Wind power generation power prediction method and system

Abstract

The present application relates to the field of wind power generation technology, and in particular to a method and system for predicting wind power generation power, the method comprising: collecting historical sequences of power generation and multiple environmental parameters, obtaining the mutation moment of any historical sequence, and obtaining the change characteristics of each historical sequence; initializing the offset of each environmental parameter, shifting the historical sequence of the environmental parameter in the positive direction along the time axis by the offset, calculating the correlation between the change characteristics of the environmental parameter and the power generation in the overlapping time period, and taking the maximum value of the correlation and the offset corresponding to the maximum value of the correlation as the influence coefficient and lag period of the environmental parameter respectively; after determining the input sequence of each environmental parameter at the current moment according to the lag period, inputting the sequence into the prediction model to obtain the prediction result of the power generation. Through the technical solution of the present application, the prediction result of the wind power generation can be accurately obtained.

Description

Technical Field

[0001] The present application relates to the field of wind power generation technology, and in particular to a wind power generation power prediction method and system. Background Art

[0002] As an intermittent clean energy source, wind power generation has large fluctuations in power generation. In order to facilitate the power dispatching department to reasonably arrange the output of wind farms and optimize grid dispatching, it is necessary to accurately predict the power generation of wind power and provide reference information for optimizing grid dispatching.

[0003] At present, a patent application document with publication number CN119578657 A discloses a method for predicting the power generation of wind power generation, which includes an environmental monitoring step, a preliminary analysis step and a power prediction step; the environmental monitoring step is used to monitor the environmental conditions of the area corresponding to the wind power generation device and obtain environmental information corresponding to the target area; the preliminary analysis step is used to perform a preliminary analysis on the environmental information corresponding to the target area and obtain a prediction difficulty coefficient for a power generation prediction period corresponding to the wind power generation device; the power prediction step is used to perform a comprehensive analysis based on the environmental information corresponding to the target area and the prediction difficulty coefficient for the power generation prediction period corresponding to the wind power generation device and obtain a power generation prediction result for the power generation prediction period corresponding to the wind power generation device.

[0004] The above method evaluates the prediction difficulty coefficient of the power generation prediction period through environmental information, and then predicts the power generation result of the power generation prediction period. However, the degree of influence of different environmental information on wind power generation varies. The above method ignores this difference between environmental information, resulting in inaccurate wind power generation prediction results and unable to provide effective reference information for optimizing power grid scheduling. Summary of the Invention

[0005] In order to solve the technical problem of inaccurate prediction results of wind power generation, the present application provides a wind power generation prediction method and system, which can accurately obtain the prediction results of wind power generation.

[0006] In a first aspect, the present application provides a method for predicting wind power generation, which includes: collecting historical sequences of power generation and multiple environmental parameters within a historical time period, obtaining the mutation moment of any historical sequence based on an ordered sample clustering algorithm, marking the mutation moment as 1, and marking other moments as 0, to obtain the change characteristics of each historical sequence; initializing the offset of each environmental parameter, shifting the historical sequence of the environmental parameter forward along the time axis by the offset, calculating the correlation between the change characteristics of the environmental parameter and the power generation within the overlapping time period, taking the offset corresponding to the maximum value of the correlation within a preset offset range as the lag period of the environmental parameter, and taking the normalized result of the maximum value of the correlation as the influence coefficient of the environmental parameter; determining the input sequence of each environmental parameter at the current moment based on the lag period, inputting each input sequence into a prediction model to obtain a prediction result of the power generation at the next adjacent moment; the prediction model extracts the time series characteristics of each input sequence, weightedly sums the time series characteristics based on the influence coefficient of each environmental parameter, and regresses the weighted sum result into a prediction result.

