Abnormal electricity utilization detection method based on load curve correlation degree

Abstract

The invention relates to an abnormal electricity utilization detection method based on load curve correlation, which comprises the following steps: firstly, collecting electricity utilization data of a user, and screening out a suspected abnormal electricity utilization date by utilizing voltage and current characteristics; calculating a correlation mean value of the characteristic load curve and the daily load curve in the period of time, and introducing an Euclidean distance d to improve the accuracy; finally, setting a self-adaptive threshold value, combining the self-adaptive threshold value with the model, and evaluating the relevancy of different users; wherein if the relevancy of the user is larger than or equal to the self-adaptive threshold value, it is judged that the power consumption condition is abnormal. According to the method, whether the electrical load of the user is greatly changed or not is judged by calculating the correlation degree between the characteristic electrical load curve and the daily electrical load curve, and the Pearson's correlation coefficient method and the Euclidean distance method are comprehensively used, so that the change condition of the electrical load of the user is evaluated more comprehensively.

Description

Technical Field

[0001] The present invention relates to the technical field of abnormal power consumption detection, and in particular to an abnormal power consumption detection method based on load curve correlation. Background Art

[0002] With the rapid development of the economy, the power consumption of users has also shown explosive growth, and higher requirements have been put forward for abnormal power consumption detection. Abnormal power consumption behavior not only directly affects the income of power supply companies, but also hinders the development of smart grids. Therefore, it is very necessary to carry out abnormal power consumption detection based on the information of user power consumption data.

[0003] At present, the research and analysis of abnormal electricity usage detection has been greatly improved. However, there are still the following problems in abnormal electricity usage monitoring: first, the difference between some abnormal electricity usage and normal electricity usage is very small, which is difficult to detect; second, there are few real abnormal samples in reality, and the model used for detection cannot be effectively trained, and the final detection effect is not ideal.

[0004] At the same time, according to data from the power department, abnormal load events for low-voltage non-residential and low-voltage residential users account for more than 90%, so most abnormal electricity consumption behaviors will be accompanied by changes in electricity load. By calculating the correlation between the characteristic electricity load curve and the daily electricity load curve, it can be determined whether the user has abnormal electricity consumption behavior.

[0005] It should be understood that the above statements merely provide background technology related to the present invention and do not necessarily constitute prior art. Summary of the invention

[0006] The purpose of the present invention is to provide a method for detecting abnormal power consumption based on load curve correlation, so as to improve the accuracy and flexibility of abnormal power consumption detection.

[0007] In order to achieve the above-mentioned object, the present invention provides an abnormal power consumption detection method based on load curve correlation, comprising the following steps: S1, collecting the power consumption data of the user within a period of time, and respectively analyzing the characteristics of the current and voltage in the power consumption data of the user within this period of time, and screening out the dates suspected of abnormal power consumption within this period of time; S2, calculating the characteristic load curve of the normal power consumption date within this period of time, and extracting the abnormal power consumption load of a suspected abnormal power consumption day within this period of time, and obtaining the daily load curve of the abnormal power consumption day; S3, calculating the correlation ρ between the characteristic load curve and the daily load curve at the same time x,y ; S4, using the sliding window value method, calculate the correlation mean of the characteristic load curve and the daily load curve S5, introduce the Euclidean distance d, and the correlation mean Combined with the correlation combination model c, the correlation combination model c of the characteristic load curve and the daily load curve is obtained; S6, setting an adaptive threshold, combined with the correlation combination model c, the correlation adaptive threshold combination model c of the characteristic load curve and the daily load curve is obtained x , evaluate the relevance of different users; if the user's relevance is ≥ the adaptive threshold, it is determined to be an abnormal power consumption situation.

[0008] Preferably, in said S1, the characteristic of the current is characterized by the current three-phase imbalance ib;

[0009] The calculation formula of the three-phase current imbalance ib is:

[0010]

[0011] Among them, I A is the user's A phase current value; I B is the user's B phase current value; I C is the user C phase current value; Mean() is the average value function; abs() is the absolute value function; Mean(I A , I B , I C ) is the average value of the user's three-phase current; abs(I i -Mean(I A ,I B ,I C )) is the absolute value of the deviation between the user's i-phase current and the average value of the user's three-phase current.

