Multi-scenario switching wind power sequence simulation method and device based on flexible time boundary

Abstract

The present invention relates to a method and device for simulating a wind power sequence with multiple scene switching based on flexible time boundaries, the method comprising: obtaining multiple original wind power output sequences output by a wind farm; determining the wind power output scenes to which the multiple original wind power output sequences belong based on typical scene output curves corresponding to existing wind power output scenes; establishing a state transition probability matrix for switching between multiple wind power output scenes based on the wind power output scenes to which the multiple original wind power output sequences belong; calculating the state transition probability of multiple original wind power output sequences based on the transition probability matrix for switching between multiple wind power output scenes to obtain a state transition matrix; generating a wind power state sequence of a specified length based on the state transition matrix, performing a wind power sequence simulation operation on the wind power state sequence in combination with a quantization noise model to obtain a wind power output simulation sequence. The technical solution provided by the present application improves the accuracy of scene classification and the efficiency of wind power sequence simulation.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy dispatching and operation of power systems, and in particular to a method and device for simulating wind power sequences with multiple scenarios switching based on flexible time boundaries. Background Art

[0002] At present, the problem of increasingly depleted energy resources has become the focus of global attention. Energy is the foundation for social development. Therefore, solving energy problems is the premise for ensuring the vitality of human social development. Many industrial revolutions in the world have occurred with the renewal of major energy sources as an opportunity. Under this situation, my country must vigorously develop renewable energy as an important strategic measure to ensure energy security and achieve industrial transformation. Clean energy such as wind energy and solar energy is the renewable energy with the greatest potential for large-scale commercial development in the world. Large-scale development and utilization of wind energy, solar energy and hydropower has become one of the effective measures for countries around the world to improve energy structure, solve energy and environmental problems, and ensure the sustainable development of the national economy. In recent years, with the increasing severity of energy crisis and environmental pollution, clean energy generation such as wind power generation and solar power generation has been vigorously developed. However, clean energy such as wind energy and solar energy not only has large fluctuations and obvious seasonal characteristics in time distribution, but also presents the characteristics of unbalanced distribution in spatial distribution. Therefore, the grid-connected power generation of such power sources is uncontrollable and will have an impact on the safe and stable operation of the power grid. When new energy is connected to the grid on a large scale, it will seriously affect the safety, stability and economy of the power system operation, resulting in the frequent occurrence of wind and light abandonment in the power grid.

[0003] It is necessary to assess the grid's ability to accept new energy in advance and plan the operation and dispatch of the power system. The time series simulation of wind power output can provide operating data that matches the actual operating status and provide a reference for the dispatch and planning of the grid. The higher the accuracy and the more closely the simulation matches the actual output process, the greater its contribution to this work.

[0004] In the calculation method of theoretical power, the functional mapping relationship between wind speed and power is very complex and is affected by many factors. The power obtained is difficult to meet the actual power distribution characteristics. The direct modeling method uses historical wind power output data and statistical principles to comprehensively consider the output characteristics, which has high reliability. Commonly used methods for building wind power output models include the Autoregressive Moving Average Model (ARMA), the Copula correlation model, and the Monte Carlo (MC) random simulation model. The ARMA algorithm needs to estimate the parameters of the model to model wind power output. When the regularity of historical data is poor, the error of the result is large. Summary of the invention

[0005] The present application provides a method and device for simulating a wind power sequence with multiple scenarios switched based on flexible time boundaries, so as to at least solve the technical problems of poor classification accuracy and low simulation efficiency in the related art.

[0006] The first embodiment of the present application proposes a multi-scenario switching wind power sequence simulation method based on flexible time boundaries, the method comprising:

[0007] Acquire multiple original wind power output sequences output by the wind farm;

[0008] Based on typical scenario output curves corresponding to existing wind power output scenarios, determining the wind power output scenarios to which the multiple original wind power output sequences respectively belong;

[0009] According to the wind power output scenarios to which the plurality of original wind power output sequences respectively belong, establishing a state transition probability matrix for switching between the plurality of wind power output scenarios;

[0010] Based on the transition probability matrix of switching between the multiple wind power output scenarios, the state transition probabilities of the multiple original wind power output sequences are calculated to obtain a state transition matrix;

[0011] A wind power state sequence of a specified length is generated according to the state transfer matrix, and a wind power sequence simulation operation is performed on the wind power state sequence in combination with a quantization noise model to obtain a wind power output simulation sequence.

[0012] The second embodiment of the present application proposes a wind power sequence simulation device based on flexible time boundary multi-scenario switching, characterized in that the device includes:

[0013] An acquisition module, used for acquiring a plurality of original wind power output sequences output by the wind farm;

[0014] A determination module, configured to determine the wind power output scene to which each of the plurality of original wind power output sequences belongs based on a typical scene output curve corresponding to an existing wind power output scene;

[0015] An establishing module, used for establishing a state transition probability matrix for switching between a plurality of wind power output scenarios according to the wind power output scenarios to which the plurality of original wind power output sequences respectively belong;

[0016] A calculation module, configured to calculate the state transition probabilities of the plurality of original wind power output sequences based on the transition probability matrix of switching between the plurality of wind power output scenarios, and obtain a state transition matrix;

[0017] The simulation module generates a wind power state sequence of a specified length according to the state transfer matrix, and performs a wind power sequence simulation operation on the state transfer matrix in combination with a quantization noise model to obtain a wind power output simulation sequence.