[0007] Preferably, obtaining the mutation moment of any historical sequence based on the ordered sample clustering algorithm includes: using ordered sample clustering to divide any historical sequence into an initial number of sequence segments, the initial number of sequence segments corresponding to the initial number of clustering results, calculating the numerical variance of each subsequence segment, and taking the maximum variance as the clustering effect of the initial number; updating the initial number, and drawing a clustering effect curve with the initial number as the horizontal coordinate and the clustering effect as the vertical coordinate, and taking the initial number corresponding to the inflection point of the clustering effect curve as the target number; and taking the segmentation point of any adjacent sequence segments in the target number of clustering results as the mutation moment of the historical sequence.

[0008] The ordered sample clustering algorithm is used to cluster the historical sequences, determine the segmentation points of multiple sequence segments, and then quickly determine all mutation moments in the historical sequences.

[0009] Preferably, the correlation is negatively correlated with the Hamming distance between the change characteristics of the environmental parameter and the generated power in the overlapping time period.

[0010] Preferably, the maximum value of the preset offset range is a preset proportion of the number of moments in the historical time period.

[0011] The length of the overlapping time period directly affects the accuracy of the correlation. As the offset increases, the length of the overlapping time period will gradually decrease. At this time, the correlation cannot accurately reflect the distribution consistency of the mutation moments between environmental parameters and power generation power. Therefore, it is necessary to set a preset offset range for the offset to ensure the accuracy of the correlation.

[0012] Preferably, the environmental parameters The influence coefficient for: , Environmental parameters The maximum correlation value of is the sum of the maximum correlation values ​​of each environmental parameter.

[0013] Environmental parameters The proportion of the possibility of causing a sudden change in power generation in all environmental parameters can accurately quantify environmental parameters The degree of impact on sudden changes in power generation.

[0014] Preferably, the method of determining the input sequence of each environmental parameter at the current moment based on the lag period includes: taking the current moment as the end point of a preset time window, and shifting the preset time window in the opposite direction along the time axis by the lag period of any environmental parameter, and taking the value within the preset time window as the input sequence of the environmental parameter.

[0015] The input information of each environmental parameter is determined according to the hysteresis period, which avoids the prediction error caused by the hysteresis of environmental parameters on wind power generation and improves the accuracy of the prediction results.

[0016] Preferably, the training method of the prediction model includes: taking the input sequence of each environmental parameter at any time in the historical time period as a training sample, and taking the power generation power at the next adjacent time as the label of the training sample; inputting the training sample into the prediction model to obtain the prediction result; using the gradient descent method to update the prediction model until the loss function is less than the preset loss, and stopping, wherein the loss function is the mean square error loss between the label and the prediction result.

[0017] Preferably, the loss function for:

[0018] ; For the moment The labels corresponding to the training samples, To predict the results, is the mean squared error loss between the label and the prediction result, is the changing characteristics of the historical sequence of power generation at the moment The numerical value of .

[0019] It is more difficult to predict the sudden changes in the historical sequence of power generation. During the training process, more attention is paid to the prediction results of the sudden changes in power generation, ensuring that the prediction model can learn the mapping relationship between multiple environmental parameters and power generation, and output accurate prediction results.

[0020] In a second aspect of the present application, a wind power generation prediction system is provided, comprising a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, a wind power generation prediction method according to the first aspect of the present application is implemented.

[0021] The technical solution of this application has the following beneficial technical effects:

[0022] Mutation point detection is performed on the historical sequences of power generation and various environmental parameters within the historical time period to obtain the change characteristics of each historical sequence. The change characteristics can reflect the distribution of the mutation moments of the corresponding historical sequences within the historical time period. Considering that the impact of environmental parameters on wind power generation has a hysteresis, the historical sequence of any environmental parameter is shifted forward along the time axis by an offset, and the correlation between the change characteristics of the environmental parameter and power generation in the overlapping time period is calculated. The greater the correlation, the greater the probability that the power generation will change when the environmental parameter changes, indicating that the change of the environmental parameter will cause the change of the power generation. The offset corresponding to the maximum correlation value is taken as the hysteresis period of the environmental parameter, and the normalized result of the maximum correlation value is taken as the influence coefficient of the environmental parameter. The influence coefficient can characterize the degree of influence of the environmental parameter on the mutation of the power generation. The input sequence of each environmental parameter at the current moment is determined based on the hysteresis period, and the time series features of each input sequence are extracted using the prediction model. The time series features are weighted and summed according to the influence coefficient of each environmental parameter to output an accurate power generation prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of a wind power generation power prediction method according to an embodiment of the present application.

