A machine learning-based method and system for wind speed prediction and correction in wind farms

Abstract

This application relates to the field of wind farm technology, and in particular to a wind speed prediction and correction method and system based on machine learning. It includes: generating multiple disturbance indices and wind speed-power correlation curves based on historical monitoring data; setting multiple wind speed segments based on the wind speed-power correlation curves, and constructing a wind speed correction model based on all disturbance indices and all wind speed segments; generating an initial wind speed curve based on the wind speed prediction model, and setting a wind speed correction strategy based on the wind speed correction model and the initial wind speed curve; constructing multiple wind speed segments based on the nonlinear relationship between wind speed and power, and judging the influence of each disturbance indices in different wind speed segments based on machine learning technology, thereby constructing a wind speed correction model corresponding to each wind speed segment, improving the correction efficiency of the predicted wind speed curve, and thus ensuring the operational efficiency of the wind farm.

Description

Technical Field

[0001] This application relates to the field of wind farm technology, and in particular to a wind speed prediction and correction method and system based on machine learning. Background Technology

[0002] Wind speed forecasting is a crucial aspect of wind farm power prediction and operation scheduling. Accurate short-term (minute to hourly) and ultra-short-term (second to minute) wind speed forecasts can significantly improve the stability of wind power grid connection, reduce wind curtailment, and lower the mechanical load on wind turbines.

[0003] However, current wind speed prediction methods for wind farms often rely on a single model and do not make full use of multi-source data (such as SCADA data, radar data, satellite data, etc.) for error correction, resulting in insufficient prediction robustness. Summary of the Invention

[0004] The purpose of this application is to provide a wind speed prediction correction method and system based on machine learning to solve the above-mentioned technical problems, aiming to improve the accuracy of wind speed prediction and the operating efficiency of wind farms.

[0005] In some embodiments of this application, a wind speed prediction and correction method for wind farms based on machine learning is provided, including:

[0006] Multiple disturbance indicators and wind speed-power correlation curves are generated based on historical monitoring data;

[0007] Multiple wind speed ranges are set based on the wind speed-power correlation curve, and a wind speed correction model is constructed based on all disturbance indices and all wind speed ranges.

[0008] An initial wind speed curve is generated based on the wind speed prediction model, and a wind speed correction strategy is set based on the wind speed correction model and the initial wind speed curve.

[0009] When setting multiple wind speed zones, these include:

[0010] Establish a sequence of wind speed segments A, A = (a1, a2, ..., a3) i …a n ), where a i Let be the i-th wind speed segment; n is the number of wind speed segments.

[0011] In some embodiments of this application, the construction of the wind speed correction model includes:

[0012] Based on the wind speed segment sequence A, a is set sequentially. i For the target wind speed range;

[0013] Training data packets for the target wind speed zone are generated based on historical monitoring data;

[0014] Primary disturbance indicators for target wind speed zones are selected based on all training data packets.

[0015] Establish a first-order disturbance index sequence B for the target wind speed range, B = (b1, b2, ..., bb2) i …b m ), where b i is the i-th primary disturbance index for the i-th target wind speed segment; m is the number of primary disturbance indices for the target wind speed segment;

[0016] Set the influence factors for each disturbance index;

[0017] A modified sub-model for the target wind speed section is constructed based on the first-level disturbance index number B;

[0018] Corrected sub-models for each wind speed range are established sequentially;

[0019] Establish a modified sub-model sequence P, P = (p1, p2, ..., p) i …p n ), where pi is the modified sub-model for the i-th wind speed segment; n is the number of wind speed segments;

[0020] A wind speed correction model is constructed based on the correction sub-model sequence P.

[0021] In some embodiments of this application, the wind speed correction strategy includes:

[0022] Multiple monitoring cycles are set based on the initial wind speed curve and wind speed segment sequence A;

[0023] Establish a monitoring cycle sequence W, W = (w1, w2, ..., w i …w r ), where w i Let be the i-th monitoring period; r is the number of monitoring periods;

[0024] Set w sequentially i The target monitoring cycle;

[0025] Generate the monitoring evaluation value c for the target monitoring cycle;

[0026] Establish a time interval series T for the target monitoring period based on the monitoring and evaluation value c;

[0027] T = (t1, t2, ..., tt) i …t r1 ), where t i r1 represents the i-th time interval of the target monitoring cycle; r1 represents the number of time intervals.