[0018] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0019] The present invention provides a method and device for simulating a wind power sequence with multiple scenarios switching based on flexible time boundaries, the method comprising: obtaining multiple original wind power output sequences output by a wind farm; determining the wind power output scenarios to which the multiple original wind power output sequences belong based on typical scenario output curves corresponding to existing wind power output scenarios; establishing a state transition probability matrix for switching between multiple wind power output scenarios based on the wind power output scenarios to which the multiple original wind power output sequences belong; calculating the state transition probability of the multiple original wind power output sequences based on the transition probability matrix for switching between the multiple wind power output scenarios to obtain a state transition matrix; generating a wind power state sequence of a specified length based on the state transition matrix, performing a wind power sequence simulation operation on the wind power state sequence in combination with a quantization noise model to obtain a wind power output simulation sequence. The technical solution provided by the present invention improves the accuracy of scenario classification and the efficiency of wind power sequence simulation.

[0020] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0022] Figure 1 is a flowchart of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided according to an embodiment of the present application;

[0023] Figure 2 It is a specific flow chart of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided according to an embodiment of the present application;

[0024] Figure 3 It is a schematic diagram of wind power sequence and time window movement in an improved scenario classification algorithm of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided according to an embodiment of the present application;

[0025] Figure 4 It is an overall schematic diagram of an improved scenario classification algorithm of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided according to an embodiment of the present application;

[0026] Figure 5 It is a basic flow chart of an improved scenario classification algorithm of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided in accordance with an embodiment of the present application;

[0027] Figure 6 It is a basic flow chart for constructing a wind power output state transfer matrix according to a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided in one embodiment of the present application;

[0028] Figure 7 It is a comparison diagram of a simulation result scenario classification curve and an original data scenario classification curve of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided according to an embodiment of the present application;

[0029] Figure 8 This is a probability distribution comparison diagram of simulated data and original data obtained according to an improved simulation algorithm of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided in one embodiment of the present application;

[0030] Figure 9a and 9b It is a correlation comparison diagram of data generated by a traditional MCMC algorithm and an improved simulation algorithm of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided in one embodiment of the present application and the original data;

[0031] Fig.10 It is a comparison chart of the average output values ​​at different times of the day of data generated by various simulation methods of the multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided in one embodiment of the present application;

[0032] Fig.11 It is a structural schematic diagram of a wind power sequence simulation device with multi-scenario switching based on flexible time boundaries provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0033] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0034] The present application proposes a method and device for simulating a multi-scenario switching wind power sequence based on flexible time boundaries, the method comprising: obtaining multiple original wind power output sequences output by a wind farm; determining the wind power output scenarios to which the multiple original wind power output sequences belong based on typical scenario output curves corresponding to existing wind power output scenarios; establishing a state transition probability matrix for switching between multiple wind power output scenarios based on the wind power output scenarios to which the multiple original wind power output sequences belong; calculating the state transition probabilities of the multiple original wind power output sequences based on the transition probability matrix for switching between the multiple wind power output scenarios to obtain a state transition matrix; generating a wind power state sequence of a specified length based on the state transition matrix, performing a wind power sequence simulation operation on the wind power state sequence in combination with a quantization noise model to obtain a wind power output simulation sequence. The technical solution provided by the present invention solves the technical problems of poor classification accuracy and low simulation efficiency of existing methods, realizes flexible scene classification through small segment splicing and recognition, and provides a wind power simulation algorithm that allows wind power scenes to have flexible time boundaries. Scene switching can be performed at multiple time points within a day, and the duration is not limited to integer days, which is very close to the actual physical process, thereby overcoming the shortcomings of existing algorithms and significantly improving the effect of wind power simulation. The noise principle in signal processing is used to process the uneven output data of wind power. In the absence of other data such as meteorological factors, wind turbine parameters, etc., a fast and accurate simulation of the wind power time series output process is achieved. At the same time, the output characteristics of wind power historical data are fully explored, and an effective wind power output model is established through effective classification and modeling, which can provide support for wind power operation scheduling planning.

[0035] Example 1

[0036] Figure 1 A flowchart of a multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided in an embodiment of the present disclosure is shown in FIG. Figure 1 As shown, the method includes:

[0037] Step 1: Obtain multiple original wind power output sequences output by the wind farm.

[0038] Step 2: Based on typical scenario output curves corresponding to existing wind power output scenarios, determine the wind power output scenarios to which the multiple original wind power output sequences each belong.

[0039] In an embodiment of the present disclosure, an improved wind power scene classification method is used to extract a typical scene output curve, and the typical scene output curve corresponding to the existing wind power output scene is used to determine the wind power output scene to which each of the multiple original wind power output sequences belongs, including:

[0040] A1: normalizing the multiple original wind power output sequences sampled at equal intervals output by the wind farm;

[0041] In this embodiment, the wind power output sequence of a wind farm is used as a sample for analysis. The sampling interval of the wind power output sequence is 15 minutes, and the data is normalized:

[0042]

[0043] Among them, P i is the input sampling power, P M is the rated capacity of the wind farm, P o It is the normalized output power. In other embodiments of the present invention, there is no restriction on the sampling interval of the wind power output sequence.

[0044] like Figure 4 As shown, the wind power sequence of wind farm output with equal interval sampling is normalized, and the number of sampling values ​​contained in each day is N D , that is, the sequence recording time l D for l D =L / N D days, and L is the total length of the wind power sampling sequence.

