Load regulation-oriented translatable load identification method and system

Abstract

A method and system for identifying transferable loads for load regulation includes: determining the initial change point time of load state change based on extracted load states; calculating the original change point time using a three-window sliding window based on the initial change point time; obtaining the load state change time based on the original change point time and the change point time calculation formula; determining the temporal correlation of non-zero load states based on the initial change point time and the load state change time using association rule learning; determining the frequency of use of different loads based on the load states at different times within the operating cycle, and constructing a strongly correlated load set based on an importance function; determining whether there are non-zero load states with temporal correlation in the strongly correlated load set, and identifying the strongly correlated load set without temporal correlation as a transferable load set, and identifying the loads in the transferable load set as transferable loads. This invention can accurately identify the load state change time and transferable loads.

Description

Technical Field

[0001] This invention relates to the field of power load regulation, and more specifically to a method and system for identifying shiftable loads for load regulation. Background Technology

[0002] In recent years, the energy internet, with the power system at its core, has developed rapidly, gradually realizing the interconnection of everything in the energy context. Relying on the underlying architecture of the Internet of Things (IoT), issues such as energy efficiency, security, and sustainable utilization can be solved automatically, interactively, informationally, and digitally. Transferable loads are important distributed devices in the power system; by modeling transferable loads, more scheduling resources can be provided for load regulation.

[0003] Currently, load control often comes at the expense of user experience, resulting in low user participation in load regulation. Furthermore, as residents' quality of life improves, home appliances are becoming more diversified and intelligent. Therefore, accurately identifying transferable loads has become a key issue affecting the effectiveness of residential load regulation. Summary of the Invention

[0004] To address the problem of inaccurate identification of transferable loads in existing technologies, this invention provides a method for identifying transferable loads for load regulation, comprising:

[0005] The initial change point time of the load state change is determined based on the extracted load state.

[0006] Centered on the initial change point, the time series before and after the initial change point are divided into two categories. The original change point is calculated using a sliding window with three windows, without considering the detection delay time and the influence of the three windows.

[0007] The load state change time is obtained by combining the original change point time without considering the detection delay time and the effect of the three windows with the change point time calculation formula.

[0008] Based on the initial change point time and the load state change time, the non-zero load state is determined, and the temporal correlation of the non-zero load state is determined by association rule learning.

[0009] The frequency of use of different loads is determined by the load status at different times within the operating cycle, and a set of strongly correlated loads is constructed by combining the importance function.

[0010] Determine whether there are any non-zero load states with time correlation in the strongly correlated load set. The strongly correlated load set without non-zero load states with time correlation is taken as the transferable load set, and the loads in the transferable load set are taken as transferable loads.

[0011] The process of determining the non-zero load state based on the initial change point time and the load state change time, and using association rule learning to determine the temporal correlation of the non-zero load state, includes:

[0012] The load state corresponding to the time between the initial change point and the change in load state is a non-zero load state. The frequent itemset of the non-zero load state within a set time period is obtained by learning association rules, wherein the set time period is less than the operating cycle.

[0013] The temporal correlation of the non-zero load state is determined by whether there is an intersection between the non-zero load state at regular intervals and the frequent itemset within the set time period.

[0014] The load state corresponding to the initial change point time and the load state change time is a non-zero load state. Frequent itemsets of non-zero load states within a set time period are obtained using association rule learning, including:

[0015] Define non-zero load states as items, define the set of all non-zero load states within the specified time period as transactions, and use FP-Growth or Apriori algorithms to mine association rules to obtain frequent items;

[0016] Perform a union operation on the frequent itemsets to obtain a frequent itemset, including:

[0017] The times corresponding to frequent items are sorted in ascending order. When the time interval between two adjacent times is less than or equal to the longest time taken for the load state to change, the frequent items corresponding to the two adjacent times are divided into the same interval, and the missing time values ​​in the middle are filled in. Otherwise, a new interval is added to complete the union processing of frequent items.

[0018] The determination of the temporal correlation of the non-zero load state by whether there is an intersection between the non-zero load state at regular intervals and the frequent itemsets within the set time period includes:

[0019] When the non-zero load state at regular intervals intersects with the frequent itemset, the value is 1; otherwise, it is 0, thus constructing a strong time correlation matrix.

