Inverter power optimization platform and method for large-capacity compressed air energy storage

Abstract

A frequency converter power optimization platform and method for large-capacity compressed air energy storage belongs to the technical field of power optimization of compressed air energy storage. It reflects the correlation and criticality between numerical points through the characteristic values ​​of numerical points, obtains the comparison component of the numerical points and the criticality of the numerical points, and then determines the smoothing factor of the numerical points to fill in the power optimization information, so that the filling of the power optimization information is more accurate, and determines the comparison component by taking into account the time point characteristics of the power optimization information, calibrates the comparison component by the correlation between the characteristics of the numerical points, determines the criticality of the numerical points by the arrangement of the missing numerical points and the complete numerical points in the near area, and uses the criticality of the incomplete numerical points and the comparison component of the complete numerical points as smoothing factors to fill in the power optimization information using the exponential smoothing method. The accuracy of the filled values ​​is improved by adjusting the smoothing factor of the numerical points.

Description

Technical Field

[0001] The present invention belongs to the technical field of power optimization of compressed air energy storage, and in particular relates to a frequency converter power optimization platform and method for large-capacity compressed air energy storage. Background Art

[0002] Compressed-Air Energy Storage (CAES) refers to a storage method that uses electrical energy to compress air during the low-load period of the power grid, seals the air at high pressure in abandoned mines, sunken seabed gas tanks, caves, expired oil and gas wells or newly built gas storage wells, and releases the compressed air to drive steam turbines to generate electricity during the peak load period of the power grid.

[0003] As for the inverter power optimization for large-capacity compressed air energy storage, the existing technical solution with patent publication number "CN103644095B" is often used to achieve inverter power optimization for large-capacity compressed air energy storage, which includes each compressor connected in series (that is, the outlet of the previous compressor is connected to the inlet of the next compressor), each air compressor is driven by a motor, and all motors are electrically connected to a frequency converter; the frequency converter, power transmitter and flow transmitter are all connected to the controller (the controller can be an industrial computer or The power transmitter is used to sample the input electric power value and transmit it to the controller. The flow transmitter is used to sample the flow value of compressed air of each air compressor and transmit it to the controller. When the input electric power value collected by the controller changes in the same power level, it will achieve corresponding detection through the collected compressed air flow value, and control the frequency converter to let the motor allocate the flow value of compressed air of each air compressor to store electric energy. The sampled input electric power and the flow rate of compressed air of each air compressor are the power optimization information.

[0004] Because the power transmitter and flow transmitter are often disturbed by the external magnetic field, the power optimization information collected by the controller for determining whether the input electric power value changes in the same power level or for sampling the flow value of compressed air is inaccurate and accompanied by the loss of its characteristic value, resulting in incomplete power optimization information collected, poor continuity of power optimization information and not conducive to determining whether the input electric power value changes in the same power level or the detection of the flow value of compressed air. Therefore, the exponential smoothing method is often used to supplement the lost characteristic values. However, there are the following defects in specific applications:

[0005] The exponential smoothing method is based on the fact that each numerical point is independent. However, the changes in the power optimization information are correlated at different time points, which makes it difficult to accurately determine the smoothing factors of the numerical points. If there are many missing features in the power optimization information, the accuracy of the power optimization information will be weakened, and the values ​​of the numerical points will be inaccurate, which is not conducive to the determination of the smoothing factors of the numerical points. Summary of the invention

[0006] In order to solve the defects in the prior art, the present invention proposes a frequency converter power optimization platform and method for large-capacity compressed air energy storage, which improves the accuracy of the filled values ​​by adjusting the smoothing factors of the numerical points, effectively avoiding the defects in the prior art that the accuracy of the power optimization information is weakened, the values ​​of the numerical points are inaccurate, and it is not conducive to the identification of the smoothing factors of the numerical points.

[0007] The present invention uses the following technical solutions.

[0008] A frequency converter power optimization method for large-capacity compressed air energy storage, comprising:

[0009] The power transmitter samples the input power value of electric energy and transmits it to the controller for processing. The flow transmitter samples the flow value of compressed air of each air compressor and transmits it to the controller for processing. When the input power value of electric energy after processing changes in the same power level, the controller detects the flow value through the flow value of compressed air after processing, and controls the inverter to let the motor adjust the flow value of compressed air of each air compressor to store electric energy. The sampled input power of electric energy and the flow rate of compressed air of each air compressor are power optimization information.