[0028] Set the end time of each time interval as the correction time node;

[0029] Determine whether to generate a correction instruction according to the correction time node;

[0030] Establish a time interval sequence for each monitoring period in turn.

[0031] In some embodiments of the present application, when generating the monitoring evaluation value c of the target monitoring period, it includes:

[0032]

[0033] where, e1 is a preset first weight coefficient; e2 is a preset second weight coefficient; Q1 is a preset first fixed coefficient; Q2 is a preset second fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; β i is the influence factor of the i-th first-level disturbance indicator within the target monitoring period; j i is the expected fluctuation value of the i-th first-level disturbance indicator within the target monitoring period; θ1 is the number of auxiliary evaluation indicators; η i is the influence factor of the i-th auxiliary evaluation indicator; v i is the reference value of the i-th auxiliary evaluation indicator within the target monitoring period.

[0034] In some embodiments of the present application, when determining whether to generate a correction instruction, it includes:

[0035] Determine the wind speed section of the target monitoring period;

[0036] Generate a first-level correction model according to the judgment result and the wind speed correction model;

[0037] Obtain the monitoring data packet at the current correction time node;

[0038] Generate the correction evaluation value f at the current correction time node according to the monitoring data packet and the first-level correction model;

[0039] Preset a first correction evaluation value threshold F1 and a second correction evaluation value threshold F2, and F1 < F2;

[0040] If f < F1, no correction instruction is generated at the current correction time node;

[0041] If F1 < f < F2, a first-level correction instruction is generated at the current correction time node;

[0042] If f > F2, a second-level correction instruction is generated at the current correction time node.

[0043] In some embodiments of the present application, when obtaining the monitoring data packet at the current correction time node, it includes:

[0044] Generate a deviation value sequence D at the current correction time node;

[0045] D = (d1, d2…di …d r2 ), where r2 is the number of time intervals between the start time node and the current correction time node within the target monitoring period; d i The deviation value within the i-th time interval of the target monitoring period;

[0046] Generate real-time reference values ​​for each primary disturbance index within the target monitoring period;

[0047] A monitoring data package is generated based on the deviation value sequence D and the real-time reference values ​​of each primary disturbance index.

[0048] In some embodiments of this application, generating the corrected evaluation value f includes:

[0049]

[0050] Where e3 is the preset third weighting coefficient; e4 is the preset fourth weighting coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; β i k is the influence factor of the i-th primary disturbance index within the target monitoring period. i k' is the real-time reference value of the i-th primary disturbance index within the target monitoring period at the current correction time point; i θi represents the standard reference value of the i-th primary disturbance index within the target monitoring period; θ2 represents the number of predicted evaluation indicators; μ i h is the influence factor of the i-th prediction and evaluation index; i This is used to generate a reference value for the i-th prediction and evaluation index based on the deviation value sequence.

[0051] In some embodiments of this application, a wind speed prediction and correction system for wind farms based on machine learning is provided, including:

[0052] The central control unit is used to generate multiple disturbance indicators and wind speed-power correlation curves based on historical monitoring data;

[0053] The monitoring unit is used to collect monitoring data from the wind farm and generate monitoring data packets;

[0054] The central control unit includes:

[0055] The first processing module is used to set multiple wind speed sections based on the wind speed-power correlation curve, and to construct a wind speed correction model based on all disturbance indices and all wind speed sections.

[0056] When setting multiple wind speed zones, these include:

[0057] Establish a sequence of wind speed segments A, A = (a1, a2, ..., a3) i …an ), where a i Let be the i-th wind speed segment; n is the number of wind speed segments.

[0058] The second processing module is used to generate an initial wind speed curve based on the wind speed prediction model.

[0059] The second processing module is also used to set a wind speed correction strategy based on the wind speed correction model and the initial wind speed curve.

[0060] In some embodiments of this application, the first processing module is further configured to:

[0061] Based on the wind speed segment sequence A, a is set sequentially. i For the target wind speed range;

[0062] Training data packets for the target wind speed zone are generated based on historical monitoring data;

[0063] Primary disturbance indicators for target wind speed zones are selected based on all training data packets.

