A single-variable-multivariable steady-state identification method based on combined heat and power units

Abstract

The present invention relates to a single-variable-multivariable steady-state identification method based on a combined heat and power unit, comprising: determining an object; data collection; information statistics and calculation; selecting a steady-state method for each operating condition feature, one by one, according to a judgment condition, while continuously sliding backward with a step size of 1 over a specific time length; judging whether the operating condition feature is in a steady state at a certain moment and marking it; stipulating that when all operating condition features in a data set at the same moment are in a steady state, the moment is the steady-state moment of the system, recording and saving relevant data of continuous steady-state moments, and thus forming a historical steady-state operating condition library. The beneficial effects of the present invention are as follows: the present invention can judge the trend changes of individual features and new steady-state operating conditions, realize the combination from single operating characteristics to multiple operating characteristics, identify the state of the system, and extract the steady-state operating conditions of the system, thereby overcoming the negative effects brought about by operating condition changes.

Description

Technical Field

[0001] The present invention relates to the field of power generation information technology, and more specifically, to a single-variable-multivariable steady-state identification method based on a combined heat and power unit. Background Art

[0002] Because combined heat and power (CHP) units provide both power and heat, their operating conditions fluctuate rapidly and frequently. These operating conditions directly impact the effectiveness of modeling, control, and optimization methods for CHP systems. To mitigate the negative impact of these operating condition fluctuations, it is necessary to identify the system's state and extract its steady-state operating conditions.

[0003] CN105809203B proposes a system steady-state detection algorithm based on hierarchical clustering. By continuously clustering the features between each pair, an expression matrix of the hierarchical clustering tree is obtained. The rationality value of the clustering result is calculated using the expression matrix, thereby calculating and obtaining the final threshold value, and then obtaining the final clustering result sequence. The system steady-state situation is obtained through joint time series judgment. CN109813978B proposes a non-intrusive load identification method that integrates transient characteristics and change characteristics between steady states. This method combines the use of convolutional neural networks to identify the transient process of electrical appliances with the C-means clustering method to identify the change process between electrical appliances between steady states. It identifies the type of electrical appliances and their working state, realizes the effect of system intelligent optimization of electrical appliance identification, and improves the recognition effect of electrical appliance transient and steady-state processes. CN113240105B proposes a grid steady-state discrimination method based on graph neural network pooling. This method preprocesses the information of each node in the grid graph by introducing a graph isomorphism network, and uses maximum likelihood estimation and pooling operations to discriminate the grid operation status.

[0004] The aforementioned patents all identify and judge the operating state of the operating condition, often using clustering algorithms or neural networks. They lack the ability to identify trends in individual features and have difficulty identifying the emergence of new steady-state conditions. Therefore, there is an urgent need to develop a steady-state identification method that combines single operating features with multiple operating features. Summary of the Invention

[0005] The purpose of the present invention is to address the deficiencies of the prior art and propose a single-variable-multivariable steady-state identification method based on a combined heat and power unit.

[0006] In a first aspect, the present invention provides a single-variable-multivariable steady-state identification method based on a combined heat and power unit, comprising:

[0007] Step 1: Determine the objects for which a historical steady-state operating condition library of cogeneration needs to be constructed, and clarify the operating condition characteristics of the relevant objects;

[0008] Step 2: Collect sensor data based on the working condition characteristics, determine the collection duration and frequency, and form a historical working condition dataset X;

[0009] Step 3: Calculate and record statistical information based on the historical working condition data set X; the statistical information includes the mean of the working condition feature i i and standard deviation std i , the historical maximum value of the relevant feature max i and minimum value min i ;

[0010] Step 4: Based on the historical operating condition dataset X and statistical information, a steady-state judgment method is selected for each operating condition feature i when it continuously slides backward with a step size of 1 over a specific time length k. The steady-state judgment methods include: a direct judgment method and a sliding sample entropy judgment method based on dynamic similarity tolerance.

[0011] Step 5: Use certain rules to determine whether the operating condition feature i is in a steady state at a certain moment and mark it;

[0012] Step 6: After determining whether each moment of all the operating condition characteristics of the historical operating condition dataset X is steady-state and marking them, it is stipulated that when all the operating condition characteristics in the dataset X are steady-state at the same moment, this moment is the steady-state moment of the system. The relevant data of the continuous steady-state moments are recorded and saved, thereby forming a historical steady-state operating condition library.