[0010] The method of processing the sampled input electric power and the flow rate of compressed air of each air compressor, that is, processing the power optimization information, includes:

[0011] Step 1: Initially obtain power optimization information;

[0012] Step 2: By analyzing the numerical value of each numerical point in the power optimization information after pre-processing in the near area, the smoothing factors of all numerical points are determined;

[0013] Step 3: Take the smoothing factor of the numerical point as the smoothing factor of the numerical filling, and use the exponential smoothing method to fill in the missing numerical points in the power optimization information after the previous processing, so as to obtain the power optimization information after processing.

[0014] Furthermore, in step 1, the power optimization information contains multiple numerical points, each numerical point contains multiple features, and the multiple features are the sampled input electric power and the flow rate of compressed air of each air compressor. The specific value of each feature is its characteristic value. According to the outlier status of the characteristic value, pre-processing is performed on the power optimization information to obtain the power optimization information after pre-processing.

[0015] Furthermore, in step 1, the method of performing pre-processing on the power optimization information includes: using the DBSCAN method to set the characteristic value of the power optimization information that is an outlier to NULL, that is, setting the characteristic value to be lost, and performing pre-processing on the characteristic value of each feature of each numerical point of the power optimization information point in sequence to obtain the power optimization information after pre-processing.

[0016] Furthermore, in step 2, the smoothing factor of the value point is determined according to the comparison component of the value point and the criticality of the value point.

[0017] Furthermore, step 2 specifically includes:

[0018] The first power optimization information used in the order of time points Value points , and choose the Value points The previous adjacent one in the time order The value point and the adjacent The total value in the time zone formed by the numerical points numerical points, and then execute the following equation:

[0019] ;

[0020] Characterization The value point The corresponding eigenvalues ​​of the features are The difference between a pair of eigenvalues ​​representing the same feature at adjacent numerical points in time order, Characterization of the The average value is obtained by calculating the difference of all characteristic values ​​in the vicinity of the time point of the numerical point. Characterize the corresponding change trend information of the characteristic value of this numerical point, The standard deviation that characterizes the trend of changes in the characteristic value of this numerical point, Represents the average trend of changes in the near area at that time point, Characterization The time point of the numerical point is near the numerical point in the The amount of loss of a feature; Characterization The value point The comparison weight of a feature.

[0021] Further, in step 2, The time point of the numerical point is near the numerical point in the The loss of the feature = The first The eigenvalue of a feature is the number of missing features ÷ the The first The total number of features.

[0022] Furthermore, step 2 specifically includes: calibrating the initial control component and identifying the control component of the numerical point by numerically analyzing each numerical point in the characteristic near area in the power optimization information after the previous processing; All the features of the numerical points construct the feature near area, that is Near area, according to the characteristics of the near area The correlation between a feature and another feature is The value point The calibration is performed on the starting reference component of the feature, and the The value point The comparison weight of a feature.

[0023] Further, in step 2, The value point The operational equation for the control component of a feature is:

[0024] ;

[0025] Here, Characterization The characteristics of the numerical points and Features The correlation between the two is obtained by obtaining the feature in the feature vicinity corresponding to the numerical point. , The cosine similarity between and Features The correlation between , Respectively characterize the The same as the other features in the feature vicinity of the numerical point The highest and lowest correlations between the features. Characterization The value point The control weight of the corresponding feature of each feature, Characterize the calibration The value point The comparison weight of a feature.

[0026] Furthermore, in step 2, the control components of the numerical points are identified:

[0027] ;

[0028] Here, Characterization Nearby numerical points; Characterization The value points are constructed as midpoints The number of numerical points in the near region, Characterization Nearby area The value point eigenvalues; Characterize intermediate quantities; Characterize the lost value; Characterize the calibrated The value point The comparison weight of the features, Characterization There is no complete number of features with missing eigenvalues ​​in the near region. Characterization The corresponding control component of each numerical point.