[0064] Establish a first-order disturbance index sequence B for the target wind speed range, B = (b1, b2, ..., bb2) i …b m ), where b i is the i-th primary disturbance index for the i-th target wind speed segment; m is the number of primary disturbance indices for the target wind speed segment;

[0065] Set the influence factors for each disturbance index;

[0066] A modified sub-model for the target wind speed section is constructed based on the first-level disturbance index number B;

[0067] Corrected sub-models for each wind speed range are established sequentially;

[0068] Establish a modified sub-model sequence P, P = (p1, p2, ..., p) i …p n ), where pi is the modified sub-model for the i-th wind speed segment; n is the number of wind speed segments;

[0069] A wind speed correction model is constructed based on the correction sub-model sequence P.

[0070] In some embodiments of this application, the second processing module is further configured to:

[0071] Multiple monitoring cycles are set based on the initial wind speed curve and wind speed segment sequence A;

[0072] Establish a monitoring cycle sequence W, W = (w1, w2, ..., w i …w r ), where w iLet be the i-th monitoring period; r is the number of monitoring periods;

[0073] Set w sequentially i The target monitoring cycle;

[0074] Generate the monitoring evaluation value c for the target monitoring cycle;

[0075]

[0076] Where e1 is the preset first weighting coefficient; e2 is the preset second weighting coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; β i The influence factor of the i-th primary disturbance index within the target monitoring period; j i θ1 represents the expected fluctuation value of the i-th primary disturbance indicator within the target monitoring period; θ1 represents the number of auxiliary evaluation indicators; η i v is the influence factor of the i-th auxiliary evaluation index; i This is the reference value for the i-th auxiliary evaluation indicator within the target monitoring period;

[0077] Establish a time interval series T for the target monitoring period based on the monitoring and evaluation value c;

[0078] T = (t1, t2, ..., tt) i …t r1 ), where t i r1 represents the i-th time interval of the target monitoring cycle; r1 represents the number of time intervals.

[0079] Set the end time of each time interval as the correction time node;

[0080] Determine whether to generate a correction instruction based on the correction time point;

[0081] Establish time interval series for each monitoring period in sequence.

[0082] Compared with existing technologies, the wind speed prediction and correction method and system for wind farms based on machine learning embodiments of this application have the following advantages:

[0083] Based on the nonlinear relationship between wind speed and power, multiple wind speed ranges are constructed, and machine learning technology is used to determine the influence of various disturbance indicators in different wind speed ranges. This allows for the construction of wind speed correction models for each wind speed range, improving the efficiency of correcting predicted wind speed curves and ensuring the operational efficiency of wind farms.

[0084] By collecting various environmental parameters of the wind farm, the environmental fluctuation status of the wind farm can be analyzed in a timely manner, thereby determining whether to dynamically correct the predicted wind speed curve, so as to avoid the decrease in wind speed prediction accuracy due to environmental factors, which would affect the operating efficiency of the wind farm. Attached Figure Description

[0085] Figure 1 This is a flowchart illustrating a preferred embodiment of a wind farm wind speed prediction and correction method based on machine learning. Detailed Implementation

[0086] The specific embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but are not intended to limit the scope of this application.

[0087] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0088] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0089] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0090] like Figure 1 As shown in the preferred embodiment of this application, a wind speed prediction correction method for wind farms based on machine learning includes:

[0091] S101: Generates multiple disturbance indicators and wind speed-power correlation curves based on historical monitoring data;

[0092] S102: Set multiple wind speed sections based on the wind speed-power correlation curve, and construct a wind speed correction model based on all disturbance indices and all wind speed sections;

[0093] S103: Generate an initial wind speed curve based on the wind speed prediction model, and set a wind speed correction strategy based on the wind speed correction model and the initial wind speed curve.

[0094] When setting multiple wind speed zones, these include:

[0095] Establish a sequence of wind speed segments A, A = (a1, a2, ..., a3) i …a n ), where a i Let be the i-th wind speed segment; n is the number of wind speed segments.

[0096] Specifically, a wind speed prediction model is constructed based on existing wind speed prediction technology to periodically predict the wind speed of wind farms.

[0097] Specifically, the initial wind speed curve refers to the predicted wind speed curve generated based on the wind speed prediction model.

