical Engineering, KTH Royal Institute of Technology, 16440, Stockholm, Sweden. zhonghai@kth.se.
Scientific reports
|June 22, 2023
概括
基于物理学的神经网络 (PINNs) 改善了电力电子产品的剩余使用寿命 (RUL) 估计. 通过结合物理定律,PINNs提高了循环神经网络 (RNN) 的准确性,从而实现更可靠的预测维护.
科学领域:
- 电气工程 电气工程
- 机器学习 机器学习
- 可靠性工程可靠性工程
背景情况:
- 对离散功率电子产品进行准确的剩余使用寿命 (RUL) 估计对于预测性维护和系统安全至关重要.
- 传统的数据驱动神经网络方法通常会产生不切实际的RUL估计,因为它们不考虑潜在的物理性质.
- 现有的方法可能会产生错误的预测,例如增加RUL估计或错误的终端预测.
研究的目的:
- 通过整合物理信息神经网络 (PINNs) 的原则来增强基于RNN的RUL估计.
- 提高动力电子设备的RUL预测的现实性和准确性.
- 使用NASA IGBT数据集验证PINN增强RNN的有效性.
主要方法:
- 应用物理信息神经网络 (PINNs) 来增强用于RUL估计的标准RNN架构.
- 在训练过程中将物理约束纳入神经网络的损失功能.
- 使用NASA IGBT数据集评估性能,与基线RNN和长短期记忆 (LSTM) 模型进行比较.
主要成果:
- 与传统神经网络相比,以物理为基础的RNN模型展示了更现实的训练行为.
- 观察到估计准确度的显著改善,其证据是平均平方误差 (MSE) 的减少.
- 与PINN增强的RNN相比,PINN增强的RNN在培训中平均提高了24.7%,在测试中提高了51.3%.
- 与基线LSTM相比,PINN增强的LSTM显示MSE平均在培训中改善15.3%,在测试中改善13.9%.
结论:
- 通过PINNs集成物理原理有效地解决了纯粹基于数据的RUL估计的局限性.
- 用PINN增强的RNN和LSTM模型为动力电子提供了更准确和可靠的RUL预测.
- 这种方法提高了预测性维护能力,并有助于提高系统安全性和运营效率.
相关概念视频
01:26Linear Approximation in Frequency Domain
116
Linear systems are characterized by two main properties: superposition and homogeneity. Superposition allows the response to multiple inputs to be the sum of the responses to each individual input. Homogeneity ensures that scaling an input by a scalar results in the response being scaled by the same scalar.
In contrast, nonlinear systems do not inherently possess these properties. However, for small deviations around an operating point, a nonlinear system can often be approximated as linear....
In contrast, nonlinear systems do not inherently possess these properties. However, for small deviations around an operating point, a nonlinear system can often be approximated as linear....
116
01:17Energy and Power Signals
354
In an electrical system with a resistor, voltage and current signals facilitate the measurement of power and energy across the resistor. For a continuous-time signal, the total energy over a time interval is defined as the integral of the square of the signal's magnitude over that interval. Mathematically, this is expressed as:
354
01:15Power in a Three-Phase Circuit
368
Three-phase systems have two configurations: the wye and delta. A star configuration can be three or four wires; in a delta configuration, the components are connected in a closed loop. Instantaneous power refers to the power value at a precise moment, and in a balanced three-phase system, it is constant. This is because the sum of the instantaneous powers in the three phases remains steady over time, despite individual fluctuations, due to the symmetry and phase relationship. The total...
368
01:07Electrical Power
3.1K
Electric power is the product of current and voltage, represented in units of joules per second, or watts. For example, cars often have one or more auxiliary power outlets with which you can charge a cell phone or other electronic devices. These outlets may be rated at 20 amps and 12 volts, so that the circuit can deliver a maximum power of 240 watts. Consider a 25 Watt bulb and a 60 Watt bulb. The conversion of electrical energy produces heat and light, while the kinetic energy lost by the...
3.1K
01:25Multimachine Stability
198
Multimachine stability analysis is crucial for understanding the dynamics and stability of power systems with multiple synchronous machines. The objective is to solve the swing equations for a network of M machines connected to an N-bus power system.
In analyzing the system, the nodal equations represent the relationship between bus voltages, machine voltages, and machine currents. The nodal equation is given by:
In analyzing the system, the nodal equations represent the relationship between bus voltages, machine voltages, and machine currents. The nodal equation is given by:
198
01:13Average Power
634
In practical electrical applications, the concept of time-varying instantaneous power is not frequently utilized. Instead, focus shifts to the more practical quantity known as average power. Average power is determined by integrating the instantaneous power over a specified time period and subsequently dividing it by that duration.
634