[0029] Furthermore, step 2 specifically includes: removing all the numerical points whose eigenvalues ​​are lost, obtaining a complete numerical point group formed by the numerical points with complete features, and constructing an interval near zone for the complete numerical points in the numerical point group, that is, constructing Near area;

[0030] Choose a nearby interval with missing value points , calculate the following equation:

[0031] ;

[0032] Here Characterizing missing value points The interval near the formed area, Represents the total amount of missing value points, Characterize the interval near the structure; Characterization The complete numerical points within The total number of fully characterized numerical points, Characterizing numerical points The control amount, Characterization The number of complete numerical points in the Characterizing numerical points of the control amount; Characterizing numerical points The criticality of .

[0033] Furthermore, step 2 specifically includes: for complete numerical points, taking the comparison components of the complete numerical points as the smoothing factors of the numerical points; for incomplete numerical points, taking the criticality of the incomplete numerical points as the smoothing factors of the numerical points.

[0034] A frequency converter power optimization platform for large-capacity compressed air energy storage, including:

[0035] Each compressor is connected in series, each air compressor is driven by a motor, and all motors are electrically connected to a frequency converter;

[0036] The frequency converter, power transmitter and flow transmitter are all connected to the controller. The power transmitter is used to sample the input power value of electric energy and transmit it to the controller for processing. The flow transmitter is used to sample the flow value of compressed air of each air compressor and transmit it to the controller for processing. When the input power value of electric energy after processing changes in the same power level, the controller detects the flow value through the flow value of compressed air after processing, and controls the frequency converter to let the motor adjust the flow value of compressed air of each air compressor to store electric energy. The sampled input power of electric energy and the flow rate of compressed air of each air compressor are power optimization information.

[0037] The inverter power optimization platform for large-capacity compressed air energy storage also includes:

[0038] A starting module, which is used to initially obtain power optimization information;

[0039] An analysis module, which is used to determine the smoothing factors of all numerical points by numerically analyzing each numerical point in the power optimization information after pre-processing in a near area;

[0040] The filling module is used to use the smoothing factor of the numerical point as the smoothing factor of the numerical filling, and use the exponential smoothing method to fill the missing numerical points in the power optimization information after the previous processing, so as to obtain the power optimization information after processing.

[0041] The beneficial effects of the present invention are that, compared with the prior art, the technical effects of the present invention include:

[0042] By analyzing the correlation and criticality between numerical points, that is, by using the characteristic values ​​of numerical points to reflect the correlation and criticality between numerical points, the comparison components and criticality of numerical points are obtained, and then the smoothing factors of numerical points are determined to fill in the power optimization information, so that the filling of power optimization information is more accurate, the comparison components are determined by considering the time point characteristics of power optimization information, and the comparison components are calibrated by the correlation between numerical point features, and the criticality of numerical points is determined by the arrangement of missing numerical points and complete numerical points in the nearby area, and the criticality of incomplete numerical points and the comparison components of complete numerical points are used as smoothing factors to fill in the power optimization information using exponential smoothing method. The accuracy of the filled values ​​is improved by adjusting the smoothing factors of numerical points. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the inverter power optimization method for large-capacity compressed air energy storage described in the present invention;

[0044] Figure 2 It is a partial structural diagram of the inverter power optimization platform for large-capacity compressed air energy storage described in the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the following will be combined with the drawings in the embodiments of the present invention to clearly and completely express the technical solution of the present invention. The embodiments expressed in this application are only partial embodiments of the present invention, not all embodiments. According to the spirit of the present invention, other embodiments obtained by technicians in this field without making creative work are all within the protection scope of the present invention.

[0046] like Figure 1 As shown, the inverter power optimization method for large-capacity compressed air energy storage described in the present invention includes:

[0047] The power transmitter samples the input power value of electric energy and transmits it to the controller for processing. The flow transmitter samples the flow value of compressed air of each air compressor and transmits it to the controller for processing. When the input power value of electric energy after processing changes in the same power level, the controller detects the flow value through the flow value of compressed air after processing, and controls the inverter to let the motor adjust the flow value of compressed air of each air compressor to store electric energy. The sampled input power of electric energy and the flow rate of compressed air of each air compressor are power optimization information.

[0048] The method of processing the sampled input electric power and the flow rate of compressed air of each air compressor, that is, processing the power optimization information, includes:

[0049] Step 1: Initially obtain power optimization information;

[0050] In a preferred but non-limiting embodiment of the present invention, in step 1, the power optimization information of the present application is features such as the sampled input electric power and the flow rate of compressed air of each air compressor. The power optimization information contains multiple numerical points, and the multiple numerical points have some continuity in the order of time points. Each numerical point contains multiple features, and the multiple features are the sampled input electric power and the flow rate of compressed air of each air compressor. The specific value of each feature is its characteristic value. According to the outlier status of the characteristic value, the power optimization information is pre-processed to obtain the power optimization information after pre-processing.