[0098] Specifically, based on the nonlinear relationship between wind speed and power, multiple wind speed ranges are defined. Within each wind speed range, the impact of a unit wind speed error on power is basically consistent, with the preferred unit wind speed error being 1 m / s. The impact is the power calculation error when a prediction error occurs.

[0099] Specifically, disturbance indicators include, but are not limited to, parameters such as temperature, humidity, terrain roughness, rainfall, air humidity, atmospheric stability, terrain complexity, and vegetation change.

[0100] Specifically, constructing a wind speed correction model includes:

[0101] Based on the wind speed segment sequence A, a is set sequentially. i For the target wind speed range;

[0102] Training data packets for the target wind speed zone are generated based on historical monitoring data;

[0103] Primary disturbance indicators for target wind speed zones are selected based on all training data packets.

[0104] Establish a first-order disturbance index sequence B for the target wind speed range, B = (b1, b2, ..., bb2) i …b m ), where b i is the i-th primary disturbance index for the i-th target wind speed segment; m is the number of primary disturbance indices for the target wind speed segment;

[0105] Set the influence factors for each disturbance index;

[0106] A modified sub-model for the target wind speed section is constructed based on the first-level disturbance index number B;

[0107] Corrected sub-models for each wind speed range are established sequentially;

[0108] Establish a modified sub-model sequence P, P = (p1, p2, ..., p) i …p n ), where pi is the modified sub-model for the i-th wind speed segment; n is the number of wind speed segments;

[0109] A wind speed correction model is constructed based on the correction sub-model sequence P.

[0110] Specifically, machine learning technology is used to determine the influence of each disturbance index in different wind speed ranges, thereby screening the primary disturbance indexes of the target wind speed range, and constructing the corresponding correction sub-model after quantifying each primary disturbance index.

[0111] Specifically, the first-level disturbance index refers to the index that fluctuates and causes significant interference to real-time wind speed, thereby reducing the accuracy of wind speed prediction.

[0112] Specifically, the modified sub-model includes standard reference values ​​for each first-level disturbance index within the wind speed range. These standard reference values ​​refer to the values ​​that have the least impact on the real-time wind speed in the wind speed range for each first-level disturbance index.

[0113] Specifically, based on the nonlinear relationship between wind speed and power, multiple wind speed ranges are constructed, and machine learning technology is used to determine the influence of each disturbance index in different wind speed ranges. This allows for the construction of wind speed correction models for each wind speed range, improving the efficiency of correcting predicted wind speed curves and ensuring the operational efficiency of wind farms.

[0114] In a preferred embodiment of this application, the wind speed correction strategy includes:

[0115] Multiple monitoring cycles are set based on the initial wind speed curve and wind speed segment sequence A;

[0116] Establish a monitoring cycle sequence W, W = (w1, w2, ..., w i …w r ), where w i Let be the i-th monitoring period; r is the number of monitoring periods;

[0117] Set w sequentially i The target monitoring cycle;

[0118] Generate the monitoring evaluation value c for the target monitoring cycle;

[0119] Establish a time interval series T for the target monitoring period based on the monitoring and evaluation value c;

[0120] T = (t1, t2, ..., tt) i …t r1 ), where t i r1 represents the i-th time interval of the target monitoring cycle; r1 represents the number of time intervals.

[0121] Set the end time of each time interval as the correction time node;

[0122] Determine whether to generate a correction instruction based on the correction time point;

[0123] Establish time interval series for each monitoring period in sequence.

[0124] Specifically, by analyzing the initial wind speed curve, multiple monitoring cycles are set according to the different wind speed sections, and the initial wind speed curve is located in different wind speed sections in two adjacent monitoring cycles.

[0125] Specifically, by analyzing the target monitoring cycle, a corresponding monitoring evaluation value is generated. The larger the monitoring evaluation value, the greater the possibility of wind speed prediction error within the target monitoring cycle. The shorter the time interval of the corresponding single time interval, the higher the efficiency of correcting the predicted wind speed, the lower the impact of environmental fluctuations on the accuracy of wind speed prediction, and the higher the operating efficiency of the wind farm.