[0024] Figure 2 This is a schematic diagram of shifting the historical sequence of environmental parameters in the positive direction along the time axis by an offset according to an embodiment of the present application.

[0025] Figure 3 This is a structural block diagram of a wind power generation prediction system according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0027] It should be understood that when the terms "first," "second," etc. are used in the claims, specification, and drawings of this application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "comprise" and "comprising" used in the specification and claims of this application indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or combinations thereof.

[0028] According to a first aspect of the present application, the present application provides a method for predicting wind power generation. Figure 1 This is a flow chart of a method for predicting wind power generation according to an embodiment of the present application. Figure 1 As shown, the wind power generation power prediction method includes steps S101 to S103, which are described in detail below.

[0029] S101, collect historical sequences of power generation and various environmental parameters within a historical time period, obtain the mutation moment of any historical sequence based on the ordered sample clustering algorithm, mark the mutation moment as 1, and mark other moments as 0, and obtain the change characteristics of each historical sequence.

[0030] In one embodiment, the historical time period can be the past day or the past month, and this application does not impose any restrictions. The longer the historical time period, the greater the amount of data contained in the collected historical sequence, which is more conducive to obtaining accurate prediction results. The environmental parameters include at least wind speed and wind direction. The power generation power and various environmental parameters at each moment in the historical time period are collected according to a preset collection frequency to obtain multiple historical sequences, and each historical sequence has the same length.

[0031] In other embodiments, the environmental parameters may also include other parameters such as temperature, humidity, and air pressure.

[0032] In one embodiment, obtaining the mutation moment of any historical sequence based on the ordered sample clustering algorithm includes: using ordered sample clustering to divide any historical sequence into an initial number of sequence segments, the initial number of sequence segments corresponding to the initial number of clustering results, calculating the numerical variance of each subsequence segment, and taking the maximum variance as the clustering effect of the initial number; updating the initial number, and drawing a clustering effect curve with the initial number as the horizontal coordinate and the clustering effect as the vertical coordinate, and taking the initial number corresponding to the inflection point of the clustering effect curve as the target number; and taking the segmentation point of any adjacent sequence segments in the target number of clustering results as the mutation moment of the historical sequence.

[0033] The initial value is 2.

[0034] Among them, ordered sample clustering is a clustering algorithm for time series data. Ordered sample clustering can continuously find split points in time series data, dividing the time series data into multiple data segments. The values ​​within each data segment are basically consistent, and the values ​​of adjacent data segments vary greatly. For example, when ordered sample clustering finds a split point in time series data, the split point can divide the time series data into two data segments with a large difference in values ​​between the two data segments. When ordered sample clustering finds three split points, the split points can divide the time series data into three data segments. The ordered sample clustering algorithm is used to cluster historical sequences, determine the split points of multiple sequence segments, and then quickly determine all mutation moments in the historical sequence.

[0035] In one embodiment, after determining the mutation moment of any historical sequence, the mutation moment is marked as 1 and other moments are marked as 0 to obtain the change characteristics of each historical sequence. The change characteristics can reflect the distribution of the mutation moments of the corresponding historical sequence within the historical time period.

[0036] S102, initialize the offset of each environmental parameter, shift the historical sequence of the environmental parameters forward along the time axis by the offset, calculate the correlation between the change characteristics of the environmental parameters and the power generation power in the overlapping time period, use the offset corresponding to the maximum value of the correlation within the preset offset range as the lag period of the environmental parameter, and use the normalized result of the maximum value of the correlation as the influence coefficient of the environmental parameter.