[0051] In a preferred but non-restrictive embodiment of the present invention, in step 1, a method for performing pre-processing on power optimization information includes: the outlier condition of the characteristic value here is the condition where the characteristic value is an outlier, and the outlier condition is often caused by errors, anomalies, different forms, etc. The DBSCAN method is used to set the characteristic value of the power optimization information that is an outlier to NULL, that is, the characteristic value is set to be lost, and the characteristic value of each feature of each numerical point of the power optimization information point is identified and pre-processed in sequence according to the order of their sampling time points to obtain the power optimization information after pre-processing.

[0052] Because the power optimization information can be used to determine whether the input power value changes within the same power level or to detect the flow value of compressed air, the input power value and the flow rate of compressed air of each air compressor and other corresponding power optimization information are collected. Whenever a power optimization information is collected, a numerical point is obtained, and each numerical point has multiple features. The outlier value corresponding to the feature of each numerical point is obtained through outlier detection, and the detected outlier value is set to NULL, that is, the feature is lost, and the pre-processing is achieved accordingly.

[0053] Step 2: By analyzing the numerical value of each numerical point in the power optimization information after pre-processing in the near area, the smoothing factors of all numerical points are determined;

[0054] In a preferred but non-limiting embodiment of the present invention, in step 2, the smoothing factor of the numerical point is determined based on the comparison component of the numerical point and the criticality of the numerical point.

[0055] Because the exponential smoothing method has simple rules, few critical parameters, and its flexible and adaptable characteristics and distribution characteristics to values, the exponential smoothing method is often used to fill in the sampled power optimization information. When using the exponential smoothing method to fill in the values, it is assumed that each value point is independent, and does not take into account the continuity of the power optimization information transmitted by the sampling in the order of sampling time points (referred to as the time point order). Therefore, the power optimization information transmitted by the combined sampling must meet the characteristics of continuity in the time point order and synchronization of changes between adjacent value points, and thus the comparison weight of the value point is determined; then the category information of the missing value point is combined in The probability in the near area is used to determine the smoothing factor of the numerical point.

[0056] In a preferred but non-limiting embodiment of the present invention, step 2 specifically includes: by analyzing the values ​​of each numerical point in the power optimization information after the previous processing in the near region of the time point, identifying the starting control component of the numerical point; in the order of the time points, The time point near zone is constructed by using numerical points, and the analysis of each numerical point in the time point near zone is The eigenvalue loss condition of the feature is obtained. The numerical point The starting control component of the feature, where and Is a positive integer, which is used to determine the comparison weight of the numerical point.

[0057] Because the existing exponential smoothing method is The power optimization information sampled at each time point in the near area is treated as a single numerical point. There is no correlation between the numerical points. However, the power optimization information transmitted by the sampling is a numerical value in the order of time points, showing a significant dependence on the time points. That is, the current power optimization information is affected by the past power optimization information, just as the changes in the input electric power and the flow rate of compressed air of each air compressor are continuous at the time point. Therefore, the numerical points adjacent to each other in the order of time points in the power optimization information transmitted by the sampling must be stable.

[0058] The first power optimization information used in the order of time points Value points , for the selected The value point The features are analyzed and the first Value points The previous adjacent one in the time order The value point and the adjacent The total value of the near zone of the time point (the near zone is also the near zone of the time point) formed by the numerical points The numerical points are analyzed, and then the following equations are executed:

[0059] ;

[0060] Characterization The value point If the feature value is missing (NULL), it will not be included in the corresponding operation. The difference between a pair of eigenvalues ​​representing the same characteristics of adjacent numerical points in the time point order (the difference is the absolute value of the quantity obtained by subtracting the pair of eigenvalues). If a pair of eigenvalues ​​adjacent in the time point order has a missing value, the operation with the next non-missing eigenvalue is performed over the missing value. Characterization of the The average is obtained by calculating the difference between all the eigenvalues ​​in the vicinity of the time point of the numerical point. Similarly, if there are missing values, they will not be included in the total (that is, in the mean calculation, the total number of divisors is not included). The corresponding change trend information of the characteristic value representing the numerical point reflects the The value point The trend information of the characteristics, The closer the value is to zero, the more stable the change of the characteristics of the numerical point is. Otherwise, the higher the value is, the stronger the change in the vicinity of the time point is. The value can be set in advance.