[0126] Specifically, when generating the monitoring evaluation value c for the target monitoring cycle, the following are included:

[0127]

[0128] Where e1 is the preset first weighting coefficient; e2 is the preset second weighting coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; β i The influence factor of the i-th primary disturbance index within the target monitoring period; j i θ1 represents the expected fluctuation value of the i-th primary disturbance indicator within the target monitoring period; θ1 represents the number of auxiliary evaluation indicators; η i v is the influence factor of the i-th auxiliary evaluation index; i This is the reference value for the i-th auxiliary evaluation indicator within the target monitoring period.

[0129] Specifically, the expected fluctuation value is generated based on the expected operating parameters of each primary disturbance indicator within the target monitoring period. The larger the expected fluctuation value, the greater the possibility that the real-time value of that primary disturbance indicator will change within the target monitoring period.

[0130] Specifically, the auxiliary evaluation indicators include, but are not limited to, the confidence level of the preset wind speed prediction model, multiple parameters such as the wind speed stability within the target monitoring period, the historical correction times (the number of corrections within the monitoring periods before the target monitoring period), the historical evaluation wind speed error (the prediction error within the monitoring periods before the target monitoring period), etc. By quantifying each auxiliary evaluation indicator, the dynamic adjustment of the correction time node is achieved, and the correction efficiency of the wind speed is improved.

[0131] Specifically, all parameters in the model are normalized by presetting a first fixed coefficient and a second fixed coefficient, so that each parameter in the model is within the same value range.

[0132] In the preferred embodiment of the present application, when judging whether to generate a correction instruction, it includes:

[0133] Judge the wind speed section of the target monitoring period;

[0134] Generate a first-level correction model according to the judgment result and the wind speed correction model;

[0135] Obtain the monitoring data packet at the current correction time node;

[0136] Generate a correction evaluation value f at the current correction time node according to the monitoring data packet and the first-level correction model;

[0137] Preset a first correction evaluation value threshold F1 and a second correction evaluation value threshold F2, and F1 < F2;

[0138] If f < F1, no correction instruction is generated at the current correction time node;

[0139] If F1 < f < F2, a first-level correction instruction is generated at the current correction time node;

[0140] If f > F2, a second-level correction instruction is generated at the current correction time node.

[0141] Specifically, according to the predicted wind speed parameters of the initial wind speed curve within the target monitoring period, the wind speed section to which the target monitoring period belongs is generated, and the correction sub-model of this wind speed section is set as the first-level correction model.

[0142] Specifically, the first correction instruction and the second correction instruction can be set according to historical parameters.

[0143] Specifically, the first-level correction instruction means regenerating the initial wind speed curve according to the change parameters of the current first-level perturbation indicators in combination with the wind speed prediction model.

[0144] Specifically, the second-level correction instruction means that the current wind speed prediction curve has serious prediction errors, and the original wind speed prediction model needs to be optimized and iterated. Based on the optimization results, a new wind speed prediction model is generated, and the initial wind speed curve is regenerated by combining the changing parameters of the current first-level disturbance indicators.

[0145] Specifically, when obtaining the monitoring data packet for the current correction time point, it includes:

[0146] Generate a sequence D of deviation values ​​at the current correction time point;

[0147] D = (d1, d2…d) i …d r2 ), where r2 is the number of time intervals between the start time node and the current correction time node within the target monitoring period; d i The deviation value within the i-th time interval of the target monitoring period;

[0148] Generate real-time reference values ​​for each primary disturbance index within the target monitoring period;

[0149] A monitoring data package is generated based on the deviation value sequence D and the real-time reference values ​​of each primary disturbance index.

[0150] Specifically, generating the corrected evaluation value f includes:

[0151]

[0152] Where e3 is the preset third weighting coefficient; e4 is the preset fourth weighting coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; β i k is the influence factor of the i-th primary disturbance index within the target monitoring period. i k' is the real-time reference value of the i-th primary disturbance index within the target monitoring period at the current correction time point; i θi represents the standard reference value of the i-th primary disturbance index within the target monitoring period; θ2 represents the number of predicted evaluation indicators; μ i h is the influence factor of the i-th prediction and evaluation index; i This is used to generate a reference value for the i-th prediction and evaluation index based on the deviation value sequence.

[0153] Specifically, standard reference values ​​for each primary disturbance index are set according to the modified sub-model within the target monitoring period.