[0013] Preferably, step 4 includes:

[0014] Step 4.1: Dataset X of operating condition feature i in a specific time period k i When (j~j+k-1) is judged by the direct judgment method, all timestamps in this period of the working condition feature are assigned a steady-state mark;

[0015] Step 4.2: The period data set X of the operating condition feature i at a specific time step k i (j~j+k-1) When the sliding sample entropy method of dynamic similarity tolerance is used for judgment, a steady-state mark or a non-steady-state mark is assigned to all time stamps of the working condition feature within the period according to the calculation results.

[0016] Preferably, step 4.2 includes:

[0017] Step 4.2.1, Standardization of time period characteristics: The time period data set X of the working condition feature i at a specific time step k is i (j~j+k-1) get the standardized data set

[0018] Step 4.2.2: Calculate the similarity of the time period features: Calculate the dataset X iThe standard deviation of (j~j+k-1) i (j~j+k-1), thus obtaining the similarity capacity coefficient r0 of feature i in this period, and then obtaining the similarity capacity r=r0×std i (j~j+k-1);

[0019] Step 4.2.3: Construct trend matrix: Using data set X i (j~j+k-1) construct trend matrices A and B;

[0020] Step 4.2.4. Calculate the steady-state trend value of the trend matrix: Count the number of elements in two different rows of A(p) and A(q) in the trend matrix A whose absolute value of the difference is less than the similarity tolerance r. Perform the same statistical operation on the trend matrix B. The statistical result is the trend value of the similarity matrix. The trend values ​​of the trend matrices A and B are denoted as a and b, respectively.

[0021] Step 4.2.5, calculation of sample entropy: use the formula to obtain the sample entropy of the working condition feature i at a specific time step k. The sample entropy formula is

[0022] Step 4.2.6, steady-state and non-steady-state marking: When the sample entropy is greater than the steady-state threshold, all timestamps of the operating condition feature within the period are assigned a steady-state mark; otherwise, all timestamps of the operating condition feature within the period are assigned a non-steady-state mark.

[0023] As a preference, in step 3, the historical maximum value of the relevant feature max i and minimum value min i It is the maximum and minimum value of the operating condition characteristic i within the acceptable time range.

[0024] Preferably, in step 4, the judgment condition is:

[0025] [max(X i (j~j+k-1))-min(X i (j~j+k-1))] <c(max i -min i )

[0026] That is, the data set X of the operating condition feature i in a specific time length k i When the maximum change amplitude of (j~j+k-1) is less than c times the maximum historical change amplitude, the direct judgment method is selected. Otherwise, the sliding sample entropy judgment method of dynamic similarity tolerance is selected, and c is 0 to 0.03.

[0027] Preferably, the steady state is marked as 1 and the non-steady state is marked as 0.

[0028] Preferably, in step 4.2.1, the normalization formula is:

[0029]

[0030] Among them, std i (j~j+k-1) is the data set X i (j~j+k-1) standard deviation, mean i (j~j+k-1) is the data set X i The average value of (j~j+k-1), X i (t) is feature i.

[0031] As a preference, in step 4.2.2, the calculation formula for the similarity coefficient is

[0032] Preferably, in step 5, the rule for judging whether the operating condition feature i is steady-state at a certain moment and marking it is: when the steady-state mark of feature i at moment t accounts for more than 50% of all marks, feature i is considered to be steady-state at moment t, otherwise it is non-steady-state.

[0033] In a second aspect, a single-variable-multivariable steady-state identification device based on a combined heat and power unit is provided, comprising:

[0034] The determination module is used to determine the objects for which a historical steady-state operating condition library of cogeneration needs to be constructed and to clarify the operating condition characteristics of the relevant objects;

[0035] The acquisition module is used to collect sensor data according to the working condition characteristics, determine the acquisition time and frequency, and form the historical working condition data set X;

[0036] The calculation module is used to calculate and record statistical information based on the historical working condition data set X; the statistical information includes the mean of the working condition feature i i and standard deviation std i , the historical maximum value of the relevant feature max i and minimum value min i ;