[0061] The standard deviation of the change trend of the eigenvalue of the numerical point reflects the stability of the numerical point in the near time point. The lower the standard deviation, the more similar the change trend information of the eigenvalue in the near time point is. Represents the average trend of changes in the near area at that time point, Characterization The time point of the numerical point is near the numerical point in the The loss of the first characteristic, that is, The first numerical point in the area near the time point constructed by the numerical point The loss status of a feature. This value is obtained by dividing the number of features lost in the time point near the time point by the total number of features in the time point near the time point. It reflects the loss status of the feature. The higher the value, the more data is lost, which reduces the criticality of the feature and acts on the comparison weight of the feature.

[0062] Characterization The value point The comparison component of each feature reflects the synchronization characteristics of the continuous change of the power optimization information of the feature in the vicinity of the corresponding time point of the numerical point. The more significant the difference, the higher the control value; the smaller the loss, the higher the control value.

[0063] In a preferred but non-limiting embodiment of the present invention, in step 2, The time point of the numerical point is near the numerical point in the The loss of the feature = The first The eigenvalue of the feature is the number of missing features ÷ the The first The total number of features.

[0064] Based on this, the starting comparison components corresponding to the different characteristics of each numerical point in the power optimization information were identified.

[0065] However, the sampled features are not independent, but correlated with each other. Just as the input electric power information will affect the power optimization information, the increase in the input electric power will increase the flow rate of compressed air in the air compressor, thereby storing electric energy. Therefore, there is a correlation between the features. So if there is power optimization information with feature missing (that is, the feature value of the feature is missing), if the feature is correlated with other features, then the missing data of the feature can be represented by the feature value of the other feature.

[0066] However, different numerical points are constructed The distribution characteristics presented by the loss of features in the near area are different, that is, different The missing features of the numerical points in the near area are different, which causes the comparison value obtained through the numerical characteristics in the time point sequence to often show significant differences in a partial characteristic near area. Therefore, the comparison value is calibrated using the appearance of the feature in the characteristic near area.

[0067] Therefore, even if a lot of values ​​on the power optimization information feature are missing, resulting in a very low control weight, its control weight should be higher because it can reflect the changes in the characteristic values ​​of other features; otherwise, if a feature is not missing much, however, because the feature cannot be used to identify the characteristic values ​​of other features, the criticality of the feature of this type of value will often be reduced, just as the increase in input electrical power increases the flow rate of compressed air in the air compressor in direct proportion. Therefore, even if a lot of values ​​of the input electrical power or the flow rate of compressed air in the air compressor are missing, a good estimate of the value can be performed through the characteristic values ​​of other features, thus causing a misunderstanding of the criticality of the feature.

[0068] In a preferred but non-limiting embodiment of the present invention, step 2 specifically further comprises: calibrating the initial control component and identifying the control component of the numerical point by numerically analyzing each numerical point in the characteristic near area in the power optimization information after the previous processing; All the features of the numerical points construct the feature near area, that is Near area, according to the characteristics of the near area The correlation between a feature and another feature is The value point The calibration is performed on the starting reference component of the feature, and the The value point The comparison component of a feature; the feature near area in this application is a feature interval near area.

[0069] In a preferred but non-limiting embodiment of the present invention, in step 2, constructing a numerical point The near-region method includes: removing the features whose eigenvalues ​​are lost, selecting an arbitrary numerical points, The preceding adjacent value of a numerical point in the time order The value point and the adjacent The numerical points are formed Total in the vicinity numerical points, thus constructing Near zone, get the near zone value group (that is, The value group formed by the values ​​in the near area), Can be set in advance, The value point The operational equation for the control component of a feature is:

[0070] ;

[0071] Here, Characterization The characteristics of the numerical points and Features The correlation between the two is obtained by obtaining the feature in the feature vicinity corresponding to the numerical point. , The cosine similarity between and Features The correlation between , Respectively characterize the The same as the other features in the feature vicinity of the numerical point The highest and lowest correlation between the features (which is also the corresponding cosine similarity), Characterization The value point The control weight of the corresponding feature of each feature, Characterize the calibration The value point The comparison weight of a feature.

