Wind Park Grid Frequency Forecasting With LSTM Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for predicting utility grid frequency are not always accurate and reliable, particularly in the context of wind parks, where fluctuations in wind power generation can destabilize the grid frequency, necessitating improved prediction methods with enhanced accuracy and reduced complexity.
Innovation Solution
A method using a recurrent neural network (RNN), specifically a Long Short-Term Memory (LSTM) network, to predict grid frequency by integrating utility grid and wind park measurement data, including active and reactive power, to provide precise forecasts of frequency deviations within seconds, enabling proactive control measures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If frequency measurements are taken only at the moment of synchronization events, then the synchronization process remains simple and computationally efficient, but the accuracy and reliability of frequency measurements deteriorate when synchronization events are infrequent or absent
Solution Approach 1:
The system performs preliminary frequency measurements continuously or at regular intervals before synchronization events occur. These preliminary measurements are stored and used to calculate the frequency measurement when the synchronization event occurs, ensuring accurate frequency data is available even when synchronization events are infrequent.
Solution Approach 2:
The system introduces an intermediary mechanism that collects frequency measurement data from multiple sources and time points, then synthesizes this data to provide an accurate frequency measurement at the synchronization event. This intermediary layer decouples the frequency measurement process from the synchronization event frequency.
2Speed
If the system waits for synchronization events to obtain frequency measurements, then the system maintains simplicity and lower computational overhead, but the responsiveness and timeliness of frequency information deteriorate
Solution Approach 1:
The system performs frequency measurements periodically at regular intervals independent of synchronization events. This periodic measurement approach ensures that frequency information is always available and up-to-date, improving responsiveness while maintaining a predictable and manageable computational workload through regular scheduling.
Solution Approach 2:
The system performs frequency measurements in advance before synchronization events occur, storing these measurements for later use. This preliminary action ensures that when a synchronization event occurs, the frequency information is already available immediately, improving responsiveness without requiring intensive real-time computation during the synchronization event.
3Reliability
If frequency measurements rely solely on synchronization events between base stations and terminal devices, then the measurement process remains simple, but the reliability and availability of frequency measurements deteriorate in scenarios with infrequent synchronization
Solution Approach 1:
The system performs frequency measurements in advance before synchronization events and stores these measurements. This ensures that even if synchronization events are infrequent or delayed, accurate frequency measurements are already available, improving reliability without requiring complex real-time measurement mechanisms.
Solution Approach 2:
The system introduces an intermediary data structure that accumulates frequency measurement data from multiple sources and time points. This intermediary layer allows the system to synthesize reliable frequency measurements even when synchronization events are sparse, decoupling measurement reliability from synchronization frequency.
Data Source
Figure 1~2
Figure 3~4
AI summary
It is described a method of predicting a frequency value of a utility grid (3) to which a wind park (1) is connected, the method comprising: obtaining plural utility grid measurement values (7) pertaining to a predetermined time range before and until a present point in time; obtaining plural wind park measurement values (12) pertaining to the time range; feeding the plural utility grid measurement values (7) and the plural wind park measurement values (12) into a recurrent neural network trained to output the frequency value at at least one next point in time, the next point in time being in particular between 0.5 s and 2 s after the present point in time.