[0154] Specifically, the prediction and evaluation indicators include, but are not limited to, multiple parameters such as the trend of deviation value changes, the variance of the deviation value series, and the average value. By quantifying each prediction and evaluation indicator, the accuracy of the initial wind speed curve is analyzed and evaluated.

[0155] Specifically, the larger the correction value, the greater the prediction error of the current initial wind speed curve.

[0156] Specifically, by pre-setting a third and a fourth fixed coefficient, all parameters in the model are normalized, so that all parameters in the model are within the same range of values.

[0157] It is understandable that in the above embodiments, by collecting various environmental parameters of the wind farm, the environmental fluctuation status of the wind farm is analyzed in a timely manner, thereby determining whether to dynamically correct the predicted wind speed curve, so as to avoid the decrease in wind speed prediction accuracy due to environmental factors, which would affect the operating efficiency of the wind farm.

[0158] In another preferred embodiment of the wind farm wind speed prediction correction method based on machine learning from any of the above preferred embodiments, this preferred embodiment provides a wind farm wind speed prediction correction method based on machine learning, including:

[0159] The central control unit is used to generate multiple disturbance indicators and wind speed-power correlation curves based on historical monitoring data;

[0160] The monitoring unit is used to collect monitoring data from the wind farm and generate monitoring data packets;

[0161] The central control unit includes:

[0162] The first processing module is used to set multiple wind speed sections based on the wind speed-power correlation curve, and to construct a wind speed correction model based on all disturbance indices and all wind speed sections.

[0163] When setting multiple wind speed zones, these include:

[0164] Establish a sequence of wind speed segments A, A = (a1, a2, ..., a3) i …a n ), where a i Let be the i-th wind speed segment; n is the number of wind speed segments.

[0165] The second processing module is used to generate an initial wind speed curve based on the wind speed prediction model.

[0166] The second processing module is also used to set the wind speed correction strategy based on the wind speed correction model and the initial wind speed curve.

[0167] Specifically, the first processing module is also used for:

[0168] Based on the wind speed segment sequence A, a is set sequentially. i For the target wind speed range;

[0169] Training data packets for the target wind speed zone are generated based on historical monitoring data;

[0170] Primary disturbance indicators for target wind speed zones are selected based on all training data packets.

[0171] Establish a first-order disturbance index sequence B for the target wind speed range, B = (b1, b2, ..., bb2) i …b m ), where b i is the i-th primary disturbance index for the i-th target wind speed segment; m is the number of primary disturbance indices for the target wind speed segment;

[0172] Set the influence factors for each disturbance index;

[0173] A modified sub-model for the target wind speed section is constructed based on the first-level disturbance index number B;

[0174] Corrected sub-models for each wind speed range are established sequentially;

[0175] Establish a modified sub-model sequence P, P = (p1, p2, ..., p) i …p n ), where pi is the modified sub-model for the i-th wind speed segment; n is the number of wind speed segments;

[0176] A wind speed correction model is constructed based on the correction sub-model sequence P.

[0177] In a preferred embodiment of this application, the second processing module is further configured to:

[0178] Multiple monitoring cycles are set based on the initial wind speed curve and wind speed segment sequence A;

[0179] Establish a monitoring cycle sequence W, W = (w1, w2, ..., w i …w r ), where w i Let be the i-th monitoring period; r is the number of monitoring periods;

[0180] Set w sequentially i The target monitoring cycle;

[0181] Generate the monitoring evaluation value c for the target monitoring cycle;

[0182]

[0183] Where e1 is the preset first weighting coefficient; e2 is the preset second weighting coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; β i The influence factor of the i-th primary disturbance index within the target monitoring period; j iθ1 represents the expected fluctuation value of the i-th primary disturbance indicator within the target monitoring period; θ1 represents the number of auxiliary evaluation indicators; η i v is the influence factor of the i-th auxiliary evaluation index; i This is the reference value for the i-th auxiliary evaluation indicator within the target monitoring period;

[0184] Establish a time interval series T for the target monitoring period based on the monitoring and evaluation value c;

[0185] T = (t1, t2, ..., tt) i …t r1 ), where t i r1 represents the i-th time interval of the target monitoring cycle; r1 represents the number of time intervals.

[0186] Set the end time of each time interval as the correction time node;

[0187] Determine whether to generate a correction instruction based on the correction time point;

[0188] Establish time interval series for each monitoring period in sequence.