[0045] A2: Classify the scenes using the original wind power output sequence with the minimum length as a unit, and use the wind power output scene corresponding to the typical scene output curve with the highest similarity among the original wind power output sequences as the wind power output scene to which each of the multiple original wind power output sequences belongs;

[0046] Furthermore, the original wind power output sequence is classified into the wind power output scenario to which the typical scenario output curve of the wind power output scenario is closest, until the wind power output scenario identification of the entire wind power output sequence is completed.

[0047] It should be noted that considering that weather changes are a slow process, in order to facilitate modeling and analysis, the following provisions are made for the switching of wind power output scenarios:

[0048] a) Wind power output scenario switching only occurs at certain times every day;

[0049] b) The duration of each wind power output scenario is not less than a certain set value;

[0050] In addition, the specific time at which the wind power output scene switching occurs and the set value that the wind power output scene duration is not less than are set according to actual conditions.

[0051] In the embodiment of the present disclosure, before determining the wind power output scene to which each of the plurality of original wind power output sequences belongs based on the typical scene output curve corresponding to the existing wind power output scene, the method further includes:

[0052] B1: Set the number of wind power scenarios K S .

[0053] B2: Using a forward moving time window, clustering is performed in each time window to obtain a typical scenario output curve segment corresponding to each wind power output scenario;

[0054] B3: splicing the typical scenario output curve fragments corresponding to each wind power output scenario to obtain the typical scenario output curve corresponding to each wind power output scenario.

[0055] Based on the fact that the chaotic dynamic mechanism followed by wind power sequence has the characteristics of short-term predictability, a forward moving time window is used, and adjacent time windows overlap. Clustering is performed in each time window to obtain typical curve fragments of each scene. These fragments are superimposed and spliced ​​to generate a typical daily curve. Figure 3 As shown, due to the window length N m It is greater than the shortest sequence length N1 of wind power, so when the time window moves, there will be overlapping parts. The output daily curves of these overlapping parts are most importantly obtained by taking the average value.

[0056] In this embodiment, each day is evenly divided into M segments, and the length of each segment N1 is recorded as N D The number of sample values ​​contained in each day, the daily sequence is numbered from 1, and the wind power scene switching is only required to occur at position i, where i = mN1, m = 1, 2, ..., M, and the shortest length N of each scene sequence is required to be N = K m N1, K m The shortest number of sequence segments contained in each scene sequence, and K m Make your choice based on experience and experimentation.

[0057] like Figures 3 to 5 As shown, in the embodiment of the present disclosure, a forward moving time window is adopted, and clustering is performed in each time window to obtain a typical scene output curve segment corresponding to each wind power output scene, including:

[0058] F1: Select the time window length as N m , the forward moving length of the adjacent time window is N1, then the calculation formula for the data fragment corresponding to the i-th time window on the k-th day is as follows:

[0059]

[0060] Where N m >N1,k=1,2,…,l D , l D is the recording duration of the sequence, and the l D =L / N D days, L is the total length of the wind power sampling sequence, N D is the number of sampling values ​​contained in each day, all l DThe data segments at the same time in the same day constitute the clustered sample set. In addition, considering that some signal segments in the time window have scene switching, that is, these data samples do not belong to a specific scene category, the clustering process allows a certain proportion of flying points. In this embodiment, the clustering process has, for example, 10% flying points, and the flying point ratio is not limited in other embodiments of the present invention.

[0061] F2: Let i = 1, use K-mean algorithm to cluster the corresponding sample set, and the cluster subscript set is denoted by G i (α), α=1,…,K S , calculate the cluster center η i (α);

[0062] The K-mean algorithm is used to calculate the cluster center η i The method of (α) is expressed as:

[0063]

[0064] Among them, |G i (α)| is the set G i The number of elements of (α).

[0065] F3: Let i = i + 1, and initialize the current clustering calculation based on the clustering result of the previous time window to obtain η′ i (α), output cluster index set and cluster center, repeat until i = (N D -N1) / ΔN+1, repeat until i=(N D -N1) / ΔN+1, that is, the time span reaches 1 full day;

[0066] Among them, the current clustering calculation is initialized based on the clustering result of the previous time window to obtain η′ i The expression of (α) is:

[0067]

[0068] In the embodiment of the present disclosure, the typical scene output curve fragments corresponding to each wind power output scene are spliced ​​to obtain the typical scene output curve corresponding to each wind power output scene, including:

[0069] For cluster center η i (α), i=1,…,(N D -N1) / ΔN+1, each element is aligned in time and then superimposed to obtain the average value, that is, a typical scenario output curve corresponding to the wind power output scenario α is generated. The calculation formula of the typical scenario output curve is as follows:

[0070]

[0071] in, It is the clustering curve of the mth time window of wind power output scenario α.

[0072] like Figure 4 and Figure 5 As shown, this method divides the entire wind power output sequence into l D =L / N D Every day is divided into Figure 2 The minimum output sequence shown, l D The time window of the day moves synchronously, and after each movement, l D The output sequence in a time window is clustered by Kmeans to generate the output curve of a typical scenario. The initial clustering center of the next time window is the clustering result of the previous time window. The chaotic idea of ​​data processing is used here, and the short-term predictability of wind power is used to improve the clustering effect and accuracy.

[0073] Step 3: According to the wind power output scenarios to which the multiple original wind power output sequences respectively belong, a state transition probability matrix for switching between the multiple wind power output scenarios is established.

[0074] In the embodiment of the present disclosure, according to the wind power output scenarios to which the multiple original wind power output sequences respectively belong, a state transition probability matrix for switching between multiple wind power output scenarios is established, including:

[0075] According to the shortest length N1 of the scene sequence in the wind power output scene classification, the state of the original wind power sampling sequence is identified to obtain the scene state sequence corresponding to the wind power sample sequence;

[0076] According to the scene state sequence, the number of occurrences of each wind power output scene state and the number of transitions between wind power output scenes are counted, and a state transition probability matrix for switching between multiple wind power output scenes is established.