[0020] The load's temporal correlation within a set time period is calculated from the strong temporal correlation matrix. When the load's temporal correlation within the set time period is not less than the set load state temporal correlation threshold, the load is not within the set time period, and the load has no strong temporal correlation with the frequent itemsets within the set time period.

[0021] When the temporal correlation of the load within a set time period is less than the set load state temporal correlation threshold, the load has a strong temporal correlation with the frequent itemsets within the set time period.

[0022] The frequency of use of different loads is determined by the load status at different times within the operating cycle, and a strongly correlated load set is constructed by combining the importance function, including:

[0023] A load state matrix is ​​constructed using different times within the operating cycle as columns and different load states within the operating cycle as rows.

[0024] The frequency of use of each load within the operating cycle is determined by the load state matrix, and the importance of each load in different time periods within the operating cycle is calculated by combining the importance function.

[0025] An importance matrix is ​​constructed with the importance of each load at different times within the operating cycle as columns and different loads as rows. Each element in the importance matrix is ​​then assigned a weight based on the sum of all elements in its row.

[0026] The importance matrix is ​​obtained by multiplying each element in the importance matrix by the weight of the sum of all elements in the row in which each element is located.

[0027] The loads corresponding to the elements in each column of the allocated importance matrix that are greater than a set threshold are combined to form a load strongly correlated set.

[0028] The importance function is shown in the following formula:

[0029]

[0030] In the formula, b is the b-th column in the i-th row of the load state matrix A. o represents the importance of the load for each time period within the cycle time. ij Let i be the minimum load state of the i-th load within the cycle time. Let k1 and k2 be the maximum load state of the i-th load within the period, where the absolute value of the difference between k1 and k2 and the average load state of the i-th row in the load state matrix A is less than 0.5.

[0031] The determination of the initial change point time of the load state based on the extracted load state includes:

[0032] Extract the state data of the same load in a time series. When the change of any two adjacent state data exceeds a set threshold, the time corresponding to the two adjacent state data is taken as the initial change point of the load change.

[0033] The process involves dividing the time series before and after the initial change point into two categories, centered on the initial change point time. A three-window sliding window method is used to calculate the original change point time, without considering detection delay time and the influence of the three windows. This includes:

[0034] Taking the initial change point as the center, the time series before the initial change point is taken as class I0, and the time series after the initial change point is taken as class I1.

[0035] The average value of all load currents corresponding to class I0 and class I1 is calculated based on the truncated nodes, the sample length within the window, and the load current within the window, respectively.

[0036] When the average load current of the time series corresponding to class I0 and class I1 satisfies the condition of reaching the minimum value of the objective function, and the absolute value of the difference between the average load current of the time series corresponding to class I0 and class I1 is greater than the set constraint condition of load current change value, the initial change point time for dividing class I0 and class I1 is identified as the original change point time without considering the detection delay time and the influence of the three windows.

[0037] The objective function is shown in the following equation:

[0038]

[0039] In the formula, m(I0) and m(I1) are the average values ​​of the load currents corresponding to time series I0 and I1, respectively, u is the number of load currents in I0 category, and x i Let i be the load current. V The length of the sample within the window.

[0040] The calculation of the acceptable translation period for each translatable load includes:

[0041] Within a set time period, the inverse time correlation interval of each load state in the set of loads that can be shifted is removed, forming a correlation interval composed of multiple time periods;

[0042] When the duration of a time period within the correlation interval is not less than the duration of the load set that can be shifted, that time period is considered an acceptable shifting period.

[0043] When the duration of a time period within the correlation interval is less than the duration of the transferable load set, that time period is considered an unacceptable transfer period.

[0044] The time of load state change is calculated using the following formula:

[0045] t=λ(t′-t d )

[0046] In the formula, λ is the correction coefficient, t′ is the original change point time without considering the detection delay time and the effect of the three windows, and t is the change point time. d This is the window detection delay time.

[0047] Based on the same inventive concept, the present invention also provides a load shifting identification system for load regulation, comprising:

[0048] The initial change point calculation module is used to determine the initial change point time of the load state change based on the extracted load state.

[0049] The original change point time calculation module is used to divide the time series before and after the initial change point time into two categories, with the initial change point time as the center, and to calculate the original change point time without considering the detection delay time and the influence of the three windows using a sliding window with three windows.

[0050] The load change time calculation module is used to obtain the load state change time based on the original change point time without considering the detection delay time and the effect of the three windows, combined with the change point time calculation formula.