[0012] Preferably, in S1, the characteristics of the current can also be characterized by a current correlation coefficient IR;

[0013]

[0014] Among them, m and n are any two phases of the three-phase current of users A, B, and C respectively; I mk Indicates the current value of the user's m-phase current at the kth moment; Indicates the average current value of the user's m-phase current at 24 moments; Indicates the standard deviation of the user's m-phase current at 24 moments.

[0015] Preferably, in S1, the voltage characteristic is characterized by the voltage three-phase unbalance vb;

[0016] The calculation formula of the voltage three-phase unbalance vb is:

[0017]

[0018] Among them, V Ais the user's A phase voltage value; V B is the user's B phase voltage value; V C is the user C phase voltage value; Mean() is the average value function; abs() is the absolute value function; Mean(V A , V B , V C ) is the average value of the three-phase voltage of the user; abs(V i -Mean(V A ,V B ,V C )) is the absolute value of the deviation between the user's i-phase voltage and the average value of the user's three-phase voltage.

[0019] Preferably, in said S1, the voltage characteristic can also be characterized by a rated voltage deviation vd;

[0020]

[0021] Among them, the rated voltage deviation vd of each phase voltage of the user p The calculation formula is:

[0022]

[0023] Where p is any phase voltage among the three-phase voltages of users A, B, and C; vu p Vr is the user's p-phase voltage value; p It is the user's p-phase rated voltage value, and the user's A-phase, B-phase, and C-phase rated voltage values ​​are the same.

[0024] Preferably, in S3, the correlation ρ x,y The calculation formula is:

[0025]

[0026] Among them, x and y represent the load values ​​of the characteristic load curve and the daily load curve at the same time; x i and i Represent the load values ​​of characteristic load and daily load at the i-th moment respectively; and Represent the mean load values ​​of characteristic load and daily load at all times respectively; cov(x,y) represents the covariance of the load values ​​of characteristic load curve and daily load curve; σ x and σ y They represent the standard deviation of the load values ​​of the characteristic load curve and the daily load curve respectively.

[0027] Preferably, in S4, the correlation mean is calculated using a sliding window value method. The specific calculation formula is:

[0028]

[0029] Among them, s is the logarithm of the load value at the corresponding moment; n is the window value size; It is the sum of the correlations between the characteristic load curves and the daily load curve calculated from the 1st window to the nth window; 1 / sn represents the averaging coefficient of the summed result.

[0030] Preferably, in S5, the calculation formula of the correlation combination model c is:

[0031]

[0032] Wherein, k represents the amplification factor.

[0033] Preferably, the calculation formula of the Euclidean distance d is:

[0034]

[0035] Among them, x n and n They represent the load values ​​at the same time on two days.

[0036] Preferably, in S6, the correlation adaptive threshold combination model c x The calculation formula is:

[0037]

[0038] Among them, ρ spe is the threshold coefficient; |s a -s x | is the absolute difference between the total daily electricity load and the total characteristic electricity load.

[0039] To sum up, compared with the prior art, the present invention provides an abnormal power consumption detection method based on load curve correlation, which determines whether there is a significant change in the user's power load by calculating the correlation between the characteristic power load curve and the daily power load curve, and comprehensively uses the Pearson correlation coefficient and Euclidean distance methods to more comprehensively evaluate the changes in the user's power load. While improving the detection efficiency, it also improves the accuracy and flexibility of abnormal power consumption detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The present invention is a flow chart of the abnormal power consumption detection method based on load curve correlation. DETAILED DESCRIPTION

[0041] The following is combined with Figure 1, the present invention is further explained by describing a preferred specific embodiment in detail.

[0042] It should be noted that the drawings are in a very simplified form and use non-precise proportions. They are only used to conveniently and clearly assist in explaining the embodiments of the present invention, and are not used to limit the conditions for the implementation of the present invention. Therefore, they have no substantive technical significance. Any structural modification, change in proportional relationship or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical content disclosed by the present invention.