[0077] It should be noted that, according to the previous agreement, the wind power scene state changes only occur at position i every day, i = mN1, m = 1, 2, ..., M, so the scene state sequence corresponding to the wind power sample sequence can be recorded as {Z m ,m=1,2,…,L / N1}; Considering that each scene is required to have a minimum duration of N=K m N1, the main calculation steps for scene state recognition of wind power sample sequence are as follows:

[0078] G1: Initialization, set s=0, m=1, p=0.

[0079] G2: If s=0, select the sequence fragment The scene state of the signal segment is determined as:

[0080]

[0081] In the formula The length of the typical daily curve for the β scenario is N m fragment, so z m =β, m=m+1, s=s+1.

[0082] G3: Because the shortest length of the output scene is N = K m N1, repeat step G2 until s = K m , thus a power output scenario is identified, p = 1; if there is a sequence in the middle that does not belong to the power output scenario β, s = 0 and re-identification is performed.

[0083] G4: If p = 1, we need to determine whether the next sequence segment belongs to scene β, select segment ξ(m+1), and calculate

[0084]

[0085] If α=β, it means that the next segment also belongs to scene β, and p=1; if α≠β, it means that the next segment does not belong to scene β, and p=0, s=0, and the next scene recognition is performed again.

[0086] G5: Repeat the above steps until the full-time scene recognition is completed and the scene state sequence corresponding to the wind power sample sequence is obtained {z m ,m=1,2,…,L / N1}.

[0087] It should be noted that the random changes in wind power output scenarios are characterized by a Markov chain probability model, which is represented by a state transition probability matrix. The state transition probability matrix is ​​expressed as follows:

[0088]

[0089] Where p(α,β)=P(z m+1 =β|z m =α), that is, the conditional probability of the wind power output scenario state transferring from α to β.

[0090] For the scene state sequence {z m} for statistical analysis, the calculation formula of p(α,β) is as follows:

[0091]

[0092] where n α,β is the event {z m =α,z m+1 =β} occurs, The number of scene conversions that do not meet the minimum wind power output scene length.

[0093] Step 4: Based on the transition probability matrix of switching between the multiple wind power output scenarios, the state transition probabilities of the multiple original wind power output sequences are calculated to obtain a state transition matrix.

[0094] In the embodiment of the present disclosure, the state transition probabilities of the multiple original wind power output sequences are calculated based on the transition probability matrix of switching between the multiple wind power output scenarios to obtain the state transition matrix, including:

[0095] H1: The original wind power output sequence is divided into non-uniform quantization intervals using quantization coding and quantization noise model;

[0096] like Figure 6 As shown in the figure, firstly, the historical wind power output data is processed, the abnormal output power is deleted, and the output power is standardized; the non-uniform quantization function Q(x) and the wind power extremely low output threshold ε are determined, and the data is non-uniformly quantized to obtain the non-uniform quantization interval and the quantized output value of each interval; the number of conversions between different intervals is counted by the original data, and the wind power state transition probability matrix P at adjacent moments when the scene state remains unchanged is constructed. S (α; t) and the wind power state transition probability matrix P at adjacent moments when the scene state changes T (α,β; t).

[0097] The method of dividing the original wind power output sequence into non-uniform quantization intervals by using quantization coding and quantization noise model specifically includes:

[0098] D1: Normalize the original wind power output sequence, denote the initialization wind power state number as K, and divide the wind power output interval into [0, q1, q2, ..., q K ], interval [q n-1 ,q n ] has an interval quantization value Q n (x), x is the wind power output sequence value, and the functional relationship x→Q(x) between the wind power output sequence value and all interval quantization values ​​is defined as the quantization function;

[0099] D2: The difference between each wind power sampling value and the corresponding quantized value Q(x) is defined as quantization noise, that is, e=xQ(x);

[0100] D3: Define the wind power sequence state whose normalized value is in this interval as 0, set the small interval to [0,ε], q1=ε<<1, q K =1;

[0101] It should be noted that, considering that there are many segments of wind power sequence that are continuously 0, that is, when the wind speed is less than a certain value, the wind turbine does not generate electricity. Therefore, a small interval [0, ε] is set, and the state of the wind power sequence with normalized values ​​in this interval is set to 0, that is, let q1 = ε << 1, qK =1.

[0102] D4: The optimal non-uniform quantization encoding is performed on the power value of the wind turbine outside the set small interval [0, ε] to obtain Q(x), requiring the overall mean square value of the quantization error to be minimized. The overall mean square value of the quantization error is expressed as follows:

[0103]

[0104] In the formula, x i is the wind power output sequence value, and the interval quantization value of K state intervals is expressed as ξ k =Q k (x), the state value is defined as S(x) = k, k = 1, 2, ..., K, and the nth interval [q n-1 ,q n ] is q n-1 =ξ n -d n ,q n =ξ n +d n , d n is the width of the nth interval.

[0105] H2: Count the number of transitions between different non-uniform quantization intervals in all segments of multiple original wind power output sequences, calculate the state transition probability between different non-uniform quantization intervals, and obtain the state transition matrix.