[0051] The time correlation module is used to determine the non-zero load state based on the initial change point time and the load state change time, and to determine the time correlation of the non-zero load state by using association rule learning.

[0052] The load strong association set construction module is used to determine the frequency of use of different loads by the load status at different times within the operating cycle, and to construct a load strong association set by combining the importance function.

[0053] The movable load determination module is used to determine whether there is a non-zero load state with time correlation in the strongly correlated load set, and to take the strongly correlated load set without non-zero load state with time correlation as the movable load set, and the load in the movable load set as the movable load.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] 1. This invention provides a method for identifying transferable loads for load regulation, comprising: determining the initial change point time of load state change based on extracted load state; calculating the original change point time using a three-window sliding window based on the initial change point time; obtaining the load state change time based on the original change point time and the change point time calculation formula; determining the temporal correlation of non-zero load states based on the initial change point time and the load state change time using association rule learning; determining the frequency of use of different loads based on the load state at different times within the operating cycle, and constructing a load strongly correlated set based on the importance function; determining whether there are time-correlated non-zero load states in the load strongly correlated set, and using the load strongly correlated set without time-correlated non-zero load states as the transferable load set, and the loads in the transferable load set as transferable loads. This invention can accurately identify the load state change time and form a load strongly correlated set, revealing the electricity consumption habits of residential users.

[0056] 2. This technical solution utilizes a three-sliding window and the rule of minimum intra-class variance to accurately obtain the moment data of load state changes.

[0057] 3. This technical solution uses the Apriori algorithm to solve for frequent terms through a layer-by-layer search and iterative method, which improves computational efficiency. Attached Figure Description

[0058] Figure 1 This is a flowchart of a load shifting identification method for load regulation according to the present invention.

[0059] Figure 2 This is a schematic diagram of three-window variable point detection;

[0060] Figure 3 This is a flowchart illustrating the specific steps of a load-control-oriented method for identifying shiftable loads according to the present invention. Detailed Implementation

[0061] This invention discloses a method and system for identifying shiftable loads for load regulation. This device accurately identifies the time of load state changes and forms a strongly correlated load set, revealing the electricity consumption habits of residential users.

[0062] Example 1: A method for identifying shiftable loads for load regulation, such as... Figure 1 As shown, it includes:

[0063] Step 1: Determine the initial change point time of the load state based on the extracted load state;

[0064] Step 2: Taking the initial change point as the center, divide the time series before and after the initial change point into two categories, and use a sliding window with three windows to calculate the original change point without considering the detection delay time and the influence of the three windows.

[0065] Step 3: Based on the original change point time without considering the detection delay time and the effect of the three windows, the change point time calculation formula is used to obtain the load state change time;

[0066] Step 4: Determine the non-zero load state based on the initial change point time and the load state change time, and use association rule learning to determine the temporal correlation of the non-zero load state;

[0067] Step 5: Determine the frequency of use of different loads based on the load status at different times within the operating cycle, and construct a set of strongly correlated loads by combining the importance function;

[0068] Step 6: Determine whether there are any non-zero load states with time correlation in the strongly correlated load set. The strongly correlated load set without non-zero load states with time correlation is taken as the transferable load set, and the loads in the transferable load set are taken as transferable loads.

[0069] The purpose of this invention is to propose a method for identifying shiftable loads for load regulation, which can provide more effective scheduling resources for residential load regulation. The steps are described in detail below:

[0070] Step 1: Determine the initial change point time of the load state based on the extracted load state, specifically including:

[0071] Define a statistical function. If the change in the statistical value exceeds a certain threshold, it indicates that the load status has changed. By combining the load information before and after the change point in step two, the load status can be obtained.

[0072] Load assignment and load shedding represent changes in load operating status. Considering that load shedding and assignment are opposite operations, the specific process for load status extraction is as follows:

[0073] 1) Let the time series of extracted load data be T={t(k)}, t=1,2,…, and define the statistical function f as:

[0074]

[0075] In the formula: μ0 is the average value before the load condition changes; θ is the random noise introduced from the outside, and the statistic when the load is stable. and It is a random variable with a mean of 0, meaning it fluctuates randomly around the value of 0. When the load is applied, t k It will increase, making It has an increasing trend, and vice versa. It will also decrease, and when the change exceeds the set threshold, the change in load status can be detected.

