A resource configuration method and device for responding to regional offshore wind power demand

Abstract

The present invention discloses a method and device for configuring demand response resources for regional offshore wind power, the method comprising: obtaining historical output data of regional offshore wind power, and preprocessing the historical output data; using the Time2Graph algorithm to extract time-aware feature subsequences from the preprocessed historical output data; mining the transfer of the feature subsequences based on a graph neural network to analyze the dynamic change characteristics of regional offshore wind power output in different periods; and selecting offshore wind power demand response resources that can match the dynamic change characteristics from a demand response resource matching library. The present invention realizes accurate analysis of feature subsequences during periods of large output fluctuations, thereby being able to match and configure DR resources with strong adaptability, so as to promote the effective consumption of regional offshore wind power, improve the stability of the power system, and reduce the impact of fluctuations in the output of new energy on the power system.

Description

Technical Field

[0001] The present invention relates to renewable energy configuration technology, and in particular to a resource configuration method and device for responding to regional offshore wind power demand. Background Art

[0002] Offshore wind power is one of the most promising areas for large-scale development in renewable energy and is also at the forefront of new energy development. With the increase of offshore wind power, the decrease of cost and the improvement of prediction accuracy, more offshore wind power will participate in market competition in the future. However, due to its inherent randomness, volatility and anti-peak characteristics, the actual output is inconsistent with the prediction, which may lead to large losses.

[0003] Demand response (DR) is an important means for the new generation of energy systems to cope with power generation uncertainty, load demand fluctuations, and promote the absorption of a high proportion of renewable energy. It has significant grid benefits. On the one hand, it helps to reduce the peak load demand of the power system and postpone the corresponding investment in power generation and transmission; on the other hand, during the off-peak period of the power grid, demand response can improve the absorption level of new energy, reduce the high start-up and shutdown costs of thermal power units, and effectively improve the safe operation level of the power grid. Therefore, how to configure sufficient demand response resources that match the characteristics of regional offshore wind power output fluctuations will become the key to promoting offshore wind power absorption and improving the safe operation level of the power grid. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies of the above-mentioned prior art and to provide a method and device for resource allocation in response to demand for regional offshore wind power, so as to promote the effective consumption of regional offshore wind power, improve the stability of the power system, reduce the impact of fluctuations in the output of renewable energy on the power system, and improve the safe operation level of the power grid.

[0005] To achieve the above object, the technical solution of the present invention is:

[0006] In a first aspect, the present invention provides a resource configuration method for responding to regional offshore wind power demand, comprising:

[0007] Obtain historical output data of regional offshore wind power and pre-process the historical output data;

[0008] The Time2Graph algorithm is used to extract time-aware feature subsequences from the pre-processed historical output data.

[0009] Based on the graph neural network, the transfer of the characteristic subsequence is mined to analyze the dynamic change characteristics of offshore wind power output in different periods;

[0010] An offshore wind power demand response resource that can match the dynamic change characteristics is selected from a demand response resource matching library.

[0011] In a second aspect, the present invention provides a resource configuration device for responding to regional offshore wind power demand, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described above when executing the computer program.

[0012] In a third aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described above are implemented.

[0013] Compared with the prior art, the present invention has the following beneficial effects:

[0014] The present invention provides a method for extracting regional offshore wind power output fluctuation characteristics based on the Time2Graph algorithm, which realizes accurate analysis of characteristic subsequences during periods of large output fluctuations, thereby being able to match and configure highly adaptable DR resources to promote the effective consumption of regional offshore wind power, improve the stability of the power system, and reduce the impact of new energy output fluctuations on the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart of a method for configuring resources in response to demand for regional offshore wind power provided in Example 1 of the present invention;

[0016] Figure 2 A schematic diagram of the composition of a resource allocation device for responding to regional offshore wind power demand provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0017] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0018] Embodiment 1:

[0019] See also Figure 1 As shown, this embodiment provides a resource configuration method for responding to regional offshore wind power demand, the method comprising:

[0020] 101. Obtain historical output data of regional offshore wind power and pre-process the historical output data.

[0021] 102. The Time2Graph algorithm is used to extract time-aware feature subsequences from the pre-processed historical output data.

[0022] 103. Mining the transfer of the characteristic subsequence based on a graph neural network to analyze the dynamic change characteristics of offshore wind power output in different periods;

[0023] 104. Select an offshore wind power demand response resource that can match the dynamic change characteristics from a demand response resource matching library.