[0189] According to the first concept of this application, multiple wind speed segments are constructed based on the nonlinear relationship between wind speed and power. The influence of each disturbance index in different wind speed segments is judged based on machine learning technology, thereby constructing a wind speed correction model corresponding to each wind speed segment, improving the correction efficiency of the predicted wind speed curve, and thus ensuring the operating efficiency of the wind farm.

[0190] According to the second concept of this application, by collecting various environmental parameters of the wind farm, the environmental fluctuation status of the wind farm can be analyzed in a timely manner, thereby determining whether to dynamically correct the predicted wind speed curve, so as to avoid the decrease in wind speed prediction accuracy due to environmental factors, which would affect the operating efficiency of the wind farm.

[0191] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of this application, and these improvements and substitutions should also be considered within the scope of protection of this application.

Claims

1. A wind speed prediction and correction method for wind farms based on machine learning, characterized in that, including: generating a plurality of disturbance indicators and a wind speed-power correlation curve based on historical monitoring data; setting a plurality of wind speed sections according to the wind speed-power correlation curve, and constructing a wind speed correction model according to all the disturbance indicators and all the wind speed sections; generating an initial wind speed curve according to a wind speed prediction model, and setting a wind speed correction strategy according to the wind speed correction model and the initial wind speed curve; wherein, when setting a plurality of wind speed sections, it includes: Establish wind speed segment sequence ,in, For the first Each wind speed zone; Number of wind speed zones; when constructing the wind speed correction model, it includes: Set sequentially according to wind speed zone sequence A For the target wind speed range; generating a training data packet for a target wind speed section according to historical monitoring data; screening first-level disturbance indicators for the target wind speed section based on all the training data packets; Establish the first-level disturbance index series B for the target wind speed range. ,in, is the i-th primary disturbance index for the i-th target wind speed segment; m is the number of primary disturbance indices for the target wind speed segment; setting influence factors for each disturbance indicator; constructing a correction sub-model for the target wind speed section according to the number B of first-level disturbance indicators; successively establishing correction sub-models for each wind speed section; Establish the modified sub-model sequence P, ,in, This is the modified sub-model for the i-th wind speed segment; n is the number of wind speed segments. constructing a wind speed correction model according to the sequence P of correction sub-models; when setting the wind speed correction strategy, it includes: setting a plurality of monitoring periods according to the initial wind speed curve and the sequence A of wind speed sections; Establish a monitoring cycle series W, ,in, Let be the i-th monitoring period; r is the number of monitoring periods; Set in sequence The target monitoring cycle; generating a monitoring evaluation value c for a target monitoring period; establishing a sequence T of time intervals for the target monitoring period according to the monitoring evaluation value c; ,in, r1 represents the i-th time interval of the target monitoring cycle; r1 represents the number of time intervals. setting the end time node of each time interval as a correction time node; judging whether to generate a correction instruction according to the correction time node; successively establishing sequences of time intervals for each monitoring period.

2. The wind speed prediction and correction method for wind farms based on machine learning as described in claim 1, characterized in that, when generating the monitoring evaluation value c for a target monitoring period, it includes: Where e1 is the preset first weight coefficient; e2 is the preset second weight coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; The influence factor of the i-th primary disturbance index within the target monitoring period; The expected fluctuation value of the i-th primary disturbance indicator within the target monitoring period; To supplement the number of evaluation indicators; Let i be the influence factor of the i-th auxiliary evaluation index; This is the reference value for the i-th auxiliary evaluation indicator within the target monitoring period.

3. The wind speed prediction and correction method for wind farms based on machine learning as described in claim 2, characterized in that, when judging whether to generate a correction instruction, it includes: judging the wind speed section of the target monitoring period; generating a first-level correction model according to the judgment result and the wind speed correction model; acquiring a monitoring data packet at the current correction time node; generating a correction evaluation value f for the current correction time node according to the monitoring data packet and the first-level correction model; presetting a first correction evaluation value threshold F1 and a second correction evaluation value threshold F2, and F1 < F2; if f < F1, no correction instruction is generated at the current correction time node; if F1 < f < F2, a first-level correction instruction is generated at the current correction time node; if f > F2, a second-level correction instruction is generated at the current correction time node.