[0037] In one embodiment, during the wind power generation process, environmental parameters will affect the speed and other operating conditions of the wind turbine, thereby affecting the power generation. However, changes in environmental parameters will not immediately cause changes in the operating conditions of the wind turbine. In other words, the impact of environmental parameters on wind power generation has a lag. For example, environmental parameters At the moment has changed, and the environmental parameters At the moment Changes will happen at any time It will cause the change of power generation. and time The time interval between them is recorded as the environmental parameter lag period.

[0038] Environmental parameters For example, the process of obtaining the environmental parameter hysteresis period is described in detail. Initialize the environmental parameters In the embodiment of the present application, the initialization offset value is 1; see Figure 2 , is a schematic diagram of shifting the historical sequence of environmental parameters along the time axis in the positive direction according to an embodiment of the present application. Assuming that the historical time period includes T moments in total, that is, each historical sequence has T values, the environmental parameters are shifted by After the historical sequence is positively shifted along the time axis by the offset, the overlapping time period includes the part from t1 to T-1 in the historical sequence of environmental parameters. Based on the part from t2 to T in the historical sequence of power generation, the correlation of the change characteristics in the overlapping time period can be calculated.

[0039] Specifically, since the value in the change feature is 0 or 1, the correlation is negatively correlated with the Hamming distance of the change feature of the environmental parameter and the power generation in the overlapping time period. At offset Time correlation for:

[0040] , Environmental parameters and the generated power at the offset The Hamming distance of the changing features in the overlapping time periods.

[0041] Understandably, when environmental parameters When the Hamming distance between the change characteristics of the environmental parameters and the power generation in the overlapping time period is 0, it means that the environmental parameters The historical sequence is shifted in the positive direction along the time axis After that, environmental parameters The distribution of the mutation time is exactly the same as that of the power generation time, indicating that the environmental parameters Changes in will cause changes in power generation, and environmental parameters and power generation have a strong correlation in the overlapping time period.

[0042] The length of the overlapping time period directly affects the accuracy of the correlation. As the offset increases, the length of the overlapping time period will gradually decrease. At this time, the correlation cannot accurately reflect the distribution consistency of the mutation moments between the environmental parameters and the generated power. Therefore, it is necessary to set a preset offset range for the offset. The maximum value of the preset offset range is a preset ratio of the number of moments in the historical time period. In the embodiment of the present application, the preset offset range is 0.5 of the number of moments in the historical time period, that is, if the historical time period includes a total of T moments, the preset offset range is , that is to say, the minimum length of the overlapping time period is 0.5T, ensuring that the relevant performance accurately reflects the distribution consistency of the mutation moments between environmental parameters and power generation power.

[0043] Calculate environmental parameters using exhaustive method within the preset offset range The correlation at each offset, the offset corresponding to the maximum correlation value is used as the environmental parameter The hysteresis period indicates that when the environmental parameters When a sudden change occurs, it will cause a sudden change in power generation after a lag period, and the correlation corresponding to the lag period (i.e., the maximum correlation value) can quantify the environmental parameters. The possibility of causing a sudden change in power generation.

[0044] Furthermore, environmental parameters The proportion of the possibility of causing a sudden change in power generation in all environmental parameters can characterize the environmental parameters The degree of impact on the sudden change of power generation. Specifically, the environmental parameters The influence coefficient for: , Environmental parameters The maximum correlation value of is the sum of the maximum correlation values ​​of each environmental parameter.

[0045] In this way, the lag period of each environmental parameter and the degree of influence of each environmental parameter on the power generation power during the lag period are determined according to the historical sequence of power generation power and environmental parameters, which provides reference data for determining the input information of the prediction model, so that the prediction model can take into account the degree of influence of each environmental parameter on wind power generation power, as well as the lag of each environmental parameter relative to wind power generation power, so that the prediction model can accurately obtain the prediction results of wind power generation power.

[0046] S103, determining an input sequence of each environmental parameter at the current moment according to the hysteresis period, and inputting each input sequence into a prediction model to obtain a prediction result of the generated power at the next adjacent moment.