[0072] According to The value point The comparison of the first characteristic and the The total loss of the features of the numerical points is obtained. Poll all the value points to obtain the comparison weights of all the value points.

[0073] In a preferred but non-limiting embodiment of the present invention, in step 2, because the missing condition of the overall feature cannot represent the missing condition of the partial feature, the comparison component connection condition of the feature in the near area is used to identify the comparison component of the numerical point, and the comparison component of the numerical point is identified:

[0074] ;

[0075] Here, Characterization Nearby numerical points; Characterization The value points are constructed as midpoints The number of numerical points in the near region, Characterization Nearby area The value point eigenvalues; Characterize intermediate quantities; Indicates missing values ​​(that is, the characteristic value is a missing NULL value); if the The numerical points are not missing values, then It is one; Characterize the calibrated The value point The comparison weight of the features, Characterization There is no complete number of features with missing eigenvalues ​​in the near region. Representation by The loss of features in the near area reflects the The corresponding comparison weight of each numerical point is used to determine the criticality of the numerical point.

[0076] By analyzing the value of each numerical point in the power optimization information after pre-processing in the near range of the interval, the criticality of the numerical point is determined. The numerical points construct the near area of ​​the interval, that is, construct Near area, perform analysis on the interval near area where the value point has a missing characteristic value (the value point with the missing characteristic value is the missing value point), analyze the interval point relationship between the complete value points and the missing value points in the near area of ​​the interval, and then combine the first The comparison value of the numerical point is obtained The criticality of the numerical points, Is a positive integer; poll all value points to obtain the criticality of the value points.

[0077] However, in order to determine the actual serial number of the missing numerical point in the power optimization information in the interval (the numerical point is a queue structure (that is, interval) arranged in the order of sampling time points, and the serial number is the serial number of the queue), it is necessary to remove the feature with the missing eigenvalue, resulting in few features for the exponential smoothing method to operate, which leads to little difference between the power optimization information in different nearby areas. If the numerical points of the power optimization information with a large difference as a whole only consider the few features with the missing eigenvalue, the power optimization information will also have a large degree of similarity. The numerical points are not significantly distinguished by only a few features, which is not conducive to the accuracy of the power optimization information filling. Therefore, in the identification of the criticality of the numerical points, the missing features of each numerical point (that is, the features with the missing eigenvalue) must be considered in the interval serial number.

[0078] Because the loss of features will make it impossible to identify the serial number of the numerical point in the interval, the range of the missing feature numerical point and the serial number value of the constructed interval nearby are used to identify the criticality of the numerical point.

[0079] In a preferred but non-limiting embodiment of the present invention, step 2 specifically further comprises: removing all numerical points whose characteristic values ​​are lost, obtaining a complete numerical point group formed by numerical points with complete characteristics, determining the serial number values ​​of the complete numerical points in the numerical point group within the interval, and constructing the interval near area for the complete numerical points (no characteristic values ​​are lost) in the numerical point group, that is, constructing In the near area, when facing the missing value points, because the values ​​are incomplete, the serial numbers in the interval are not constant. Therefore, the serial numbers in the interval are a category information. The category of the value is used in The ratio in the near area determines the criticality of the numerical point;

[0080] If there is no missing value point in the near area of ​​the interval, that is, the eigenvalue is the missing value point, then the criticality of the value point identified based on this is ineffective. Therefore, only the missing value point in the near area of ​​the interval is considered to be effective. The value point of the execution focus is selected to use a range near the interval with the missing value point (The midpoint value point of this near zone is ), calculate the following equation:

[0081] ;

[0082] Here Characterizing missing value points The near area of ​​the interval formed is the missing value point of Nearby area, Represents the total amount of missing value points, The interval near the characterization structure; if the interval near When there is no missing value point in the interval near the value area, The value is zero; The higher the value, the The higher the criticality; Characterization The complete numerical points within The total number of fully characterized numerical points, Characterizing numerical points The higher the control component in the near area of ​​the interval, the higher the criticality of the corresponding numerical point. Characterization The number of complete value points in , Characterize complete numerical points With missing value points The corresponding criticality; Characterizing numerical points of the control amount; Characterizing numerical points The criticality of .