[0024] It can be seen that this embodiment provides a method for extracting regional offshore wind power output fluctuation characteristics based on the Time2Graph algorithm, which realizes accurate analysis of characteristic subsequences during periods of large output fluctuations, thereby being able to match and configure highly adaptable DR resources to promote the effective consumption of regional offshore wind power, improve the stability of the power system, and reduce the impact of new energy output fluctuations on the power system.

[0025] In a specific embodiment, in the above step 101, the historical output data is preprocessed, including:

[0026] The historical output data of regional seawater wind power obtained and input is T = {t 1 ,...,t |T|}, where each t contains n elements arranged in time order, that is, t = {x 1 ,...,x n}, n is the number of recording periods for each output data; the segment s of t is a continuous subsequence, that is, s = {x i ,...,x j}; Suppose t has m segments of equal length l, then t={{x l*k+1 ,...,x l*k+l},0≤k≤m-1}, k represents the kth fragment; two fragments s i and j The distance between them is expressed as d(s i ,s j ), where d(·,·) selects the dynamic time warping distance DTW;

[0027] For a feature subsequence Shapelet v, it represents a time series segment of a certain category, that is, by selecting certain specific feature indicators, the time series set T can be divided into two smaller sets, one with features close to v and the other far away. In this step 101, the offshore wind power data set is classified into two categories by setting relevant threshold indicators, so as to subsequently extract the typical feature subsequences of the offshore wind farm during the period of large output fluctuations. The classification indicators mainly include the curve fluctuation amplitude index and the curve fluctuation frequency index f 0 , the calculation formula is as follows:

[0028]

[0029] Among them, Q max It represents the maximum installed capacity of the offshore wind farm in the region; i is a time point on the output curve, i≤n.

[0030]

[0031] Among them, c is the fluctuation amplitude of the time series curve greater than For f≥f 0 Offshore wind power output curve, marked as 1 otherwise marked as 0.

[0032] In the time series classification task, positive samples and negative samples can be put into different groups. The specific form can be transformed into:

[0033] L=-g(S pos (v,T),S neg (v,T)) (3)

[0034] Here, L is the dissimilarity measure between positive and negative samples for the feature subsequence Shapelet v; S * (v,T) means relative to a specific group T * The function g accepts two finite sets as input and returns a scalar value to indicate how far apart the two sets are, which can be information gain or a measure of dissimilarity on the sets, namely KL divergence.

[0035] The traditional Shapelets method ignores the dynamic changes that may exist in feature subsequences at different times. To this end, in the above step 102, according to the Time2Graph algorithm, two factors are defined to measure the time effect of feature subsequence Shapelets at different levels. Specifically, the local factor w is introduced. n To represent the internal importance of the nth element of a specific Shapelet. The distance between Shapelet v and the time segment s of the offshore wind power output time series curve is calculated as:

[0036]

[0037] Among them, τ(v,s|a * ,w) is the sequence alignment method a * The best alignment is defined by the predefined dissimilarity of the two time series, here DTW.

[0038] At the same time, the global factor u is set to obtain the cross-segment influence weight, and the distance between Shapelet v and the offshore wind power output timing curve segment t is calculated as:

[0039]

[0040] Given a classification task, a supervised learning method is established to select the most important time-aware Shapelets and learn the corresponding time factor w for each Shapelet. n and u. In particular, we set a pool of segments selected from all subsequences as Shapelet candidates and a set of labeled time series T. For each candidate v, we have the following objective function:

[0041]

[0042] λ and ε are hyperparameters of the penalty, and the differentiable function g(·,·) measures the distance between the distributions of two finite sets.

[0043] After learning the time factor from the candidate shapelets, the top K shapelets with the smallest loss are selected as the typical feature subsequences for the period of large output fluctuations of the offshore wind farm. The time scale of the feature subsequence is determined according to the set Shapelet segment length.

[0044] In this way, the above method can achieve accurate analysis of the characteristic subsequences during the period of large output fluctuations.