4. The wind speed prediction and correction method for wind farms based on machine learning as described in claim 3, characterized in that, when acquiring the monitoring data packet at the current correction time node, it includes: generating a sequence D of deviation values at the current correction time node; Where r2 is the number of time intervals between the start time node and the current correction time node within the target monitoring period; The deviation value within the i-th time interval of the target monitoring period; generating real-time reference values for each first-level disturbance indicator within the target monitoring period; generating a monitoring data packet according to the sequence D of deviation values and the real-time reference values of each first-level disturbance indicator.

5. The wind speed prediction and correction method for wind farms based on machine learning as described in claim 4, characterized in that, when generating the correction evaluation value f, it includes: Where e3 is the preset third weighting coefficient; e4 is the preset fourth weighting coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; The influence factor of the i-th primary disturbance index within the target monitoring period; This is the real-time reference value of the i-th primary disturbance index within the target monitoring period at the current correction time point; This is the standard reference value for the i-th primary disturbance index within the target monitoring period; To predict the number of evaluation indicators; Let be the influence factor of the i-th prediction and evaluation index; This is used to generate a reference value for the i-th prediction and evaluation index based on the deviation value sequence.

6. A wind farm wind speed prediction and correction system based on machine learning, employing the wind farm wind speed prediction and correction method based on machine learning as described in any one of claims 1-5, characterized in that, including: a central control unit for generating a plurality of disturbance indicators and a wind speed-power correlation curve according to historical monitoring data; a monitoring unit for collecting monitoring data of a wind farm and generating a monitoring data packet; the central control unit includes: a first processing module for setting a plurality of wind speed sections according to the wind speed-power correlation curve, and constructing a wind speed correction model according to all the disturbance indicators and all the wind speed sections; wherein, when setting a plurality of wind speed sections, it includes: Establish a wind speed range sequence A, ,in, Let be the i-th wind speed segment; n is the number of wind speed segments; a second processing module for generating an initial wind speed curve according to a wind speed prediction model; the second processing module is further used for setting a wind speed correction strategy according to the wind speed correction model and the initial wind speed curve.

7. The wind speed prediction and correction system for wind farms based on machine learning as described in claim 6, characterized in that, the first processing module is further used for: Set sequentially according to wind speed zone sequence A For the target wind speed range; generating a training data packet for a target wind speed section according to historical monitoring data; Primary disturbance indicators for target wind speed zones are selected based on all training data packets. Establish the first-level disturbance index series B for the target wind speed range. ,in, is the i-th primary disturbance index for the i-th target wind speed segment; m is the number of primary disturbance indices for the target wind speed segment; Set the influence factors for each disturbance index; A modified sub-model for the target wind speed section is constructed based on the first-level disturbance index number B; Corrected sub-models for each wind speed range are established sequentially; Establish the modified sub-model sequence P, Where pi is the modified sub-model for the i-th wind speed segment; n is the number of wind speed segments; A wind speed correction model is constructed based on the correction sub-model sequence P.

8. The wind speed prediction and correction system for wind farms based on machine learning as described in claim 7, characterized in that, The second processing module is also used for: Multiple monitoring cycles are set based on the initial wind speed curve and wind speed segment sequence A; Establish a monitoring cycle series W, ,in, Let be the i-th monitoring period; r is the number of monitoring periods; Set in sequence The target monitoring cycle; Generate the monitoring evaluation value c for the target monitoring cycle; Where e1 is the preset first weight coefficient; e2 is the preset second weight coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; m1 is the number of first-level disturbance indicators within the target monitoring period; The influence factor of the i-th primary disturbance index within the target monitoring period; The expected fluctuation value of the i-th primary disturbance indicator within the target monitoring period; To supplement the number of evaluation indicators; Let i be the influence factor of the i-th auxiliary evaluation index; This is the reference value for the i-th auxiliary evaluation indicator within the target monitoring period; Establish a time interval series T for the target monitoring period based on the monitoring and evaluation value c; ,in, r1 represents the i-th time interval of the target monitoring cycle; r1 represents the number of time intervals. Set the end time of each time interval as the correction time node; Determine whether to generate a correction instruction based on the correction time point; Establish time interval series for each monitoring period in sequence.

Images