[0076] Step 2: Centered on the initial change point, divide the time series before and after the initial change point into two categories. Calculate the original change point time using a three-window sliding window method, without considering the detection delay and the influence of the three windows. Specifically, this includes:

[0077] Variable point detection is performed by using a sliding window and connecting the three windows end to end.

[0078] After extracting the load operating status, the threshold effect can make the time points corresponding to load status changes in step 1 inaccurate. To accurately determine the times when the load starts and stops, a sliding window is used for change point detection.

[0079] Load state extraction window C1 is used to calculate the mean μ0 of the sampled sequence, and load state extraction window C2 is used to detect change points. However, two-window change point detection is prone to missed detections. Therefore, a three-window approach is used, with the first and last windows connected, i.e., adding window C3. Figure 2 As shown, the load current data samples are centered on the time point corresponding to the load change state in step 1, and two types of time series are taken before and after it, which are defined as I0 types {x1, x2, ..., x...}. u} and I1 class {x u+1 x u+2 , ..., x v The I1 class data is processed later than the I0 class data in time. The I0 class data is computed in window C2, and the I1 class data is computed in window C3, where u is the truncated node and V is the sample length within the window.

[0080]

[0081]

[0082] In the formula: m(I0) and m(I1) are the average values ​​of the load currents corresponding to time series I0 and I1, respectively. When the objective function is satisfied:

[0083]

[0084] When the minimum value of |m(I1)-m(I0)| is reached and |m(I1)-m(I0)| is greater than the set load current change value, it can be considered as the original change point without considering the detection delay time and the influence of the three windows.

[0085] Step 3: Based on the original change point time (excluding detection delay and the impact of the three-window method) and the change point time calculation formula, the load state change time is obtained, specifically including:

[0086] Considering the window detection delay time t d In addition to the impact of the three-window detection, the change point time t will also be affected. Taking all factors into account, the change point time t is:

[0087] t=λ(t′-t d )

[0088] Among them, the correction coefficient λ is to reduce the impact of the large time deviation caused by the three-window detection, 0.9≤λ≤1; t′ is the original change point time without considering the detection delay time and the impact of the three windows.

[0089] Step 4: Based on the initial change point time and the load state change time, determine the non-zero load state, and use association rule learning to determine the temporal correlation of the non-zero load state, specifically including:

[0090] Establish association rules, solve for frequent terms, connect the frequent terms into frequent time intervals, and calculate the time data corresponding to the non-zero state of load i at regular intervals T. s A certain time period [t] of frequent itemsets s ,t e The probability of intersection exists. If the probability of intersection is higher than the load state time correlation threshold, then the load is considered to be related to [t]. s ,t e It has a strong time correlation.

[0091] 1) Compile the daily non-zero state data for load i. This data can be obtained by combining steps one and two (in step one). The load input is increasing and is in a non-zero state. If the load is relatively small, it represents a zero-load state. Step two obtains the accurate time points when the load state changes (the data between these two time points represents the time points when the load is non-zero). The obtained state sequence is divided into daily units, and the sampling time points for each day are recorded as t1, t2, ..., t H Define each non-zero state sampling point of the load as an item. The transactions of load i on a given day include all sampling points of that day, denoted as... Where d represents day d, association rules are established, and each association rule contains two parameters: support and confidence. Support describes the importance of the association rule, and confidence describes the accuracy of the association rule. Non-zero load states are defined as terms, and the set of all non-zero load states within a given time period is defined as transactions. The Apriori algorithm is used to mine association rules and solve for frequent terms.

[0092] 2) Perform a union operation on the frequent items obtained in 1), sort the time points in ascending order, and then connect them to form a time interval. The specific rule is that if the data of two adjacent time points are less than or equal to t... g If the values ​​belong to the same interval, fill in any missing time values ​​in between; otherwise, create a new interval. Here, t... g It represents the longest interruption time during which a user completes a set of electricity consumption behaviors, and its value is equal to the longest time experienced when the load state changes.

[0093] 3) Determine whether load i has strong temporal correlation. The specific method is as follows: First, define the strong temporal correlation matrix:

[0094]

[0095] Where: Y mn This represents the time data corresponding to the non-zero state of load i at regular intervals T (e.g., one day, i.e., m=24, etc.) and the n time periods [t] of the frequent itemset within a certain time period (usually less than the period T). s ,t e Does Y have an intersection?mn =1 indicates that there is an intersection. The temporal correlation is determined by the following method.

