A photovoltaic power generation power prediction method using long short-term memory network

Abstract

The invention discloses a photovoltaic power generation power prediction method utilizing a long short-term memory network and belongs to the technical field of photovoltaic power generation power prediction. Using the photovoltaic power generation power parameters of the long and short-term memory network to build a long-term and short-term memory network prediction model: Build a long-term and short-term memory network with a hidden layer containing several neurons, and use the relevant five-dimensional feature vectors: accumulated days, ambient temperature, and ambient humidity. , wind speed and solar irradiance, and the photovoltaic power value and weather data at 24 hours a day in the 30 days before the forecast point of the next day as the original data, these five-dimensional vectors are formed into an input matrix and input into the long-term and short-term memory network. Power prediction at the prediction point; compared with all prediction methods, the present invention establishes a connection between the photovoltaic power change at the current moment and the previous photovoltaic power change, realizes the dynamic modeling of time series data, and can more fully reflect the photovoltaic power The change rule of power can realize more accurate photovoltaic power prediction.

Description

technical field [0001] The invention belongs to the technical field of photovoltaic power generation power prediction, and in particular relates to a photovoltaic power generation power prediction method using a long short-term memory network. Background technique [0002] Photovoltaic system power generation has volatility and periodicity due to the influence of external environmental factors such as weather conditions, day-night alternation and seasonal changes. , periodic shocks. Accurate photovoltaic power prediction is the premise to ensure the safe and stable operation of photovoltaic grid-connected power generation, and it is also an important basis for rational distribution and scheduling of photovoltaic system power generation. [0003] According to the classification of forecast time, photovoltaic power forecasting can be divided into short-term photovoltaic power forecasting and mid- and long-term photovoltaic power forecasting. For short-term photovoltaic power...

AI Technical Summary

Problems solved by technology

[0009] Step 4: Establish a long-short-term memory network prediction model: the long-short-term memory network (Long-Short TermMemory, LSTM) is an improvement on the traditional cyclic neural network. In order to solve the problem that the traditional cyclic neural network cannot be optimized because the sequence is too long, LSTM adds memory units while maintaining the traditional cyclic neural network structure; there is a cell (cell) in each LSTM unit, which is regarded as a memory unit of LSTM and is used to describe the current state of the LSTM unit; the current state of the LSTM unit The state is controlled by 3 control gates, the 3 control gates are the input gate, the output gate and the forgetting gate, and the three gates respectively control the input, output and the state of the cell unit of the network; specifically, at each moment, the LSTM unit receives the input After information, each gate will perform calculations on inputs from different sources to determine whether the input information is passed; the input of the input gate is transformed by a nonlinear function, and then superimposed with the state of the memory unit processed by the forget gate to form a new state of the memory unit; In the end, the state of the memory unit forms the output of the LSTM unit through the operation of the nonlinear function and the dynamic control of the output gate; the gate unit is an operation that uses a neural network and a bitwise multiplication, and these two operations together are a Gate unit, the weight of the neural network in the gate unit is learned through the training process, LSTM relies on the gate unit to allow information to selectively affect the state of each moment in the recurrent neural network

Images