[0037] The selection module is used to select a steady-state method for each operating condition feature i based on the historical operating condition dataset X and statistical information, when it continuously slides backward with a step size of 1 over a specific time length k. The steady-state judgment methods include: direct judgment method and sliding sample entropy judgment method based on dynamic similarity tolerance;

[0038] The judgment module is used to use certain rules to determine whether the working condition feature i is in a steady state at a certain moment and mark it;

[0039] The recording module is used to determine whether all the working condition characteristics of the historical working condition dataset X are in steady state at each moment and mark them. It stipulates that when all the working condition characteristics in the dataset X are in steady state at the same moment, this moment is the steady state moment of the system. The relevant data of the continuous steady state moments are recorded and saved, thereby forming a historical steady-state working condition library.

[0040] The beneficial effects of the present invention are: the present invention can judge the trend changes of individual characteristics and new steady-state operating conditions, realize the combination from single operating characteristics to multiple operating characteristics, identify the state of the system and extract the steady-state operating conditions of the system, and thus overcome the negative effects brought about by changes in operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a method for calculating the steady-state entropy of sliding samples based on dynamic similarity capacity;

[0042] Figure 2 This is a schematic diagram of the load change trend and steady-state identification of Unit #2;

[0043] Figure 3 This is a schematic diagram of the low-pressure heating flow rate change trend and steady-state identification of Unit #2;

[0044] Figure 4 This is a schematic diagram of the change trend and steady-state identification of the medium-pressure heating flow rate of Unit #2;

[0045] Figure 5 This is a schematic diagram of the operating condition changes and steady-state judgment of Unit #2;

[0046] Figure 6 A schematic diagram showing the contents of the steady-state operating condition library for unit #2. DETAILED DESCRIPTION

[0047] The present invention will be further described below with reference to the following examples. The following examples are provided only to facilitate understanding of the present invention. It should be noted that, without departing from the principles of the present invention, it is possible for a person skilled in the art to make various modifications to the present invention, and such improvements and modifications fall within the scope of the claims of the present invention.

[0048] Example 1:

[0049] Taking a certain group's data standardization project as an example, this application provides a single-variable-multivariable steady-state identification method for a combined heat and power unit, including:

[0050] Step 1: Determine the objects for which a historical steady-state operating condition library of cogeneration needs to be constructed, and clarify the operating condition characteristics of the relevant objects.

[0051] Specifically, the object in the embodiment of the present application is the combined heat and power unit #2, and the operating characteristics of the object include: the load of the #2 unit, the low-pressure heating flow rate, and the medium-pressure heating flow rate.

[0052] Step 2: Collect sensor data based on the working condition characteristics, determine the collection duration and frequency, and form the historical working condition dataset X.

[0053] Specifically, the collection time is 1 year, the sampling frequency is 1 minute, and a total of 525,600 rows of historical operating condition data are collected to form the historical operating condition dataset X.

[0054] Step 3: Calculate and record statistical information based on the historical working condition data set X; the statistical information includes the mean of the working condition feature i i and standard deviation std i , the historical maximum value of the relevant feature max i and minimum value min i .

[0055] It should be noted that the historical maximum value of the relevant feature max i and minimum value min i It is a relative value, which is the maximum and minimum value of feature i within the acceptable time range.

[0056] Step 4: Based on the historical operating condition dataset X and statistical information, a steady-state judgment method is selected one by one according to the judgment conditions for each operating condition feature i when it continuously slides backward with a step size of 1 over a specific time length k. The steady-state judgment methods include: a direct judgment method and a sliding sample entropy judgment method based on dynamic similarity tolerance.

[0057] The judgment conditions are:

[0058] [max(X i (j~j+k-1))-min(X i (j~j+k-1))] <c(max i -min i )

[0059] That is, the data set X of the operating condition feature i in a specific time length k i When the maximum change amplitude of (j~j+k-1) is less than c=0.02 of the historical maximum change amplitude, the direct judgment method is selected; otherwise, the sliding sample entropy judgment method of dynamic similarity tolerance is selected.

[0060] The present application embodiment is described with k=30, j∈[0,525571], and the sliding sample entropy judgment method of the dynamic similarity tolerance is selected for the working condition feature i. The working condition feature i is a period data set X at a specific time step k. i(j~j+k-1) When the sliding sample entropy method of dynamic similarity tolerance is used for judgment, a steady-state mark or a non-steady-state mark is assigned to all time stamps of the working condition feature within the period according to the calculation results.