[0045] The Shapelet evolution graph is a directed weighted graph G = (V, E), where the graph consists of K vertices, each vertex represents a Shapelet, and each directed edge e i,j ∈E and its weight w i,j , which means that in the same time series, Shapelet v i ∈V is followed by another Shapelet v of the same timing curve j ∈V. The path of the graph can naturally reflect the evolution of shapelet and its transition mode. To this end, the above step 103 includes:

[0046] First, each segment s of each time series is divided into i Assign to several i The Shapelet closest to the segment s is assigned a predefined threshold δ to determine whether the two are close or far away. In actual implementation, this can be determined by statistics of the data set. i The Shapelet is represented by v i,* , use v i,j Indicates i The jth allocation of . The allocation probability p i,j The normalized calculation is:

[0047]

[0048] in, At the same time Normalize the edge weight of each node to 1, then each pair of nodes v i and v j The edge weight between them is represented by Shapelet v i Convert to Shapelet v j The conditional probability P(v j |v i ). Finally, the Shapelet evolution graph constructed as above is used to learn the representation of the Shapelet and the given time series, and the conversion between the Shapelets is obtained to mine the maximum possible conversion of the feature subsequence as the dynamic change feature of the period of large fluctuation of offshore wind power output.

[0049] For example, when the length of the feature subsequence is set to 96 points (time scale is 15 minutes), the Shapelet learned from the constructed Shapelet evolution graph can be used to characterize the typical output day during the period of large fluctuations in offshore wind power output, and the conversion between Shapelets can characterize the weight of the typical feature subsequence on each day of the week and the weight of each point element within the feature.

[0050] In this way, by using the Time2Graph algorithm and graph neural network in the above way, it is possible to accurately identify and analyze the maximum possible conversion of characteristic subsequences during periods of large output fluctuations, and through time series analysis, determine the dynamic change characteristics of offshore wind power output in different periods.

[0051] In a specific embodiment, the demand response resource matching library in step 104 includes a demand response resource library and a maximum possible characteristic subsequence library for periods of large fluctuations in offshore wind power output; the demand response resource library stores and records the characteristics of various types of demand response resources, including ramp characteristics, response time, response speed, response reliability, and resource capacity. The main matching steps are as follows:

[0052] (1) Analyze the characteristic subsequences representing different time periods in the maximum possible characteristic subsequence library, including calculating the fluctuation and frequency of the subsequences and performing corresponding time-frequency domain analysis.

[0053] (2) Based on the time-frequency domain analysis results of the characteristic subsequences in different time periods in the maximum possible characteristic subsequence library, a list of demand response resources that can best meet the characteristic subsequence absorption needs is provided.

[0054] Embodiment 2:

[0055] See also Figure 2As shown, the demand response resource configuration device for regional offshore wind power provided in this embodiment includes a processor 21, a memory 22, and a computer program 23 stored in the memory 22 and executable on the processor 21, such as a demand response resource configuration program for regional offshore wind power. When the processor 21 executes the computer program 23, the steps of the above-mentioned embodiment 1 are implemented, such as Figure 1 Steps shown.

[0056] Exemplarily, the computer program 23 may be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 23 in the resource configuration device for responding to demand response for offshore wind power in the region.

[0057] The processor 21 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0058] The memory 22 may be an internal storage unit of the demand response resource configuration device for regional offshore wind power, such as a hard disk or memory of the demand response resource configuration device for regional offshore wind power. The memory 22 may also be an external storage device of the demand response resource configuration device for regional offshore wind power, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the demand response resource configuration device for regional offshore wind power. Further, the memory 22 may also include both an internal storage unit and an external storage device of the demand response resource configuration device for regional offshore wind power. The memory 22 is used to store the computer program and other programs and data required by the demand response resource configuration device for regional offshore wind power. The memory 22 may also be used to temporarily store data that has been output or is to be output.

[0059] Embodiment 3:

[0060] This embodiment provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the method described in Embodiment 1 are implemented.

[0061] The computer-readable medium shown can be any device that can contain, store, communicate, propagate or transmit a program for use with an instruction execution system, device or apparatus or in conjunction with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection portion with one or more wirings (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium can even be paper or other suitable media on which the program can be printed, such as by optically scanning the paper or other medium, then editing, interpreting or processing in other suitable ways as necessary to obtain the program electronically, and then storing it in a computer memory.

[0062] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.