[0106] A cyclic stationary Markov chain model is established for the wind power state sequence under multi-scenario switching conditions, which is represented by the wind power state transition probability matrix under the same scenario and the wind power state transition probability matrix when the wind power scene is switched. S (α; t) is the wind power state transition probability matrix from time t to time t+1 when the wind power scene state α remains unchanged, that is, the wind power state transition probability matrix at adjacent moments when the scene state remains unchanged, specifically expressed as:

[0107]

[0108] where p α；t (i,j) represents the probability of transitioning from state i at time t to state j at time t+1;

[0109] Note: T (α, β; t) is the wind power state transition probability matrix corresponding to the wind power scenario switching from α to β from time t to t+1, that is, the wind power state transition probability matrix at adjacent moments when the scenario state changes, which can be specifically expressed as:

[0110]

[0111] where pα,β；t (i,j) represents the probability of transitioning from state i at time t to state j at time t+1 when the wind power scenario switches from α to β.

[0112] Specifically, for the wind power sequence {x i}The corresponding wind power state sequence {s i} and wind power scenario state sequence {z m} to conduct statistical analysis and establish the wind power state transition probability matrix P from time t to time t+1 when the wind power scenario state α remains unchanged S (α; t) and the corresponding wind power state transition probability matrix P when the wind power scenario switches from α to β from t to t+1 T (α,β; t);

[0113] The probability p of transitioning from state i at time t to state j at time t+1 is α；t (i,j), that is, the matrix P S (α; t) element p α；t The calculation formula for (i,j) is as follows:

[0114]

[0115] Where t = (m-1)N1+n is the time of the day, and the range of m and n is m = 1, 2, ..., M and n = 1, ..., N1-1; n α；t (i,j) is the number of times the wind power state changes from i to j in the α wind power scenario from time t to time t+1 within the day.

[0116] When the wind power scenario switches from α to β, the probability p of transitioning from state i at time t to state j at time t+1 is α,β；t (i,j) calculation formula, that is, matrix P T The element P of (α, β; t) α,β；t (i,j) is as follows:

[0117]

[0118] Where t = mN1 is the time of the day, m = 1, 2, ..., M; n α,β；t (i,j) is the number of times the wind power scenario switches from α to β and the wind power state transfers from i to j from time t to t+1 within the day.

[0119] Calculate the state transition probabilities between different non-uniform quantization intervals and obtain the state transfer matrix.

[0120] The expression of the state transfer matrix P is as follows:

[0121]

[0122] Step 5: Generate a wind power state sequence of a specified length according to the state transfer matrix, and perform a wind power sequence simulation operation on the wind power state sequence in combination with a quantization noise model to obtain a wind power output simulation sequence.

[0123] In the embodiment of the present disclosure, the wind power state sequence of a specified length is generated according to the state transfer matrix, and a wind power sequence simulation operation is performed on the wind power state sequence in combination with a quantization noise model to obtain a wind power output simulation sequence, including:

[0124] M1: Generate a scene state sequence of specified length using the state transfer matrix;

[0125] The state transition matrix is ​​used to generate a scene state sequence of a specified length, including:

[0126] Calculate the cumulative transition probability matrix Q corresponding to the state transfer matrix, record:

[0127]

[0128]

[0129] Where P is the state transfer matrix;

[0130] Then the elements of the cumulative transition probability matrix Q are calculated as:

[0131]

[0132] According to the cumulative transition probability matrix Q, the wind power scenario state sequence simulation is performed to generate a scenario state sequence of specified length.

[0133] It should be noted that the above calculation process also applies to the matrix P W , P S (α; t) and P T (α, β; t). M2: Generate a wind power state sequence of specified length using the state transfer matrix in each wind power output scenario

[0134] Among them, the state transfer matrix in each wind power output scenario is used to generate a wind power state sequence of a specified length, including:

[0135] According to the scenario state sequence {zm} and the cumulative transition probability matrix Q, the wind power state sequence simulation is performed to obtain a wind power state sequence of a specified length.

[0136] Among them, according to the cumulative transfer probability matrix of wind power output scenario Q = [q(i,j)] 1≤i,j≤K , simulate the wind power state sequence, and obtain the wind power state sequence of specified length including:

[0137] a) Initialization. Generate a random integer α∈{1,2,…,K S}, let z0=α, m=1, s=0.

[0138] b) Iterative process. Generate uniformly distributed random numbers v∈[0,1] and record z m-1 =α, determine β so that q α,β ≤v α,β+1 .

[0139] For s = 0, let z m+k =β,k=0,…,K m -1, and update m=m+K m , s = 1;

[0140] For s = 1, generate a uniformly distributed random number v'∈[0,1], if q β,β ≤v' β,β+1 , that is, z m =β, then set m=m+1 and keep s=1; otherwise update s=0.

[0141] c) Repeat the previous step until a sequence of the required length is generated.

[0142] Among them, based on the simulated wind power scenario state sequence {z m}, according to the cumulative transition probability matrix Q of wind power state under multi-scenario switching conditions S (α; t) = [q α；t (i,j)] 0≤i,j≤K-1 and Q T (α,β;t)=[q α,β；t (i,j)] 0≤i,j≤K-1 , realize the wind power state sequence simulation, the specific steps are as follows:

[0143] Ⅰ) Initialization. Generate a random integer l∈{1,2,…,K}, let s0=0, t=1.

[0144] II) Iteration process. Determine τ from the current time t d =tmodN d , m = [τ d / N1],τ1=τ d modN1, and read z m-1 =α,z m =β,s t-1 =i, generate uniformly distributed random numbers v∈[0,1].

[0145] For τ1 = 0, determine j such that Order t =j, update t=t+1;

[0146] ​​For τ1>0, determine j such that Order t =j, update t=t+1.