[0096]

[0097] in, P represents the temporal correlation of load i within this time period. thre This indicates the set load state time correlation threshold; r i =0 indicates that the load is not within this cycle, and the load i and [t] are considered to be outside this cycle. s ,t e There is no strong time correlation; the load can be shifted within this period. i =1 indicates that this load within this cycle, i and [t] s ,t e It has a strong time correlation, and the load cannot be shifted within this period.

[0098] 4) Repeat steps 1) to 3) until all loads are identified based on time correlation.

[0099] By introducing a frequency factor and defining an importance function, the frequency of load usage for each load in each time period is analyzed, and a set of strongly correlated loads is formed using a correlation matrix.

[0100] Suppose there are m different loads A1, A2, A3, ..., A m Each load includes different states of the load at different times. A load state matrix is ​​established, where each column represents a different time period and different rows represent different loads (except for the first row):

[0101]

[0102] In definition A, the first element is the frequency factor C, which represents the frequency of use of household appliances in each time period. n = n, where a larger n indicates more frequent use of appliances. For example, appliances are used more frequently in the morning, noon, and evening, so the corresponding n value is larger. At other times, the frequency factor n is smaller. This division better reflects the actual usage of household appliances and is beneficial for defining strongly correlated load sets; A mn This represents the load state at the nth time interval of the m-th type of load; [a 1n ,…,a mn ] T It indicates the load status within the same time period under different loads; the load status is represented by the load current value.

[0103] Define importance function It indicates the importance of each period within the cycle. for:

[0104]

[0105] In the formula, o ij =min{k|a ik ≠0, k=1,2,…,n} B = {1,2,…,n}; k1 and k2 are elements in set K, and k1 and k2 belong to B. Elements in set K represent two adjacent columns of matrix A, and it is stipulated that k1 ≤ k2. The elements in set K are the load state average μ(A) of the i-th row of the load state matrix A. i The value of ), that is:

[0106] K = {k∈B:|k-μ(A)} i )|≤0.5}

[0107] In the formula, if |k-μ(A) i If | < 0.5, then set K contains only one element, i.e., k1 = k2. Considering the influence of the frequency factor, the mean μ(A) i )for:

[0108]

[0109] In the formula, a ik c represents the k-th element in the i-th row of the load matrix A. k It is the value of the k-th column in the frequent factors of the first row of matrix A.

[0110] Based on this, the importance of each load at each time period is obtained. for:

[0111]

[0112] in, For two values ​​in the frequency factor C in the first row of the load state matrix A, Let k1 be the value of the frequent factors in the first row of matrix A. Let be the value of the k2th column of the frequent factors in the first row of matrix A.

[0113] By listing the importance of each load across different time periods into a matrix, we obtain an m×n correlation matrix Z, where each column represents a different time period and each row represents the importance of the same load across different time periods.

[0114]

[0115] Assign elements in each row of the matrix according to their weights:

[0116]

[0117] In the formula, z ij z represents the importance of load i in the j-th time period. ik Let i represent the importance of load i in the k-th time period, where k and j are both time period numbers.

[0118] The loads corresponding to each element in each column that exceed a set threshold are combined to form a strongly correlated load set. For example, induction cookers and dishwashers have a strong correlation over a period of time. Here, the threshold can be set to 1 / n, where n represents the number of elements in each row.

[0119] Step 6: Determine whether there are any time-related non-zero load states in the strongly correlated load set. The strongly correlated load set without time-related non-zero load states is designated as the transferable load set, and the loads in the transferable load set are designated as transferable loads. Specifically, this includes:

[0120] The set of strongly correlated loads is divided into movable loads and non-movable loads. The set containing load states with time correlation is the non-movable load set, and the other sets are the movable load sets.

[0121] The invention also includes: calculating the acceptable translation period for each transferable load, specifically including:

[0122] Within a set time period, the inverse time correlation interval of each load state in the set of loads that can be shifted is removed, forming a correlation interval composed of multiple time periods;

[0123] When the duration of a time period within the correlation interval is not less than the duration of the load set that can be shifted, that time period is considered an acceptable shifting period.

[0124] When the duration of a time period within the correlation interval is less than the duration of the transferable load set, that time period is considered an unacceptable transfer period.