[0043] It should be noted that, in the present invention, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only the elements explicitly listed, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0044] like Figure 1 As shown, the present invention provides an abnormal power consumption detection method based on load curve correlation, which specifically comprises the following steps:

[0045] S1. Collect the user's electricity consumption data over a period of time, analyze the characteristics of current and voltage in the user's electricity consumption data during this period, and filter out the dates of suspected abnormal electricity consumption during this period;

[0046] S2. Calculate the characteristic load curve of the normal power consumption date within the period, and extract the abnormal power consumption load of a suspected abnormal power consumption day within the period to obtain the daily load curve of the abnormal power consumption day;

[0047] S3. Calculate the correlation ρx between the characteristic load curve and the daily load curve at the same time. ,y ;

[0048] S4. Using the sliding window value method, calculate the correlation mean of the characteristic load curve and the daily load curve

[0049] S5, introduce the Euclidean distance d, and the average correlation between the characteristic load curve and the daily load curve obtained in S4 Combined, the correlation combination model c of characteristic load curve and daily load curve is obtained;

[0050] S6. Set an adaptive threshold value and combine it with the correlation combination model c to obtain the characteristic load curve and the daily load curve correlation adaptive threshold combination model c. x , evaluate the relevance of different users; if the user's relevance is ≥ the adaptive threshold, it is determined to be an abnormal power consumption situation.

[0051] Furthermore, in the above S1, by analyzing the three-phase current of normal power consumption and abnormal power consumption, it can be found that the current value of each phase of the three-phase current of normal power consumption is very close, and the change trend is very similar, while the three-phase current of abnormal power consumption has power theft behavior in one or more phases, which will make the current value of each phase have a large difference, and in extreme cases, the current of one phase will be 0. Therefore, two characteristics that characterize the difference between the two can be proposed, namely: the three-phase current imbalance degree ib and the current correlation coefficient IR.

[0052] Furthermore, the calculation formula of the current three-phase imbalance degree ib is:

[0053]

[0054] Among them, I A is the user's A phase current value; I B is the user's B phase current value; I C is the user C phase current value; Mean() is the average value function; abs() is the absolute value function; Mean(I A , I B , I C ) is the average value of the user's three-phase current; abs(I i -Mean(I A ,I B ,I C )) is the absolute value of the deviation between the user's i-phase current and the average value of the user's three-phase current.

[0055] It can be understood that the smaller the three-phase current imbalance degree ib is, the lower the probability of current abnormality is, and vice versa.

[0056] Furthermore, the current correlation coefficient IR is represented by the minimum value of the Pearson correlation coefficient between the two-phase currents, wherein the formula of the Pearson correlation coefficient between the two phases of the current is:

[0057]

[0058] Among them, m and n are any two phases of the three-phase current of users A, B, and C respectively; I mk Indicates the current value of the user's m-phase current at the kth moment; Indicates the average current value of the user's m-phase current at 24 moments; Indicates the standard deviation of the user's m-phase current at 24 moments.

[0059] Then the current correlation coefficient IR is:

[0060]

[0061] Among them, the larger the current correlation coefficient IR is, the lower the probability of current abnormality is, and vice versa.

[0062] In addition, it should be noted that in the calculation formulas of the above-mentioned current three-phase imbalance degree ib and the current correlation coefficient IR, the situation where the user's single-phase current is valid or the user's three-phase currents are all invalid will only occur under extremely extreme conditions. User behavior generally does not produce such conditions, so it is not considered here.

[0063] Furthermore, in S1, by analyzing the three-phase voltages of normal power consumption and abnormal power consumption, it can be found that the three-phase voltages of normal power consumption fluctuate around the rated voltage, while the abnormal power consumption, due to the theft of electricity on one or more phases, will make the measured voltage much lower than the rated voltage, and in extreme cases, it will be 0. Therefore, two characteristics that characterize the voltage difference can be proposed: the rated voltage deviation vd and the voltage three-phase imbalance vb.

[0064] Furthermore, the rated voltage deviation vd of each phase voltage of the user p The calculation formula is:

[0065]

[0066] Where p is any phase voltage among the three-phase voltages of users A, B, and C; vu p Vr is the user's p-phase voltage value; p is the user's p-phase rated voltage value, where the user's A-phase, B-phase, and C-phase rated voltage values ​​are the same.