[0061] Specifically, the sliding sample entropy method of dynamic similarity tolerance includes the following steps:

[0062] Step 4.1: Standardization of time period characteristics: The time period data set X of the working condition feature i at a specific time step k is i (j~j+k-1) get the standardized data set

[0063] The normalization formula is:

[0064]

[0065] Among them, std i (j~j+k-1) is the data set X i (j~j+k-1) standard deviation, mean i (j~j+k-1) is the data set X i The average value of (j~j+k-1), X i (t) is feature i.

[0066] Step 4.2: Calculate the similarity of the time period features: Calculate the dataset X i The standard deviation of (j~j+k-1) i (j~j+k-1), thus obtaining the similarity capacity coefficient r0 of feature i in this period, and then obtaining the similarity capacity r=r0×std i (j~j+k-1);

[0067] Among them, the calculation formula of the similarity capacity coefficient is:

[0068] Step 4.3: Construct trend matrix: Using data set X i (j~j+k-1) construct trend matrices A and B;

[0069]

[0070] (where n = km, m = 2);

[0071] Step 4.4: Calculate the steady-state trend value of the trend matrix: Count the number of elements in the trend matrix A whose absolute value of the difference between the elements in two different rows A(p) and A(q) is less than the similarity tolerance r. Perform the same statistical operation on the trend matrix B. The statistical result is used as the trend value of the similarity matrix. The trend values ​​of the trend matrices A and B are denoted as a and b, respectively.

[0072] Step 4.5, calculation of sample entropy: Use the formula to obtain the sample entropy of the working condition feature i at a specific time step k. The sample entropy formula is

[0073] Step 4.6, steady-state and non-steady-state marking: When the sample entropy is greater than the steady-state threshold, all timestamps of the working condition feature in this period are assigned a steady-state mark; otherwise, all timestamps of the working condition feature in this period are assigned a non-steady-state mark, where the steady-state mark is 1 and the non-steady-state mark is 0. The steady-state threshold is a set value, which can be set according to different working conditions. For example, the threshold can be 1.

[0074] Step 5: Use certain rules to determine whether the operating condition feature i is steady at a certain moment and mark it.

[0075] Specifically, the rule is: when the proportion of steady-state marks of feature i at time t accounts for more than 50% of all marks, feature i is considered to be steady-state at time t, otherwise it is considered to be non-steady-state.

[0076] Step 6: After determining whether each moment of all the operating condition characteristics of the historical operating condition data set X is steady and marking them, it is stipulated that when all the operating condition characteristics in the data set X are steady at the same moment, this moment is the steady state moment of the system. The steady state moment of unit #2 in one day is marked as follows: Figure 5 As shown, the relevant data of the continuous steady-state moments are recorded and saved, thereby forming the historical steady-state operating condition library of unit #2. Part of the contents of the operating condition library are as follows: Figure 6 shown.

[0077] Example 2:

[0078] Based on Example 1, Example 2 of the present application provides another single-variable-multivariable steady-state identification method based on a combined heat and power unit, the method comprising:

[0079] Step 1: Determine the objects for which a historical steady-state operating condition library of cogeneration needs to be constructed, and clarify the operating condition characteristics of the relevant objects.

[0080] Step 2: Collect sensor data based on the working condition characteristics, determine the collection duration and frequency, and form the historical working condition dataset X.

[0081] Step 3: Calculate and record statistical information based on the historical working condition data set X; the statistical information includes the mean of the working condition feature i i and standard deviation std i , the historical maximum value of the relevant feature max i and minimum value min i .

[0082] Step 4: Based on the historical operating condition data set X and statistical information, the steady-state method selection is performed one by one according to the judgment conditions for each operating condition feature i when it continuously slides backward with a step size of 1 over a specific time length k.

[0083] Specifically, the embodiment of the present application selects a direct judgment method, and the data set X of the period of the working condition feature i in a specific time length k is i When (h~j+k-1) is judged by a direct judgment method, it can be directly determined to be a steady state, and all time stamps within this period of the working condition feature are assigned a steady state mark.

[0084] Step 5: Use certain rules to determine whether the operating condition feature i is steady at a certain moment and mark it.