[0147] Repeat the above-mentioned steps of implementing the wind power state sequence simulation and matching the wind power output state sequence using the scenario state sequence {zm} and the cumulative transition probability matrix Q until a wind power output state sequence of the required length is generated.

[0148] M3: Map the wind power state sequence to a discrete quantization sequence ξ of each state, and add quantization noise to the discrete quantization sequence ξ according to the quantization noise probability distribution function corresponding to the quantization noise model to obtain a wind power output simulation sequence.

[0149] Furthermore, a probability model of quantization noise with respect to quantized output values ​​is established, specifically:

[0150] N1: For the code value ξ1=0, e=0;

[0151] N2: For the encoding value ξ i ,i=2,…,K-1, the quantization noise satisfies uniform distribution, and its probability distribution function is:

[0152]

[0153] Where e is the quantization noise;

[0154] N3: For ξ K , the quantization noise generally does not satisfy the uniform distribution and needs statistical analysis to obtain

[0155]

[0156] like Figure 7 As shown in the figure, the solid line and the dotted line represent the typical output curves of the original data and the simulated data obtained by the improved method, respectively. It can be seen that the simulated data obtained by the wind power sequence simulation method based on the wind power sequence probability model in multiple wind power scenarios with flexible time boundaries are roughly the same as the original data after scene classification, indicating the accuracy of the simulation effect of this method; Figure 8 As shown, the output data obtained by the wind power sequence simulation method of the wind power sequence probability model under multiple wind power scenarios with flexible time boundaries is roughly consistent with the probability density distribution of the original data.

[0157] Under the single condition of only having historical wind power output data, this application can first classify the historical output days, and improve the simulation accuracy by separately modeling different output days. At the same time, the shortest wind power scene is identified by using wind power output sequence fragmentation and moving time windows, making the wind power scene length more flexible and the classification accuracy higher. Then, on the basis of scene clustering, the imbalance of wind power output is further processed by using the signal noise principle, and the output sequence is simulated using the non-uniform quantization MCMC algorithm. The Markov state transfer matrix of the cycle is established in combination with the correlation in wind power output time, which can further improve the simulation accuracy and make the simulation process more in line with the actual situation.

[0158] like Figure 9a and 9b As shown in the figure, the temporal autocorrelation of wind power output is periodic and roughly cycles in days. The original MCMC simulation cannot simulate this periodic autocorrelation characteristic. The simulation results obtained by the wind power sequence simulation method based on the wind power sequence probability model in multiple wind power scenarios with flexible time boundaries have a high degree of correlation fit with the original data.

[0159] See also Fig.10 , SS, CS, CS2, KMSS, KMCS and Measured in the figure represent the first-order traditional MCMC method, the cyclostationary first-order MCMC method, the cyclostationary second-order MCMC method, the first-order traditional MCMC method after KM scenario classification, the cyclostationary first-order MCMC method after KM scenario classification and the original data of wind power output, respectively. This figure obtains the mean output at the same time on all output days. It can be seen that the MCMC algorithm without scenario classification can only fit the original data well at the second order, but it requires a larger computing space. The multi-scenario MCMC method after KM clustering has higher computing efficiency while having similar simulation effects.

[0160] Table 1 shows the distance between the historical data of wind power in this application and the scene classification curves of the data obtained by simulating using the first-order traditional MCMC method, the cyclostable first-order MCMC method, the cyclostable second-order MCMC method, the first-order traditional MCMC method after KM scene classification, and the cyclostable first-order MCMC method after KM scene classification. It can be seen that the distance between the data obtained by the method of this application and the original wind power data is the smallest, and the simulation effect is closer to the actual situation.

[0161]

[0162] In addition, under the condition that only historical wind power output sequence data exists, this application can improve the modeling accuracy by fully extracting data features; on the other hand, it can provide the operation data that conforms to the actual situation for the operation and dispatch of the power grid by constructing a wind power generation simulation model, thereby contributing to the stable and safe operation of the power system. First, the wind power output is fragmented, and the shortest fragment splicing is achieved through time window movement, so as to construct a wind power scenario with flexible length, so that the classification is more in line with the actual situation. By analyzing the output data of different scenarios, different data models are established to minimize the modeling error; combining the autocorrelation of wind power output and the imbalance of wind power output, a periodic Markov transition matrix is ​​established, and the wind power output sequence is reconstructed using non-uniform quantization and noise principles, further improving the simulation efficiency, so that the data generated by the established wind power output model is closer to the historical data. This application improves the modeling accuracy and can provide support for the operation and dispatch of wind power, thereby providing a reference for improving the capacity of new energy consumption.

[0163] The effect of the output day scene division will directly affect the quality of the wind power model establishment. This application fragments the wind power output sequence to obtain wind power output scenes of flexible lengths. At the same time, using a moving time window to process the wind power sequence fragments one by one can improve the clustering effect and clustering speed, thereby improving the speed and accuracy of the entire classification process.

[0164] In summary, the multi-scenario switching wind power sequence simulation method based on flexible time boundaries provided by the embodiment of the present disclosure includes: obtaining multiple original wind power output sequences output by the wind farm; determining the wind power output scenarios to which the multiple original wind power output sequences belong based on the typical scenario output curves corresponding to the existing wind power output scenarios; establishing a state transition probability matrix for switching between multiple wind power output scenarios based on the wind power output scenarios to which the multiple original wind power output sequences belong; calculating the state transition probabilities of the multiple original wind power output sequences based on the transition probability matrix for switching between the multiple wind power output scenarios to obtain a state transition matrix; generating a wind power state sequence of a specified length based on the state transition matrix, performing a wind power sequence simulation operation on the wind power state sequence in combination with a quantization noise model to obtain a wind power output simulation sequence. The technical solution provided by the present invention improves the accuracy of scene classification and the efficiency of wind power sequence simulation.

