阅读说明：本技术 一种基于滑动窗口技术的暂态负荷启停事件快速捕捉方法 (Method for rapidly capturing transient load start-stop event based on sliding window technology ) 是由 肖涛 陆春光 孙钢 李亦龙 王朝亮 蒋群 马立骏 葛岳军 于 2020-12-01 设计创作，主要内容包括：本发明公开了一种基于滑动窗口技术的暂态负荷启停事件快速捕捉方法,非侵入负荷辨识技术主要存在以下问题：复杂工况下的算法不稳定、电器大类细分不清晰、暂态负荷启停时间判断不准确。为此,本发明提出了一种基于滑动窗口技术的暂态负荷启停事件快速捕捉方法,通过计算相邻时间点或者时间段内负荷特征的变化并将其与设定的阈值相比较,当变化超过阈值时,即判定有事件发生,并根据负荷启停事件快速捕捉模型,分析负荷开关引起的系统暂态过程,结合特征数据库,实现快速负荷事件捕捉及识别过程,实现简单,可快速准确的获取检测结果。该方法的提出能够为暂态负荷识别技术的实现提供准确的识别时刻,具有良好的经济效益和实用价值。(The invention discloses a method for quickly capturing a transient load start-stop event based on a sliding window technology, and a non-invasive load identification technology mainly has the following problems: the algorithm under the complex working condition is unstable, the subdivision of the large class of the electric appliances is unclear, and the judgment of the transient load start-stop time is inaccurate. The invention provides a method for quickly capturing a transient load start-stop event based on a sliding window technology, which comprises the steps of calculating the change of load characteristics in adjacent time points or time periods and comparing the change with a set threshold value, judging that an event occurs when the change exceeds the threshold value, analyzing a system transient process caused by a load switch according to a load start-stop event quick capture model, combining a characteristic database, realizing the quick load event capture and identification process, and being simple to realize and capable of quickly and accurately obtaining a detection result. The method can provide accurate identification time for realizing the transient load identification technology, and has good economic benefit and practical value.)

1. A method for rapidly capturing a transient load start-stop event based on a sliding window technology is characterized by comprising the following steps:

1) acquiring data; the data includes: real-time voltage and current acquisition data, real-time active power data, active power minimum variation and oscilloscope sampling frequency;

2) the acquired data is brought into a transient load event rapid capturing model to be solved;

analyzing a system transient process caused by a load switch by using collected transient signals when different loads are put into and cut out and combining a transient load event rapid capture model, acquiring mean values in a sliding window and a calculation window, and calculating and comparing; when no load is cut out or put in, the two mean values are the same; when mutation occurs, the mean values are different, whether mutation occurs is judged according to the mean value condition, and when the change exceeds a set threshold value, a mutation event is judged to occur; and after the sudden change event is detected, extracting the current waveform characteristics of the power system at the moment, comparing the current waveform characteristics with a characteristic library, and quickly identifying the load type.

2. The method for rapidly capturing the transient load start-stop event based on the sliding window technology according to claim 1, wherein the method comprises the following steps: in step 2), the method comprises the following steps:

201) initializing a transient event minimum variation and environmental noise;

202) calculating a threshold h;

203) reading a data sequence;

204) initializing a sampling moment;

205) judging whether k is<n-n_m-n_dIs there a If yes, go to step 207); if not, entering the next step; in the formula: k is the number of collected signal data points, n is the number of collected signal points, n_mIs the average calculation window length, n_dThe number of sample data of an event detection window;

206) reading the next data length and returning to the step 205);

207) calculating a sequence mean value;

208) k meter_process<k+n_m(ii) a In the formula: k is a radical of_processIs the number of times an event is processed;

209) judging whether k is_process<n+n_m+n_d(ii) a If not, returning to step 207); if yes, entering the next step;

210) computingAnd d;positive and negative shift accumulation amounts at point k, respectively; in the formula: d is the detection delayTime;

211) judging whether g is present⁺>h or g^->h; if not, k_process＝k_process+1 and return to step 209); if yes, entering the next step;

212)k_start＝k_process-d; in the formula: k is a radical of_startIs the time at which the mutation occurred; when a mutational event occurs, according to k_startMarking the corresponding position of the oscillogram by using a mark and displaying alarm information;

213)k＝k_process+ 1; and returns to step 205).

3. The method for rapidly capturing the transient load start-stop event based on the sliding window technology according to claim 2, wherein the method comprises the following steps: in the step 2), the transient load event rapid capture model carries out variable point detection, and the variable point detection is that a time point or a change point of the change of the condition density can be rapidly detected after the condition density in the signal is changed; it is expressed by the following formula:

d＝inf(k:g_k(i₁,…,i_k)≥h)

in the formula: d is the detection delay time, h is a threshold value, g is a statistical function, and k is the detected time or point number of the change point.

4. The method for rapidly capturing the transient load start-stop event based on the sliding window technology as claimed in claim 3, wherein: the statistical function g is:

in the formula: s_k＝x_k-μ₀-σ，μ₀Is the average value before the occurrence of the change point, which is usually assumed to be known or estimable, and σ is the environmental noise when x is_kIf the amount of change is larger than σ, it is considered that the load level may be abruptly changed.