[0085] Step 6: After determining whether each moment of all the operating condition characteristics of the historical operating condition dataset X is steady-state and marking them, it is stipulated that when all the operating condition characteristics in the dataset X are steady-state at the same moment, this moment is the steady-state moment of the system. The relevant data of the continuous steady-state moments are recorded and saved, thereby forming a historical steady-state operating condition library.

[0086] It should be noted that the parts in this embodiment that are the same or similar to those in Example 1 can be referenced to each other and will not be described in detail in this application.

[0087] Example 3:

[0088] Based on Example 1, Example 3 of the present application provides a single-variable-multivariable steady-state identification device based on a combined heat and power unit, comprising:

[0089] The determination module is used to determine the objects for which a historical steady-state operating condition library of cogeneration needs to be constructed and to clarify the operating condition characteristics of the relevant objects;

[0090] The acquisition module is used to collect sensor data according to the working condition characteristics, determine the acquisition time and frequency, and form the historical working condition data set X;

[0091] The calculation module is used to calculate and record statistical information based on the historical working condition data set X; the statistical information includes the mean of the working condition feature i i and standard deviation std i , the historical maximum value of the relevant feature max i and minimum value min i ;

[0092] The selection module is used to select a steady-state method for each operating condition feature i based on the historical operating condition dataset X and statistical information, when it continuously slides backward with a step size of 1 over a specific time length k. The steady-state judgment methods include: direct judgment method and sliding sample entropy judgment method based on dynamic similarity tolerance;

[0093] The judgment module is used to use certain rules to determine whether the working condition feature i is in a steady state at a certain moment and mark it;

[0094] The recording module is used to determine whether all the working condition characteristics of the historical working condition dataset X are in steady state at each moment and mark them. It stipulates that when all the working condition characteristics in the dataset X are in steady state at the same moment, this moment is the steady state moment of the system. The relevant data of the continuous steady state moments are recorded and saved, thereby forming a historical steady-state working condition library.

[0095] Specifically, the system provided in this embodiment is a device corresponding to the method provided in Example 1. Therefore, the parts in this embodiment that are the same or similar to those in Example 1 can be referenced to each other and will not be repeated in this application.

Claims

1. A single-variable-multivariable steady-state identification method based on a combined heat and power unit, characterized in that: include Step 1: Determine the objects for which a historical steady-state operating condition library of cogeneration needs to be constructed, and clarify the operating condition characteristics of the relevant objects; Step 2: Collect sensor data based on the working condition characteristics, determine the collection duration and frequency, and form a historical working condition dataset X; Step 3: Calculate and record statistical information based on the historical working condition data set X; the statistical information includes the mean of the working condition feature i i and standard deviation std i , the historical maximum value of the relevant feature max i and minimum value min i ; Step 4: Based on the historical operating condition data set X and statistical information, a steady-state method is selected for each operating condition feature i when it continuously slides backward with a step size of 1 over a specific time length k. The steady-state judgment methods include: a direct judgment method and a sliding sample entropy judgment method based on dynamic similarity tolerance. Step 4 includes: Step 4.1: Dataset X of operating condition feature i in a specific time period k i When (j~j+k-1) is judged by the direct judgment method, all timestamps in this period of the working condition feature are assigned a steady-state mark; Step 4.2: The period data set X of the operating condition feature i at a specific time step k i (j~j+k-1) When the sliding sample entropy method of dynamic similarity tolerance is used for judgment, a steady-state mark or a non-steady-state mark is assigned to all timestamps of the working condition feature within the period according to the calculation results; In step 4, the judgment conditions are: [max(X i (j~j+k-1))-min(X i (j~j+k-1))] <c(max i -min i ) That is, the data set X of the operating condition feature i in a specific time length k i When the maximum change amplitude of (j~j+k-1) is less than c times of the maximum historical change amplitude), the direct judgment method is selected, otherwise the sliding sample entropy judgment method of dynamic similarity tolerance is selected; c is 0 to 0.03; Step 5: Use a certain rule to determine whether the operating condition feature i is in a steady state at a certain moment and mark it. In step 5, the rule for determining whether the operating condition feature i is in a steady state at a certain moment and marking it is: if the proportion of steady-state marks of feature i at time t accounts for more than 50% of all marks, then feature i is considered to be in a steady state at time t; otherwise, it is considered to be in a non-steady state. Step 6: After determining whether each moment of all the operating condition characteristics of the historical operating condition dataset X is steady-state and marking them, it is stipulated that when all the operating condition characteristics in the dataset X are steady-state at the same moment, this moment is the steady-state moment of the system. The relevant data of the continuous steady-state moments are recorded and saved, thereby forming a historical steady-state operating condition library.