[0165] Example 2

[0166] Fig.11 A schematic diagram of the structure of a multi-scenario switching wind power sequence simulation device based on flexible time boundaries provided in an embodiment of the present disclosure is shown in FIG. Fig.11 As shown, the device comprises:

[0167] An acquisition module 100 is used to acquire a plurality of original wind power output sequences output by a wind farm;

[0168] A determination module 200, configured to determine the wind power output scene to which each of the plurality of original wind power output sequences belongs based on a typical scene output curve corresponding to an existing wind power output scene;

[0169] An establishing module 300 is used to establish a state transition probability matrix for switching between multiple wind power output scenarios according to the wind power output scenarios to which the multiple original wind power output sequences belong respectively;

[0170] A calculation module 400 is used to calculate the state transition probabilities of the multiple original wind power output sequences based on the transition probability matrix of switching between the multiple wind power output scenarios to obtain a state transition matrix;

[0171] The simulation module 500 generates a wind power state sequence of a specified length according to the state transfer matrix, and performs a wind power sequence simulation operation on the state transfer matrix in combination with a quantization noise model to obtain a wind power output simulation sequence.

[0172] In summary, the wind power sequence simulation device based on multi-scenario switching with flexible time boundaries proposed in this application includes: an acquisition module 100, a determination module 200, an establishment module 300, a calculation module 400 and a simulation module 500. This device improves the accuracy of scene classification and enhances the efficiency of wind power sequence simulation.

[0173] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0174] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0175] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A multi-scenario switching wind power sequence simulation method based on flexible time boundaries, characterized in that: The method comprises: Acquire multiple original wind power output sequences output by the wind farm; Set the number of wind power scenarios K S , using a forward moving time window, clustering in each time window to obtain typical scenario output curve fragments corresponding to each wind power output scenario, splicing the typical scenario output curve fragments corresponding to each wind power output scenario to obtain the typical scenario output curve corresponding to each wind power output scenario, and based on the typical scenario output curve corresponding to each wind power output scenario, determining the wind power output scenario to which each of the multiple original wind power output sequences belongs; According to the wind power output scenarios to which the plurality of original wind power output sequences respectively belong, establishing a state transition probability matrix for switching between the plurality of wind power output scenarios; Based on the transition probability matrix of switching between the multiple wind power output scenarios, the state transition probabilities of the multiple original wind power output sequences are calculated to obtain a state transition matrix; Generate a wind power state sequence of a specified length according to the state transfer matrix, perform a wind power sequence simulation operation on the wind power state sequence in combination with a quantization noise model, and obtain a wind power output simulation sequence; The forward moving time window is used to cluster in each time window to obtain a typical scene output curve fragment corresponding to each wind power output scene, including: Select the time window length as N m , the forward moving length of the adjacent time window is N1, then the calculation formula for the data fragment corresponding to the i-th time window on the k-th day is as follows: Where N m >N1,k=1,2,…,l D , l D is the recording duration of the sequence, and the l D =L / N D days, L is the total length of the wind power sampling sequence, N D is the number of sampling values ​​contained in each day, all l D The data segments at the same time of the day constitute the clustering sample set; Let i = 1, use K-mean algorithm to cluster the corresponding sample set, and the cluster subscript set is denoted by G i (α), α=1,…,K S , calculate the cluster center η i (α); Let i = i + 1, and initialize the current clustering calculation based on the clustering result of the previous time window to obtain η′ i (α), output cluster index set and cluster center, repeat until i = (N D -N1) / ΔN+1; The typical scene output curve fragments corresponding to each wind power output scene are spliced ​​to obtain the typical scene output curve corresponding to each wind power output scene, including: For cluster center η i (α), i=1,…,(N D -N1) / ΔN+1, each element is aligned in time and then superimposed to obtain the average value to generate the typical scenario output curve corresponding to the wind power output scenario α. The calculation formula of the typical scenario output curve is as follows: in, It is the clustering curve of the mth time window of wind power output scenario α.

2. The method according to claim 1, characterized in that Determining the wind power output scene to which each of the plurality of original wind power output sequences belongs based on the typical scene output curves corresponding to the respective wind power output scenes includes: Normalizing the plurality of original wind power output sequences sampled at equal intervals output by the wind farm; The scene classification is performed in units of original wind power output sequences of minimum length, and the wind power output scene corresponding to the typical scene output curve with the highest similarity among the original wind power output sequences is taken as the wind power output scene to which each of the multiple original wind power output sequences belongs.

3. The method according to claim 1, characterized in that According to the wind power output scenarios to which the multiple original wind power output sequences belong, a state transition probability matrix for switching between the multiple wind power output scenarios is established, including: According to the shortest length of the scene sequence in the wind power output scene classification, the state of the wind power original sampling sequence is identified to obtain the scene state sequence corresponding to the wind power sample sequence; According to the scene state sequence, the number of occurrences of each wind power output scene state and the number of transitions between wind power output scenes are counted, and a state transition probability matrix for switching between multiple wind power output scenes is established.