[0125] Let the time of a day be {t1,t2,…,t} H Remove the inverse time correlation interval Φ for each load state in the set of loads that can be shifted. r (Inverse time correlation refers to the fact that certain states of a load never appear within a specific time period), forming... The form is (t) q1 ,t z1 )∪(t q2 ,t z2 )…(t qr ,t zr )…(t qR ,t zR The method for determining whether a certain time period is an acceptable shift period is as follows:

[0126]

[0127] Among them, L Φ Let t be the time length of a certain movable load set. zr -t qr Let λ represent a certain period of a day for the set of loads that can be moved; λ = 1 indicates that the period is an acceptable time for moving, otherwise it is not.

[0128] The beneficial result of this invention is that it provides a method for identifying shiftable loads oriented towards load regulation, such as... Figure 3 As shown, firstly, load status extraction is performed, specifically by defining a statistical function and using changes in statistical quantities to extract the load operating status. Then, the time of load status change is obtained, specifically by using three-window change point detection to collect the time of load status change. Next, the temporal correlation of load status is identified. Then, a frequency factor is introduced, and a load status importance function is defined to form a strongly correlated load set. Specifically, this includes describing the usage frequency of each load within each time period to form a strongly correlated load set. Then, the load set is divided, specifically by using the temporal correlation of loads to divide the strongly correlated load set into movable loads and non-movable loads. Finally, the acceptable movable period for each movable load is calculated, specifically by calculating the movable load for the analysis day based on the load characteristic identification results, and outputting the movable load set for that day and the corresponding acceptable movable period.

[0129] Example 2

[0130] Based on the same inventive concept, the present invention also provides a load-shifting load identification system for load regulation, comprising:

[0131] The initial change point calculation module is used to determine the initial change point time of the load state change based on the extracted load state.

[0132] The original change point time calculation module is used to divide the time series before and after the initial change point time into two categories, with the initial change point time as the center, and to calculate the original change point time without considering the detection delay time and the influence of the three windows using a sliding window with three windows.

[0133] The load change time calculation module is used to obtain the load state change time based on the original change point time without considering the detection delay time and the effect of the three windows, combined with the change point time calculation formula.

[0134] The time correlation module is used to determine the non-zero load state based on the initial change point time and the load state change time, and to determine the time correlation of the non-zero load state by using association rule learning.

[0135] The load strong association set construction module is used to determine the frequency of use of different loads by the load status at different times within the operating cycle, and to construct a load strong association set by combining the importance function.

[0136] The movable load determination module is used to determine whether there is a non-zero load state with time correlation in the strongly correlated load set, and to take the strongly correlated load set without non-zero load state with time correlation as the movable load set, and the load in the movable load set as the movable load.

[0137] The time correlation module includes:

[0138] The frequent itemset determination submodule is used to obtain frequent itemsets of non-zero load states within a set time period by learning association rules when the load state is non-zero between the initial change point time and the load state change time. The set time period is less than the operating cycle.

[0139] The correlation submodule is used to determine the temporal correlation of the non-zero load state by whether there is an intersection between the non-zero load state at regular intervals and the frequent itemset within the set time period.

[0140] The frequent itemset determination submodule includes:

[0141] The frequent item calculation unit is used to define non-zero load states as items, define the set of all non-zero load states within the set time period as transactions, and use FP-Growth or Apriori algorithms to mine association rules to obtain frequent items.

[0142] The union processing unit is used to perform union processing on the frequent items to obtain a frequent itemset.

[0143] The union processing unit is specifically used to sort the time corresponding to the frequent items in ascending order. When the time interval between two adjacent time data is less than or equal to the longest time taken for the load state to change, the frequent items corresponding to the two adjacent time times are divided into the same interval, and the missing time values ​​in the middle are filled in. Otherwise, a new interval is added to complete the union processing of frequent items.

[0144] The original change point time calculation module includes:

[0145] The category division and mean processing submodule is used to take the time series before the initial change point as the center, the time series after the initial change point as the I0 class, and the time series after the initial change point as the I1 class. Based on the intermediate truncation node, the sample length in the window, and the load current in the window, the average value of all load currents corresponding to the time series of the I0 class and the I1 class are calculated respectively.