[0047] In one embodiment, the method of determining the input sequence of each environmental parameter at the current moment based on the lag period includes: taking the current moment as the end point of a preset time window, and shifting the preset time window in the opposite direction along the time axis by the lag period of any environmental parameter, and taking the value within the preset time window as the input sequence of the environmental parameter.

[0048] For example, the length of the preset time window is 10, and the current time is recorded as , environmental parameters The lag period is 1, then the environmental parameters The input sequence is the time At the time Environmental parameters The numerical value of .

[0049] After obtaining the input sequences of each environmental parameter, each input sequence is input into a prediction model. The prediction model extracts the time series characteristics of each input sequence, performs a weighted summation of the time series characteristics based on the influence coefficient of each environmental parameter, and regresses the weighted summation result into a prediction result. The prediction model includes multiple time series sub-models and regression sub-models. Each time series sub-model corresponds to an environmental parameter and is used to extract the time series characteristics of the corresponding environmental parameter input sequence. After performing a weighted summation of the time series characteristics based on the influence coefficient of each environmental parameter, the weighted summation result is input into the regression sub-model, and the prediction result of the generated power is output.

[0050] Among them, the timing sub-model can adopt a recurrent neural network such as LSTM or Transformer, and the regression sub-model adopts a fully connected neural network.

[0051] In one embodiment, in order to ensure that the prediction model can output accurate prediction results, the prediction model needs to be trained. The training method of the prediction model includes: taking the input sequence of each environmental parameter at any time in the historical time period as a training sample, and taking the power generation power at the next adjacent time as the label of the training sample; inputting the training sample into the prediction model to obtain the prediction result; using the gradient descent method to update the prediction model until the loss function is less than the preset loss, and stopping. The loss function is the mean square error loss between the label and the prediction result.

[0052] Wherein, the preset loss is 0.01.

[0053] In another embodiment, it is difficult to predict the mutation moment in the historical sequence of power generation. In order to ensure that the prediction model can output accurate prediction results at the mutation moment, more attention should be paid to the mutation moment of power generation in the historical time period during the training process. Therefore, the loss function for: ; For the moment The labels corresponding to the training samples, To predict the results, is the mean squared error loss between the label and the prediction result, is the changing characteristics of the historical sequence of power generation at the moment The numerical value of .

[0054] In this way, the training of the prediction model is completed. During the training process, more attention is paid to the prediction results at the moment of sudden change in power generation, ensuring that the prediction model can learn the mapping relationship between multiple environmental parameters and power generation, and output accurate prediction results.

[0055] The above describes the technical principles and implementation details of a wind power generation prediction method of the present application through specific embodiments. First, mutation point detection is performed on the historical sequences of power generation and multiple environmental parameters within a historical time period to obtain the change characteristics of each historical sequence, which can reflect the distribution of the mutation moments of the corresponding historical sequences within the historical time period; considering that the impact of environmental parameters on wind power generation has a hysteresis, the historical sequence of any environmental parameter is shifted forward along the time axis by an offset, and the correlation between the change characteristics of the environmental parameter and power generation in the overlapping time period is calculated. The greater the correlation, the greater the probability that the power generation will change when the environmental parameter changes, indicating that the change of the environmental parameter will cause the change of power generation. The offset corresponding to the maximum value of the correlation is used as the hysteresis period of the environmental parameter, and the normalized result of the maximum value of the correlation is used as the influence coefficient of the environmental parameter. The influence coefficient can characterize the degree of influence of the environmental parameter on the mutation of power generation; the input sequence of each environmental parameter at the current moment is determined based on the hysteresis period, the time series features of each input sequence are extracted using a prediction model, and the time series features are weighted and summed according to the influence coefficient of each environmental parameter to output an accurate power generation prediction result.

[0056] According to the second aspect of the present application, the present application also provides a wind power generation power prediction system. Figure 3 This is a structural block diagram of a wind power generation prediction system according to an embodiment of the present application. Figure 3 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions. When the computer program instructions are executed by the processor, a method for predicting wind power generation according to the first aspect of the present application is implemented. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are well known in the art and are therefore not described in detail here.