4. The method according to claim 1, characterized in that: Based on the transition probability matrix of switching between the multiple wind power output scenarios, the state transition probabilities of the multiple original wind power output sequences are calculated to obtain a state transition matrix, including: The original wind power output sequence is divided into non-uniform quantization intervals using quantization coding and quantization noise model; The number of transitions between different non-uniform quantization intervals in all segments of multiple original wind power output sequences is counted, the state transition probability between different non-uniform quantization intervals is calculated, and the state transition matrix is ​​obtained.

5. The method according to claim 4, characterized in that The original wind power output sequence is divided into non-uniform quantization intervals using quantization coding and quantization noise model, including: The original wind power output sequence is normalized, the number of initialized wind power states is denoted as K, and the wind power output interval is divided into [0, q1, q2, ..., q K ], interval [q n-1 ,q n ] has an interval quantization value Q n (x), x is the wind power output sequence value, and the functional relationship x→Q(x) between the wind power output sequence value and all interval quantization values ​​is defined as the quantization function; The difference between each wind power sampling value and the corresponding quantized value Q(x) is defined as quantization noise, that is, e=xQ(x); Define the wind power sequence state whose normalized value is in this interval as 0, set the small interval as [0,ε], q1=ε<<1, q K =1; The optimal non-uniform quantization encoding is performed on the power value of the wind turbine outside the set small interval [0, ε] to obtain Q(x), requiring the overall mean square value of the quantization error to be minimized. The overall mean square value of the quantization error is expressed as follows: In the formula, x i is the wind power output sequence value, and the interval quantization value of K state intervals is expressed as ξ k =Q k (x), the state value is defined as S(x) = k, k = 1, 2, ..., K, and the nth interval [q n-1 ,q n ] is q n-1 =ξ n -d n ,q n =ξ n +d n , d n is the width of the nth interval.

6. The method according to claim 1, characterized in that Generate a wind power state sequence of a specified length according to the state transfer matrix, perform a wind power sequence simulation operation on the wind power state sequence in combination with a quantization noise model, and obtain a wind power output simulation sequence, including: Generate a scene state sequence of specified length using the state transfer matrix; Generate a wind power state sequence of specified length using the state transfer matrix in each wind power output scenario; The wind power state sequence is mapped into a discrete quantization sequence ξ of each state, and quantization noise is added to the discrete quantization sequence ξ according to the quantization noise probability distribution function corresponding to the quantization noise model to obtain the wind power output simulation sequence.

7. The method according to claim 6, characterized in that Use the state transition matrix to generate a scene state sequence of specified length, including: Calculate the cumulative transition probability matrix Q corresponding to the state transfer matrix, record: Where P is the state transfer matrix; Then the elements of the cumulative transition probability matrix Q are calculated as: According to the cumulative transition probability matrix Q, a wind power scenario state sequence simulation is performed to generate a scenario state sequence of a specified length; The state transfer matrix in each wind power output scenario is used to generate a wind power state sequence of a specified length, including: According to the scenario state sequence {zm} and the cumulative transition probability matrix Q, the wind power state sequence simulation is performed to obtain a wind power state sequence of a specified length.

8. A multi-scenario switching wind power sequence simulation device based on flexible time boundaries, characterized in that: The device comprises: An acquisition module, used for acquiring a plurality of original wind power output sequences output by the wind farm; Determine module, used to set the number of wind power scenarios K S , using a forward moving time window, clustering in each time window to obtain typical scenario output curve fragments corresponding to each wind power output scenario, splicing the typical scenario output curve fragments corresponding to each wind power output scenario to obtain the typical scenario output curve corresponding to each wind power output scenario, and based on the typical scenario output curve corresponding to each wind power output scenario, determining the wind power output scenario to which each of the multiple original wind power output sequences belongs; An establishing module, used for establishing a state transition probability matrix for switching between a plurality of wind power output scenarios according to the wind power output scenarios to which the plurality of original wind power output sequences respectively belong; A calculation module, configured to calculate the state transition probabilities of the plurality of original wind power output sequences based on the transition probability matrix of switching between the plurality of wind power output scenarios, and obtain a state transition matrix; A simulation module, which generates a wind power state sequence of a specified length according to the state transfer matrix, and performs a wind power sequence simulation operation on the wind power state sequence in combination with a quantization noise model to obtain a wind power output simulation sequence; The forward moving time window is used to cluster in each time window to obtain a typical scene output curve fragment corresponding to each wind power output scene, including: Select the time window length as N m , the forward moving length of the adjacent time window is N1, then the calculation formula for the data fragment corresponding to the i-th time window on the k-th day is as follows: Where N m >N1,k=1,2,…,l D , l D is the recording duration of the sequence, and the l D =L / N D days, L is the total length of the wind power sampling sequence, N D is the number of sampling values ​​contained in each day, all l D The data segments at the same time of the day constitute the clustering sample set; Let i = 1, use K-mean algorithm to cluster the corresponding sample set, and the cluster subscript set is denoted by G i (α), α=1,…,K S , calculate the cluster center η i (α); Let i = i + 1, and initialize the current clustering calculation based on the clustering result of the previous time window to obtain η′ i (α), output cluster index set and cluster center, repeat until i = (N D -N1) / ΔN+1; The typical scene output curve fragments corresponding to each wind power output scene are spliced ​​to obtain the typical scene output curve corresponding to each wind power output scene, including: For cluster center η i (α), i=1,…,(N D -N1) / ΔN+1, each element is aligned in time and then superimposed to obtain the average value to generate the typical scenario output curve corresponding to the wind power output scenario α. The calculation formula of the typical scenario output curve is as follows: in, It is the clustering curve of the mth time window of wind power output scenario α.

Images