[0146] The original change point identification submodule is used to identify the initial change point of classifying I0 and I1 as the original change point without considering the detection delay time and the influence of the three windows when the average load current corresponding to the time series I0 and I1 satisfies the condition of reaching the minimum value of the objective function and the absolute value of the difference between the average load current corresponding to the time series I0 and I1 is greater than the set constraint condition of the load current change value.

[0147] The load-strongly correlated set construction module includes:

[0148] The load matrix construction submodule is used to construct a load state matrix with different times within the operating cycle as columns and different load states within the operating cycle as rows.

[0149] The importance calculation submodule is used to determine the frequency of use of each different load within the operating cycle from the load state matrix, and to calculate the importance of each load in different time periods within the operating cycle in combination with the importance function;

[0150] The weight allocation submodule is used to construct an importance matrix with the importance of each load in different time periods within the operating cycle as columns and different loads as rows. The elements in the importance matrix are allocated according to the weight of each element relative to the sum of all elements in the row. The allocated importance matrix is ​​obtained by multiplying each element in the importance matrix by the weight of each element relative to the sum of all elements in the row.

[0151] The combination submodule is used to combine the loads corresponding to the elements in each column of the allocated importance matrix that are greater than a set threshold, to form a load strongly correlated set.

[0152] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0153] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0154] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0155] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0156] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A method for identifying shiftable loads for load regulation, characterized in that, include: The initial change point time of the load state change is determined based on the extracted load state. Centered on the initial change point, the time series before and after the initial change point are divided into two categories. The original change point is calculated using a sliding window with three windows, without considering the detection delay time and the influence of the three windows. The load state change time is obtained by combining the original change point time without considering the detection delay time and the effect of the three windows with the change point time calculation formula. Based on the initial change point time and the load state change time, the non-zero load state is determined, and the temporal correlation of the non-zero load state is determined by association rule learning. The frequency of use of different loads is determined by the load status at different times within the operating cycle, and a set of strongly correlated loads is constructed by combining the importance function. Determine whether there are any non-zero load states with time correlation in the strongly correlated load set. The strongly correlated load set without non-zero load states with time correlation is taken as the set of loads that can be moved. The loads in the set of loads that can be moved are taken as loads that can be moved. The process involves dividing the time series before and after the initial change point into two categories, centered on the initial change point time. A three-window sliding window method is used to calculate the original change point time, without considering detection delay time and the influence of the three windows. This includes: Centered on the initial change point, the time series before the initial change point is classified as class I0, and the time series after the initial change point is classified as class I0. kind; The time series was calculated based on the intermediate truncated nodes, the sample length within the window, and the load current within the window. Class and The average value of all load currents corresponding to the class; When the time series is Class and The average load current corresponding to the class satisfies the requirement of reaching the minimum value of the objective function, and the time series is... Class and When the absolute value of the difference between the average load current values ​​corresponding to a class is greater than the set constraint condition of the load current change value, the class will be divided. Class and The initial change point of the class is defined as the original change point without considering the detection delay time and the effect of the three windows; The objective function is shown in the following equation: In the formula, and The time series are respectively Class and The average value of the load current corresponding to the class, u is The number of load currents of each class. Let i be the load current. The length of the sample within the window.

2. The method as described in claim 1, characterized in that, The process of determining the non-zero load state based on the initial change point time and the load state change time, and using association rule learning to determine the temporal correlation of the non-zero load state, includes: The load state corresponding to the time between the initial change point and the change in load state is a non-zero load state. The frequent itemset of the non-zero load state within a set time period is obtained by learning association rules, wherein the set time period is less than the operating cycle. The temporal correlation of the non-zero load state is determined by whether there is an intersection between the non-zero load state at regular intervals and the frequent itemset within the set time period.

3. The method as described in claim 2, characterized in that, The load state corresponding to the time between the initial change point and the time of load state change is a non-zero load state. Frequent itemsets of non-zero load states within a set time period are obtained using association rule learning, including: Define non-zero load states as items, define the set of all non-zero load states within the specified time period as transactions, and use FP-Growth or Apriori algorithms to mine association rules to obtain frequent items; Perform a union operation on the frequent itemsets to obtain the frequent itemset.

4. The method as described in claim 3, characterized in that, The process of performing a union operation on the frequent itemsets to obtain a frequent itemset includes: The times corresponding to frequent items are sorted in ascending order. When the time interval between two adjacent times is less than or equal to the longest time taken for the load state to change, the frequent items corresponding to the two adjacent times are divided into the same interval, and the missing time values ​​in the middle are filled in. Otherwise, a new interval is added to complete the union processing of frequent items.