[0083] In a preferred but non-limiting embodiment of the present invention, step 2 specifically further includes: the smoothing factor of the numerical point is determined based on the comparison component of the numerical point and the criticality of the numerical point, and the method is: for complete numerical points, the comparison component of the complete numerical point is used as the smoothing factor of the numerical point; for incomplete numerical points (that is, missing numerical points), the criticality of the incomplete numerical point is used as the smoothing factor of the numerical point. Thus, the smoothing factor of the numerical point is determined.

[0084] If the numerical points are not complete, then their criticality is their comparison component, and the criticality is used as the smoothing factor in the exponential smoothing method.

[0085] So the smoothing factor of the numerical point is determined.

[0086] Step 3: Take the smoothing factor of the numerical point as the smoothing factor of the numerical filling, and use the exponential smoothing method to fill in the missing numerical points in the power optimization information after the previous processing, so as to obtain the power optimization information after processing.

[0087] like Figure 2As shown, the inverter power optimization platform for large-capacity compressed air energy storage described in the present invention includes:

[0088] Each compressor is connected in series (that is, the outlet of the previous compressor is connected to the inlet of the next compressor), each air compressor is driven by a motor, and all motors are electrically connected to a frequency converter;

[0089] The frequency converter, power transmitter and flow transmitter are all connected to the controller (the controller can be an industrial computer or PLC). The power transmitter is used to sample the input power value of electric energy and transmit it to the controller for processing. The flow transmitter is used to sample the flow value of compressed air of each air compressor and transmit it to the controller for processing. The controller is used to detect the flow value through the flow value of compressed air after processing when the input power value of electric energy after processing changes in the same power level (it can be sent to the display screen connected to the controller for display to achieve detection), and control the frequency converter to let the motor adjust the flow value of compressed air of each air compressor to store electric energy. The sampled input power of electric energy and the flow rate of compressed air of each air compressor are power optimization information;

[0090] The inverter power optimization platform for large-capacity compressed air energy storage also includes:

[0091] A starting module, which is used to initially obtain power optimization information;

[0092] An analysis module, which is used to determine the smoothing factors of all numerical points by numerically analyzing each numerical point in the power optimization information after pre-processing in a near area;

[0093] The filling module is used to use the smoothing factor of the numerical point as the smoothing factor of the numerical filling, and use the exponential smoothing method to fill the missing numerical points in the power optimization information after the previous processing, so as to obtain the power optimization information after processing.

[0094] The beneficial effects of the present invention are that, compared with the prior art, the technical effects of the present invention include:

[0095] By analyzing the correlation and criticality between numerical points, that is, by using the characteristic values ​​of numerical points to reflect the correlation and criticality between numerical points, the comparison components and criticality of numerical points are obtained, and then the smoothing factors of numerical points are determined to fill in the power optimization information, so that the filling of power optimization information is more accurate, the comparison components are determined by considering the time point characteristics of power optimization information, and the comparison components are calibrated by the correlation between numerical point features, and the criticality of numerical points is determined by the arrangement of missing numerical points and complete numerical points in the nearby area, and the criticality of incomplete numerical points and the comparison components of complete numerical points are used as smoothing factors to fill in the power optimization information using exponential smoothing method. The accuracy of the filled values ​​is improved by adjusting the smoothing factors of numerical points.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not deviate from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for frequency converter power optimization for large-capacity compressed air energy storage, characterized in that: include: The power transmitter samples the input power value of electric energy and transmits it to the controller for processing. The flow transmitter samples the flow value of compressed air of each air compressor and transmits it to the controller for processing. When the input power value of electric energy after processing changes in the same power level, the controller detects the flow value through the flow value of compressed air after processing, and controls the inverter to let the motor adjust the flow value of compressed air of each air compressor to store electric energy. The sampled input power of electric energy and the flow rate of compressed air of each air compressor are power optimization information. The method of processing the sampled input electric power and the flow rate of compressed air of each air compressor, that is, processing the power optimization information, includes: Step 1: Initially obtain power optimization information; In step 1, the power optimization information contains multiple numerical points, each numerical point contains multiple features, and the multiple features are the sampled input electric power and the flow rate of compressed air of each air compressor. The specific value of each feature is its characteristic value. According to the outlier status of the characteristic value, the power optimization information is pre-processed to obtain the power optimization information after pre-processing; In step 1, a method for performing pre-processing on the power optimization information includes: using the DBSCAN method to set the characteristic value of the power optimization information that is an outlier to NULL, that is, setting the characteristic value to be lost, and performing pre-processing on the characteristic value of each characteristic of each numerical point of the power optimization information point in sequence to obtain the power optimization information after pre-processing; Step 2: by analyzing the values ​​of each value point in the power optimization information after pre-processing in the near area, the smoothing factors of all value points are determined; Step 2 specifically includes: for complete value points, the comparison component of the complete value point is regarded as the smoothing factor of the value point; for incomplete value points, the criticality of the incomplete value point is regarded as the smoothing factor of the value point; Step 3: Take the smoothing factor of the numerical point as the smoothing factor of the numerical filling, and use the exponential smoothing method to fill in the missing numerical points in the power optimization information after the previous processing, so as to obtain the power optimization information after processing.