[0067] It is worth noting that, for the user's three-phase voltage, the rated voltage deviation vd is the maximum deviation of the user's three-phase voltage, that is, the calculation formula is:

[0068]

[0069] The smaller the rated voltage deviation vd is, the lower the probability of voltage abnormality is, and vice versa.

[0070] Furthermore, the voltage three-phase imbalance vb is similar to the current three-phase imbalance ib, and its calculation formula is:

[0071]

[0072] Among them, VA is the user's A phase voltage value; V B is the user's B phase voltage value; V C is the user C phase voltage value; Mean() is the average value function; abs() is the absolute value function; Mean(V A , V B , V C ) is the average value of the three-phase voltage of the user; abs(V i -Mean(V A ,V B ,V C )) is the absolute value of the deviation between the user's i-phase voltage and the average value of the user's three-phase voltage.

[0073] It can be understood that the smaller the voltage three-phase imbalance vb is, the lower the probability of voltage abnormality is, and vice versa. In other words, the data of each phase of normal voltage is relatively concentrated, and the rated voltage deviation and imbalance are relatively small, tending to 0; while the data of each phase of abnormal voltage is relatively scattered, and the rated voltage deviation and imbalance are relatively large.

[0074] Similarly, in the above calculation formulas for rated voltage deviation vd and voltage three-phase imbalance vb, the situation where the user's single-phase voltage is valid or the user's three-phase voltage is invalid will only occur under extremely extreme conditions. User behavior generally does not produce such conditions, so they are not considered here.

[0075] Furthermore, in the S2, it should be noted that, in the present invention, 15 minutes is taken as one moment, and 24 hours a day is divided into 96 moments in total. The load values ​​of 96 moments on the day of abnormal power consumption are extracted to obtain the daily load curve.

[0076] Furthermore, in S2, 5 to 10 normal electricity consumption days without obvious abnormal events and equipment failures are taken, and the load average value at the corresponding time of each normal electricity consumption day is calculated to obtain the characteristic load curve.

[0077] Furthermore, in S3, the correlation between the characteristic load curve and the daily load curve at the same time is x,y The calculation formula is:

[0078]

[0079] Where x and y represent the load values ​​of the characteristic load curve and the daily load curve at the same time; x i and i Represent the load values ​​of characteristic load and daily load at the i-th moment respectively; and Represent the mean load values ​​of characteristic load and daily load at all times respectively; cov(x,y) represents the covariance of the load values ​​of characteristic load curve and daily load curve; σ x and σ y They represent the standard deviation of the load values ​​of the characteristic load curve and the daily load curve respectively.

[0080] Furthermore, in the above ρ x,y In the calculation, only the load values ​​of the time values ​​corresponding to the single set of the characteristic load curve and the daily load curve are used. Therefore, in order to reduce the error, in the S4, the sliding window value method is used to calculate the correlation mean of the characteristic load curve and the daily load curve. The specific calculation formula is:

[0081]

[0082] Among them, s is the logarithm of the load value at the corresponding moment; n is the window value size; It is the sum of the correlations between the characteristic load curves and the daily load curve calculated from the 1st window to the nth window; 1 / sn represents the averaging coefficient of the summed result.

[0083] Furthermore, in S5, it should be noted that when the load curves for any two consecutive days have a high correlation, but the total load changes greatly, the above It is impossible to accurately judge the correlation between the characteristic load curve and the daily load curve, so the Euclidean distance d is introduced to reflect the change in the total amount.

[0084]

[0085] Among them, x n and n They represent the load values ​​at the same time on two days.

[0086] It can be understood that a smaller value of the Euclidean distance d indicates a higher correlation, and a larger value indicates a lower correlation.

[0087] Furthermore, due to the The value range of is [-1,1]. In order to ensure that the obtained correlation is a positive value, The value range of is shifted to [0,2] through the operation of ρ+1, and is multiplied by the amplification factor k to ensure that the final characteristic load curve and the daily load curve correlation combination model c has sufficient weight to facilitate subsequent intuitive comparison.

[0088] Therefore, the calculation formula of the correlation combination model c of the characteristic load curve and the daily load curve is:

[0089]

[0090] Wherein, k represents the amplification factor.

