基于无线传感器网络的风电场在线监测和动态风速预测

Wind speed forecasting is of great significance for wind turbine control, power transmission and distribution, and the stable operation of the grid. However, the wind speed has great stochastic volatility, due to the geographical and other environmental factors, which is considered as one of the most difficult parameters to be predicted. There have been a lot of researches on wind speed forecast recently. However, the prediction error is still unacceptable, and with low temporal-spatial resolution. Therefore, we propose a distributed on-line monitoring technology, and study the corresponding dynamic wind speed forecasting method based on the monitoring information. Considering the remote location of the wind farm, we use the inexpensive, ad-hoc wireless sensor networks to on-line monitor the wind speed and its surrounding environment, studying the key issues such as sensor node localization, and network energy management; Based on the limited information sensed by the sensors, we reconstruct the large-scale field distribution of the wind speed using the Computational Fluid Dynamics （CFD) method. By dividing the whole farm into separated computation areas, and performing parallel computation, the speed of CFD has been greatly promoted, while improving the spatial resolution of the problem; Based on the variational Bayesian method, we fuse the monitoring data of multi-mode, uncertainty and imperfection, and forecast the wind speed in real time. Specifically, considering the random, nonlinear, and dynamic characteristics of the wind speed, we integrate a variety of uncertain stochastic distribution functions to construct a variational model, weakening uncertainty and imperfection of the monitoring information. Then we forecast the dynamic wind speed distribution by Bayesian inference, based on the variational model. According to the performance difference, the model parameters are automatically and dynamically corrected online, thus ensuring the forecasting accuracy and increasing the computation speed, realizing a breakthrough of the wind speed forecasting method. This research will provide new technology to monitor the wind farm, leading to the formation of a more systematic wind speed forecasting theory, thereby establishing the theoretical basis for decision-making systems to implement control programs and response plans, improving the quality of the output power, and finally resolving wind power connection problems.

风速预测对于风力发电机组控制、电能输配送乃至电网稳定运行具有重要意义。然而，风速受地理环境等因素影响，具有很大的随机波动性，被认为是最难准确预测的参数之一，相关研究普遍存在着预测误差大，空间分辨率低等问题。本课题从在线监测、风速分布构建和预测多个层次提升数据精度和计算速度，完善风速预测理论。基于无线传感器网络实现对风电场及周边环境的多参数和全方位在线监测；在有限测点的基础上构建时-空解析的风场分布，将风场尺度的流体力学计算分解为风机尺度的区域并行计算，提升计算速度和空间分辨率；构造变分模型，融合传感器监测数据和风场分布信息进行贝叶斯推理，在动态预测风速的同时修正模型参数，保障预测精度，使控制系统在所需时间分辨率下获得准确的风速分布预测，从而改善风机运行状态，提高输出电能质量。本课题研究将为风电场监测提供新的技术模式，形成较为系统的风速预测理论，为电能质量评估奠定理论基础，消除风电并网瓶颈。

风速的随机性是风力发电机面对的主要问题。在波动的风速中，虽然风电机叶轮仍然显得是在平滑运转，但是由于桨叶上的力矩变化使得风电机输出的电功率具有剧烈的波动性。对于分钟级别或更长时间尺度，风能向电能的转换往往可以参考标准功率曲线。但是在短时间尺度上，转换曲线却可能会偏离标准功率曲线，具有更加复杂的模式，例如出现多重分性等湍流特征。如果能动态以及预先提供风速和风向，将风速矢量信息作为参数进入控制系统，则可以辅助风车进行主动调整姿态，使发电机能运行在最优效率区，并能对随机的风力变化提前准备，及时调整，这对于风力发电机组控制系统的研发和完善具有非常重要的意义。本课题从在线监测、风速分布构建和预测等多个层次提升数据精度和计算速度，完善风速预测理论。基于无线传感器网络实现对风电场及周边环境的多参数和全方位在线监测；在有限测点的基础上构建时-空解析的风场分布，将风场尺度的流体力学计算分解为风机尺度的区域并行计算，从而帮助改善风机运行状态，提高输出电能质量。项目研究成果主要包括基于变分滤波算法的无线传感器网络节点定位，基于多重信号分类算法的三维风速风向估计，以及基于计算流体力学的动态风速场构建三个方面。本课题的研究对于风力发电机组控制系统的研发和完善具有非常重要的意义，通过准确地提供输出风速、风向等矢量预测值，从而辅助风力发电机组控制风车主动调整姿态，使发电机运行在最优效率区，及时应对随机变化的风力，为系统输出电能质量的评估和运行时设备重构提供依据，为风电大规模并网奠定了理论基础。