5. The method as described in claim 2, characterized in that, The determination of the temporal correlation of the non-zero load state by whether there is an intersection between the non-zero load state at regular intervals and the frequent itemsets within the set time period includes: When the non-zero load state at regular intervals intersects with the frequent itemset, the value is 1; otherwise, it is 0, thus constructing a strong time correlation matrix. The load's temporal correlation within a set time period is calculated from the strong temporal correlation matrix. When the load's temporal correlation within the set time period is not less than the set load state temporal correlation threshold, the load is not within the set time period, and the load has no strong temporal correlation with the frequent itemsets within the set time period. When the temporal correlation of the load within a set time period is less than the set load state temporal correlation threshold, the load has a strong temporal correlation with the frequent itemsets within the set time period.

6. The method as described in claim 2, characterized in that, The frequency of use of different loads is determined by the load status at different times within the operating cycle, and a strongly correlated load set is constructed by combining the importance function, including: A load state matrix is ​​constructed using different times within the operating cycle as columns and different load states within the operating cycle as rows. The frequency of use of each load within the operating cycle is determined by the load state matrix, and the importance of each load in different time periods within the operating cycle is calculated by combining the importance function. An importance matrix is ​​constructed with the importance of each load at different times within the operating cycle as columns and different loads as rows. Each element in the importance matrix is ​​then assigned a weight based on the sum of all elements in its row. The importance matrix is ​​obtained by multiplying each element in the importance matrix by the weight of the sum of all elements in the row in which each element is located. The loads corresponding to the elements in each column of the allocated importance matrix that are greater than a set threshold are combined to form a load strongly correlated set.

7. The method as described in claim 1, characterized in that, The importance function is shown in the following formula: In the formula, b is the b-th column in the i-th row of the load state matrix A. The importance of the load for each time period within the cycle time. Let i be the minimum load state of the i-th load within the cycle time. This represents the maximum load state of the i-th load within a given period. The absolute value of the difference between the load state mean and the load state mean in the i-th row of the load state matrix A is less than 0.5; .

8. The method as described in claim 1, characterized in that, The determination of the initial change point time of the load state based on the extracted load state includes: Extract the state data of the same load in a time series. When the change of any two adjacent state data exceeds a set threshold, the time corresponding to the two adjacent state data is taken as the initial change point of the load change.

9. The method as described in claim 1, characterized in that, Also includes: Calculate the acceptable relocation period for each relocatable load.

10. The method as described in claim 9, characterized in that, The calculation of the acceptable translation period for each translatable load includes: Within a set time period, the inverse time correlation interval of each load state in the set of loads that can be shifted is removed, forming a correlation interval composed of multiple time periods; When the duration of a time period within the correlation interval is not less than the duration of the load set that can be shifted, that time period is considered an acceptable shifting period. When the duration of a time period within the correlation interval is less than the duration of the transferable load set, that time period is considered an unacceptable transfer period.

11. The method as described in claim 10, characterized in that, The time of load state change is calculated using the following formula: In the formula, For correction factor, The original change point time is t, which does not consider the detection delay time and the effect of the three windows. This is the window detection delay time.

12. A system for implementing the load-oriented shiftable load identification method as described in any one of claims 1-11, characterized in that, include: The initial change point calculation module is used to determine the initial change point time of the load state change based on the extracted load state. The original change point time calculation module is used to divide the time series before and after the initial change point time into two categories, with the initial change point time as the center, and to calculate the original change point time without considering the detection delay time and the influence of the three windows using a sliding window with three windows. The load change time calculation module is used to obtain the load state change time based on the original change point time without considering the detection delay time and the effect of the three windows, combined with the change point time calculation formula. The time correlation module is used to determine the non-zero load state based on the initial change point time and the load state change time, and to determine the time correlation of the non-zero load state by using association rule learning. The load strong association set construction module is used to determine the frequency of use of different loads by the load status at different times within the operating cycle, and to construct a load strong association set by combining the importance function. The movable load determination module is used to determine whether there is a non-zero load state with time correlation in the strongly correlated load set, and to take the strongly correlated load set without non-zero load state with time correlation as the movable load set, and the load in the movable load set as the movable load.

Images