[0091] Furthermore, in S6, in order to evaluate the correlation of different users, an adaptive threshold needs to be set, and the adaptive threshold is combined with the correlation combination model c of the characteristic load curve and the daily load curve to obtain the correlation adaptive threshold combination model c of the characteristic load curve and the daily load curve x To evaluate user relevance, the following steps are included:

[0092] S61. Analyze the correlation between the user's historical daily power load curve and the characteristic power load curve x,y , when the number of selected corresponding time points, that is, the number of selected corresponding load values, is greater than 300, more than 90% of ρ x,y Less than a certain value ρ spe , and the specific value ρ spe As ρ x,y Adaptive threshold of

[0093] S62: When the difference between the total daily power load and the total characteristic power load exceeds one fifth of the total characteristic power load, it is regarded as a potential load change abnormality, thereby obtaining a characteristic load curve and a daily load curve correlation adaptive threshold combination model c. x The calculation formula is:

[0094]

[0095] Among them, ρ spe is the adaptive threshold; |s a -s x | is the absolute difference between the total daily power load and the total characteristic power load, where c x That is, it represents the obtained user relevance.

[0096] It is worth noting that the adaptive thresholds of different users are different. When the adaptive threshold is small, it means that the user's electricity usage is highly correlated with the characteristic load, indicating that the user's electricity usage behavior conforms to the normal electricity usage pattern and the degree of deviation from the characteristic load is small; when the adaptive threshold is large, it means that the user's electricity usage is less correlated with the characteristic load, the possibility of abnormal electricity usage by the user is relatively high, and the user's electricity usage is greatly different from the normal pattern.

[0097] Furthermore, abnormal power usage can be determined by comparing the adaptive threshold with the correlation. If the correlation is ≥ the adaptive threshold, it is determined to be abnormal power usage; if the correlation is < the adaptive threshold, it is determined to be normal power usage.

[0098] In a preferred embodiment of the present invention, taking the power consumption data of a user in a certain period of time in 2024 as an example, the date with a high probability of abnormal power consumption is located as July 15 through voltage and current factor characteristic analysis, and then the load at the corresponding time of several days (for example, the 10th) with normal daily power load curve is averaged, the characteristic power load curve is extracted, and the correlation within this period is calculated, as shown in Table 1:

[0099] Table 1: User relevance over time

[0100] date Relevance July 10 0.845 July 11 0.730 July 12 1.612 July 13 1.076 July 14 0.728 July 15 0.086 July 16 0.723 July 17 0.945 July 18 1.012 July 19 0.673 July 20 1.133

[0101] Therefore, according to the above table, there is a large difference between the correlation on July 15 and the correlation on other days. It is judged that there is an abnormal power consumption event, which causes the correlation between the load curve on that day and the characteristic power load curve to be much smaller than the correlation of nearby dates.

[0102] In summary, the abnormal power consumption detection method based on load curve correlation of the present invention determines whether the user's power load has undergone significant changes by calculating the correlation between the characteristic power load curve and the daily power load curve, and comprehensively uses the Pearson correlation coefficient and Euclidean distance methods to more comprehensively evaluate the changes in the user's power load, which has great practical value.

[0103] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be appreciated that the above description should not be considered as a limitation of the present invention. After reading the above content, it will be apparent to those skilled in the art that various modifications and substitutions of the present invention will occur. Therefore, the protection scope of the present invention should be limited by the appended claims.

Claims

1. A method for detecting abnormal power consumption based on load curve correlation, characterized in that: The following steps are involved: S1. Collect the user's electricity consumption data over a period of time, analyze the characteristics of current and voltage in the user's electricity consumption data during this period, and filter out the dates of suspected abnormal electricity consumption during this period; S2. Calculate the characteristic load curve of the normal power consumption date within the period, and extract the abnormal power consumption load of a suspected abnormal power consumption day within the period to obtain the daily load curve of the abnormal power consumption day; S3. Calculate the correlation ρ between the characteristic load curve and the daily load curve at the same time. x,y ; S4. Using the sliding window value method, calculate the correlation mean of the characteristic load curve and the daily load curve S5, introduce the Euclidean distance d, and the correlation mean Combined, the correlation combination model c of the characteristic load curve and the daily load curve is obtained; S6. Set an adaptive threshold value and combine it with the correlation combination model c to obtain the characteristic load curve and the daily load curve correlation adaptive threshold combination model c. x , evaluate the relevance of different users; if the user's relevance is ≥ the adaptive threshold, it is determined to be an abnormal power consumption situation.