2. The inverter power optimization method for large-capacity compressed air energy storage according to claim 1 is characterized in that: In step 2, the smoothing factor of the value point is determined according to the comparison component of the value point and the criticality of the value point.

3. The inverter power optimization method for large-capacity compressed air energy storage according to claim 2 is characterized in that: Step 2 specifically includes: The first power optimization information used in the order of time points Value points , and choose the Value points The previous adjacent one in the time order The value point and the adjacent The total value in the time zone formed by the numerical points numerical points, and then execute the following equation: ； Characterization The value point The corresponding eigenvalues ​​of the features are The difference between a pair of eigenvalues ​​representing the same feature at adjacent numerical points in time order, Characterization of the The average value is obtained by calculating the difference of all characteristic values ​​in the vicinity of the time point of the numerical point. Characterize the corresponding change trend information of the characteristic value of this numerical point, The standard deviation that characterizes the trend of changes in the characteristic value of this numerical point, Represents the average trend of changes in the near area at that time point, Characterization The time point of the numerical point is near the numerical point in the The amount of loss of a feature; Characterization The value point The comparison weight of each feature; In step 2, The time point of the numerical point is near the numerical point in the The loss of the feature = The first The eigenvalue of the feature is the number of missing features ÷ the The first The total number of features.

4. The inverter power optimization method for large-capacity compressed air energy storage according to claim 3 is characterized in that: Step 2 specifically includes: calibrating the initial reference component and identifying the reference component of the numerical point by numerically analyzing the value of each numerical point in the characteristic near area in the power optimization information after the previous processing; All the features of the numerical points construct the feature near area, that is Near area, according to the characteristics of the near area The correlation between a feature and another feature is The value point The calibration is performed on the starting reference component of the feature, and the The value point The comparison weight of each feature; In step 2, The value point The operational equation for the control component of a feature is: ； Here, Characterization The characteristics of the numerical points and Features The correlation between the two is obtained by obtaining the feature in the feature vicinity corresponding to the numerical point. , The cosine similarity between and Features The correlation between , Respectively characterize the The same as the other features in the feature vicinity of the numerical point The highest and lowest correlations between the features. Characterization The value point The comparison weight of the features, Characterize the calibration The value point The comparison weight of a feature.

5. The inverter power optimization method for large-capacity compressed air energy storage according to claim 4 is characterized in that: In step 2, identify the control components of the numerical points: ； Here, Characterization Nearby numerical points; Characterization The value points are constructed as midpoints The number of numerical points in the near region, Characterization Nearby area The value point eigenvalues; Characterize intermediate quantities; Characterize the lost value; Characterize the calibrated The value point The comparison weight of the features, Characterization There is no complete number of features with missing eigenvalues ​​in the near region. Characterizes the value point in the near area The corresponding control component of each numerical point.

6. The inverter power optimization method for large-capacity compressed air energy storage according to claim 5 is characterized in that: Step 2 specifically includes: removing all the missing numerical points with eigenvalues, obtaining a complete numerical point group formed by the numerical points with complete features, and constructing an interval near zone for the complete numerical points in the numerical point group, that is, constructing Near area; Choose a nearby interval with missing value points , calculate the following equation: ； Here Characterizing missing value points The interval near the formed area, Represents the total amount of missing value points, Characterize the interval near the structure; Characterization The complete numerical points within The total number of fully characterized numerical points, Characterizing numerical points The control amount, Characterization The number of complete numerical points in the Characterizing numerical points of the control amount; Characterizing numerical points The criticality of .

Images