Claims

1. A resource allocation method for responding to regional offshore wind power demand, characterized in that: include: Obtain historical output data of regional offshore wind power and pre-process the historical output data; The Time2Graph algorithm is used to extract time-aware feature subsequences from the pre-processed historical output data. Based on graph neural network, the transfer of the characteristic subsequence is mined to analyze the dynamic change characteristics of offshore wind power output in different periods; Selecting an offshore wind power demand response resource that can match the dynamic change characteristics from a demand response resource matching library; The preprocessing of the historical output data includes: The historical output data of regional seawater wind power obtained and input is T = {t1,...,t |T| }, where each t contains n elements arranged in time order, that is, t = {x1,...,x n }, n is the number of recording periods for each output data; the segment s of t is a continuous subsequence, that is, s = {x i ,...,x j }; Suppose t has m segments of equal length a, then t={{x a*k+1 ,...,x a*k+a },0≤k≤m-1}, k represents the kth fragment; two fragments s i and j The distance between them is expressed as d(s i ,s j ), where d(·,·) selects the dynamic time warping distance DTW; The threshold of the classification index is set to classify the historical output data of seawater wind power into two categories; The classification index includes a curve fluctuation amplitude index And the curve fluctuation frequency index f0, the calculation formula is as follows: Among them, Q max represents the maximum installed capacity of the offshore wind farm in the region; i is a time point on the output curve, i≤n; Among them, c is the fluctuation amplitude of the time series curve greater than The number of times; for the offshore wind power output curve that satisfies f≥f0, it is marked as 1, otherwise it is marked as 0; In the time series classification task, positive samples and negative samples are placed in different groups, which can be formalized as follows: L=-g(S pos (v,T),S neg (v,T)) (3) Where L is the dissimilarity measure between positive and negative samples for the feature subsequence Shapelet v; S * (v,T) means relative to a specific group T * The function g accepts two finite sets as input and returns a scalar value to indicate how far the two sets are, which can be information gain or a measure of dissimilarity on the sets, namely KL divergence; The Time2Graph algorithm is used to extract time-aware feature subsequences from the pre-processed historical output data, including: Introducing the local factor w n To represent the internal importance of the nth element of a specific feature subsequence Shapelet; the distance between Shapeletv and the time series segment s of the offshore wind power output time series curve is calculated as: Among them, τ(v,s|a * ,w) is the sequence alignment method a * The predefined dissimilarity of the next two time series is used to define the best alignment using DTW; At the same time, the global factor u is set to obtain the cross-segment influence weight, and the distance between Shapelet v and the offshore wind power output timing curve segment t is calculated as: To select the most important time-aware Shapelet and learn the corresponding time factor w for each Shapelet n and u; Set up a pool of segments selected from all subsequences as Shapelet candidates, and a set of labeled time series T; For each candidate v, there is the following objective function: λ and ε are hyperparameters of the penalty, and the differentiable function g(·,·) measures the distance between the distributions of two finite sets; After learning the time factor from the candidate shapelets, the top K shapelets with the smallest loss are selected as the typical feature subsequences of offshore wind farm output fluctuations.

2. The resource allocation method for responding to regional offshore wind power demand according to claim 1, characterized in that: The graph neural network is used to mine the transfer of the characteristic subsequences to analyze the dynamic change characteristics of regional offshore wind power output in different periods, including: Each segment s of each time series is divided into i Assign to several i The Shapelet that is closest to the nearest shapelet has a predefined threshold δ to determine whether the two are close or far away; will be assigned to segment s i The Shapelet is represented by v i,* , use v i,j Indicates i The jth allocation of i,j The normalized calculation is: in, At the same time Normalize the edge weight of each node to 1, then each pair of nodes v i and v j The edge weight between them is represented by Shapelet v i Convert to Shapelet v j The conditional probability P(v j |v i ) to construct the Shapelet evolution graph; The constructed Shapelet evolution graph is used to learn the representation of the Shapelet and the given time series, and the conversion between the Shapelets is obtained to mine the maximum possible conversion of the feature subsequence as the dynamic change feature of the offshore wind power output fluctuation.

3. The resource allocation method for responding to regional offshore wind power demand according to claim 1, characterized in that: The demand response resource matching library includes a demand response resource library and a maximum possible characteristic subsequence library of offshore wind power output fluctuations; The demand response resource library stores and records the characteristics of various types of demand response resources, including ramp characteristics, response time, response speed, response reliability and resource capacity.

4. A resource configuration device for responding to regional offshore wind power demand, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

5. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images