2. The abnormal power consumption detection method based on load curve correlation according to claim 1, characterized in that: In the above S1, the current characteristic is characterized by the current three-phase imbalance degree ib; The calculation formula of the three-phase current imbalance ib is: Among them, I A is the user's A phase current value; I B is the user's B phase current value; I C is the user C phase current value; Mean() is the average value function; abs() is the absolute value function; Mean(I A , I B , I C ) is the average value of the user's three-phase current; abs(I i -Mean(I A ,I B ,I C )) is the absolute value of the deviation between the user's i-phase current and the average value of the user's three-phase current.

3. The abnormal power consumption detection method based on load curve correlation according to claim 1, characterized in that: In the above-mentioned S1, the characteristics of the current can also be characterized by the current correlation coefficient IR; Wherein, the calculation formula of the current correlation coefficient IR is: Among them, m and n are any two phases of the three-phase current of users A, B, and C respectively; I mk Indicates the current value of the user's m-phase current at the kth moment; Indicates the average current value of the user's m-phase current at 24 moments; Indicates the standard deviation of the user's m-phase current at 24 moments.

4. The abnormal power consumption detection method based on load curve correlation according to claim 1, characterized in that: In the above S1, the voltage characteristic is characterized by the voltage three-phase unbalance vb; The calculation formula of the voltage three-phase unbalance vb is: Among them, V A is the user's A phase voltage value; V B is the user's B phase voltage value; V C is the user C phase voltage value; Mean() is the average value function; abs() is the absolute value function; Mean(V A , V B , V C ) is the average value of the three-phase voltage of the user; abs(V i -Mean(V A ,V B ,V C )) is the absolute value of the deviation between the user's i-phase voltage and the average value of the user's three-phase voltage.

5. The abnormal power consumption detection method based on load curve correlation according to claim 1, characterized in that: In S1, the voltage characteristic can also be characterized by the rated voltage deviation vd: Among them, the rated voltage deviation vd of each phase voltage of the user p The calculation formula is: Where p is any phase voltage among the three-phase voltages of users A, B, and C; v u p Vr is the user's p-phase voltage value; p It is the user's p-phase rated voltage value, and the user's A-phase, B-phase, and C-phase rated voltage values ​​are the same.

6. The abnormal power consumption detection method based on load curve correlation according to claim 1, characterized in that: In S3, the correlation ρ x,y The calculation formula is: Among them, x and y represent the load values ​​of the characteristic load curve and the daily load curve at the same time; x i and i Represent the load values ​​of characteristic load and daily load at the i-th moment respectively; and Represent the mean load values ​​of characteristic load and daily load at all times respectively; cov(x,y) represents the covariance of the load values ​​of characteristic load curve and daily load curve; σ x and σ y They represent the standard deviation of the load values ​​of the characteristic load curve and the daily load curve respectively.

7. The abnormal power consumption detection method based on load curve correlation according to claim 6, characterized in that: In S4, the sliding window value method is used to calculate the correlation mean The specific calculation formula is: Among them, s is the logarithm of the load value at the corresponding moment; n is the window value size; It is the sum of the correlations between the characteristic load curves and the daily load curve calculated from the 1st window to the nth window; 1 / sn represents the averaging coefficient of the summed result.

8. The abnormal power consumption detection method based on load curve correlation according to claim 7, characterized in that: In the above S5, the calculation formula of the correlation combination model c is: Wherein, k represents the amplification factor.

9. The abnormal power consumption detection method based on load curve correlation according to claim 8, characterized in that: The calculation formula of the Euclidean distance d is: Among them, x n and n They represent the load values ​​at the same time on two days.

10. The abnormal power consumption detection method based on load curve correlation according to claim 8, characterized in that: In the above S6, the correlation adaptive threshold combination model c x The calculation formula is: Among them, ρ spe is the threshold coefficient; |s a -s x | is the absolute difference between the total daily electricity load and the total characteristic electricity load.

Images