[0057] It should be noted that a person skilled in the art may make a number of modifications and improvements without departing from the concept of the present application, and these modifications and improvements are all within the scope of protection of the present application. Therefore, the scope of protection of the patent application shall be based on the appended claims.

Claims

1. A method for predicting wind power generation, characterized in that: The prediction method comprises: Collect historical sequences of power generation and various environmental parameters within a historical period, and use the ordered sample clustering algorithm to obtain the mutation moment of any historical sequence. Mark the mutation moment as 1 and other moments as 0 to obtain the change characteristics of each historical sequence. Initialize the offset of each environmental parameter, shift the historical sequence of the environmental parameters forward along the time axis by the offset, calculate the correlation between the change characteristics of the environmental parameters and the generated power in the overlapping time period, use the offset corresponding to the maximum correlation value within the preset offset range as the lag period of the environmental parameter, and use the normalized result of the maximum correlation value as the influence coefficient of the environmental parameter; Determine the input sequence of each environmental parameter at the current moment based on the hysteresis period, and input each input sequence into the prediction model to obtain the prediction result of the power generation at the next adjacent moment; The prediction model extracts the time series features of each input sequence, performs weighted summation on the time series features according to the influence coefficient of each environmental parameter, and regresses the weighted summation result into the prediction result.

2. A wind power generation power prediction method according to claim 1, characterized in that: The mutation moments of any historical sequence are obtained based on the ordered sample clustering algorithm, including: Using ordered sample clustering, any historical sequence is divided into an initial number of sequence segments, wherein the initial number of sequence segments corresponds to the initial number of clustering results, the numerical variance of each subsequence segment is calculated, and the maximum variance is taken as the initial number of clustering effects; Update the initial quantity, and draw a clustering effect curve with the initial quantity as the horizontal axis and the clustering effect as the vertical axis. The initial quantity corresponding to the inflection point of the clustering effect curve is used as the target quantity. The split points of any adjacent sequence segments in the target number of clustering results are used as the mutation moments of the historical sequence.

3. The method for predicting wind power generation according to claim 1, wherein: The correlation is negatively correlated with the Hamming distance between the variation characteristics of the environmental parameters and the generated power in the overlapping time period.

4. A wind power generation power prediction method according to claim 1, characterized in that: The maximum value of the preset offset range is a preset ratio of the number of moments in the historical time period.

5. The method for predicting wind power generation according to claim 1, wherein: Environmental parameters The influence coefficient for: , Environmental parameters The maximum correlation value of is the sum of the maximum correlation values ​​of each environmental parameter.

6. A wind power generation power prediction method according to claim 1, characterized in that: The input sequence of determining each environmental parameter at the current moment according to the hysteresis period includes: The current moment is taken as the end point of the preset time window, and after the preset time window is shifted in the reverse direction along the time axis by the lag period of any environmental parameter, the values ​​within the preset time window are taken as the input sequence of the environmental parameter.

7. A wind power generation power prediction method according to claim 1, characterized in that: The training methods for the prediction model include: The input sequence of each environmental parameter at any time in the historical time period is used as a training sample, and the power generation power at the next adjacent time is used as the label of the training sample; Input the training samples into the prediction model to obtain the prediction results; The prediction model is updated using the gradient descent method until the loss function is less than the preset loss, where the loss function is the mean square error loss between the label and the prediction result.

8. A wind power generation power prediction method according to claim 7, characterized in that: The loss function for: ; For the moment The labels corresponding to the training samples, To predict the results, is the mean squared error loss between the label and the prediction result, is the changing characteristics of the historical sequence of power generation at the moment The numerical value of .

9. A wind power generation power prediction method according to claim 1, characterized in that: The environmental parameters include at least wind speed and wind direction.

10. A wind power generation power prediction system, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a wind power generation power prediction method according to any one of claims 1 to 9 is implemented.

Images