2. The single-variable-multivariable steady-state identification method based on a combined heat and power unit according to claim 1, characterized in that: Step 4.2 includes: Step 4.2.1, Standardization of time period characteristics: The time period data set X of the working condition feature i at a specific time step k is i (j~j+k-1) get the standardized data set Step 4.2.2: Calculate the similarity of the time period features: Calculate the dataset X i The standard deviation of (j~j+k-1) i (j~j+k-1), thus obtaining the similarity capacity coefficient r0 of feature i in this period, and then obtaining the similarity capacity r=r0×std i (j~j+k-1); Step 4.2.3: Construct trend matrix: Using data set X i (j~j+k-1) construct trend matrices A and B; Step 4.2.

4. Calculate the steady-state trend value of the trend matrix: Count the number of elements in two different rows of A(p) and A(q) in the trend matrix A whose absolute value of the difference is less than the similarity tolerance r. Perform the same statistical operation on the trend matrix B. The statistical result is used as the trend value of the similarity matrix. The trend values ​​of the trend matrices A and B are denoted as a and b, respectively. Step 4.2.5, calculation of sample entropy: use the formula to obtain the sample entropy of the working condition feature i at a specific time step k. The sample entropy formula is Step 4.2.6, steady-state and non-steady-state marking: When the sample entropy is greater than the steady-state threshold, all timestamps of the operating condition feature within the period are assigned a steady-state mark; otherwise, all timestamps of the operating condition feature within the period are assigned a non-steady-state mark.

3. The single-variable-multivariable steady-state identification method based on a combined heat and power unit according to claim 1, characterized in that: In step 3, the historical maximum value of the relevant feature max i and minimum value min i It is the maximum and minimum value of the operating condition characteristic i within the acceptable time range.

4. The single-variable-multivariable steady-state identification method based on a combined heat and power unit according to claim 1, characterized in that: The stable state is marked as 1 and the unstable state is marked as 0.

5. The single-variable-multivariable steady-state identification method based on a combined heat and power unit according to claim 1, characterized in that: In step 4.2.1, the normalization formula is: Among them, std i (j~j+k-1) is the data set X i Standard deviation of (j~j+k-1), mean i (j~j+k-1) is the data set X i The average value of (j~j+k-1), X i (t) is feature i.

6. The single-variable-multivariable steady-state identification method based on a combined heat and power unit according to claim 1, characterized in that: In step 4.2.2, the calculation formula for the similarity capacity coefficient is:

7. A single-variable-multivariable steady-state identification device based on a combined heat and power unit, characterized in that: include: The determination module is used to determine the objects for which a historical steady-state operating condition library of cogeneration needs to be constructed and to clarify the operating condition characteristics of the relevant objects; The acquisition module is used to collect sensor data according to the working condition characteristics, determine the acquisition time and frequency, and form the historical working condition data set X; The calculation module is used to calculate and record statistical information based on the historical working condition data set X; the statistical information includes the mean of the working condition feature i i and standard deviation std i , the historical maximum value of the relevant feature max i and minimum value min i ; The selection module is used to select a steady-state method for each operating condition feature i based on the historical operating condition dataset X and statistical information, when it continuously slides backward with a step size of 1 over a specific time length k. The steady-state judgment methods include: direct judgment method and sliding sample entropy judgment method based on dynamic similarity tolerance; The judgment module is used to use certain rules to determine whether the working condition feature i is in a steady state at a certain moment and mark it; The recording module is used to determine whether all the working condition characteristics of the historical working condition dataset X are in steady state at each moment and mark them. It stipulates that when all the working condition characteristics in the dataset X are in steady state at the same moment, this moment is the steady state moment of the system. The relevant data of the continuous steady state moments are recorded and saved, thereby forming a historical steady-state working condition library.

Images