5. The method for rapidly capturing the transient load start-stop event based on the sliding window technology as claimed in claim 4, wherein: the calculation formula of the threshold value is as follows:

in the formula: t is t_sIs a sampling interval, n_maxΔ ch is the minimum amount of change of the sampled data, α, for the maximum allowable delay time₁，α₂Is a correction factor.

6. The method for rapidly capturing the transient load start-stop event based on the sliding window technology as claimed in claim 5, wherein: considering that the load level of the power system may be increased or decreased, the transient load event rapid capture model adopts a bilateral CUSUM algorithm, and the specific formula is as follows:

in the formula:positive and negative shift accumulations, mu, respectively₀The average value of the variable points can be estimated and assumed through sequence process information;

the model assumes that the mean value of the sequence is μ when no change occurs in the power system₀The bilateral CUSUM algorithm is:

1) when no change has occurred in the signal or signals,andclose to 0, mu₀＝0，g_k＝0；

2) When the load is switched in or out, the sequence changes, if the forward shift is generated, the mean value after the shift rises to mu₁＞μ₀+σ，Continuously increasing; if it is negative, the shifted μ becomes μ₁＜μ₀-σ，Continuously increasing;

3)moving d +1 backwards, otherwise d is 0, when d is greater than h, a signal mutation event is detected, and the occurrence time of the mutation event in the signal is estimated to be tau-k-d; in the same way, whenAnd moving d to d +1, otherwise d to 0, and when d is larger than h, detecting the signal mutation event and estimating the occurrence time of the mutation event in the signal to be tau to k-d.

Technical Field

The invention relates to the field of non-invasive load detection, in particular to a method for quickly capturing a transient load start-stop event based on a sliding window technology.

Background

Along with the continuous improvement of social ecological civilization consciousness, the attention of people to energy consumption is also continuously increased, and the requirement of a user on power supply quality is continuously improved, so that the traditional extensive management can not meet the requirement of a family user on intellectualization and lean electric energy supply, and the intelligent power utilization trend is inevitable. The monitoring and identification of the power utilization load is a key technology in the field of intelligent power utilization, and from the current theoretical research, detection results and application conditions, the non-invasive load identification technology mainly has the following problems: the algorithm under the complex working condition is unstable, the subdivision of the large class of the electric appliances is unclear, and the judgment of the transient load start-stop time is inaccurate.

Disclosure of Invention

The technical problem to be solved and the technical task to be solved by the invention are to perfect and improve the prior technical scheme, and provide a method for rapidly capturing the transient load start-stop event based on the sliding window technology so as to realize the purpose of accurately identifying the transient load start-stop moment. Therefore, the invention adopts the following technical scheme.

A method for rapidly capturing a transient load start-stop event based on a sliding window technology is characterized by comprising the following steps:

1) acquiring data; the data includes: real-time voltage and current acquisition data, real-time active power data, active power minimum variation and oscilloscope sampling frequency;

2) the acquired data is brought into a transient load event rapid capturing model to be solved;

analyzing a system transient process caused by a load switch by using collected transient signals when different loads are put into and cut out and combining a transient load event rapid capture model, acquiring mean values in a sliding window and a calculation window, and calculating and comparing; when no load is cut out or put in, the two mean values are the same; when mutation occurs, the mean values are different, whether mutation occurs is judged according to the mean value condition, and when the change exceeds a set threshold value, a mutation event is judged to occur; and after the sudden change event is detected, extracting the current waveform characteristics of the power system at the moment, comparing the current waveform characteristics with a characteristic library, and quickly identifying the load type. According to the technical scheme, the change of the load characteristics in the adjacent time points or time periods is calculated and compared with the set threshold, when the change exceeds the threshold, an event is judged to occur, the model is rapidly captured according to the load start-stop event, the system transient process caused by the load switch is analyzed, the rapid load event capturing and identifying process is realized by combining the characteristic database, the realization is simple, and the detection result can be rapidly and accurately obtained. The method can provide accurate identification time for realizing the transient load identification technology, and has good economic benefit and practical value.

As a preferable technical means: in step 2), the method comprises the following steps:

201) initializing a transient event minimum variation and environmental noise;

202) calculating a threshold h;

203) reading a data sequence;

204) initializing a sampling moment;

205) judging whether k is<n-n_m-n_dIs there a If yes, go to step 207); if not, entering the next step; k is the number of collected signal data points, n is the number of collected signal points, n_mIs the average calculation window length, n_dThe number of sample data of an event detection window;

206) reading the next data length and returning to the step 205);

207) calculating a sequence mean value;

208) k meter_process<k+n_m(ii) a In the formula: k is a radical of_processIs the number of times an event is processed;

209) judging whether k is_process<n+n_m+n_d(ii) a If not, returning to step 207); if yes, entering the next step;

210) computingAnd d;positive and negative shift accumulation amounts at point k, respectively; in the formula: d is a detection delay time;

211) judging whether g is present⁺>h or g^->h; if not, k_process＝k_process+1 and return to step 209); if yes, entering the next step;

212)k_start＝k_process-d; in the formula: k is a radical of_startIs the time at which the mutation occurred; when a mutational event occurs, according to k_startMarking the corresponding position of the oscillogram by using a mark and displaying alarm information;

213)k＝k_process+ 1; and returns to step 205).

As a preferable technical means: in the step 2), the transient load event rapid capture model carries out variable point detection, and the variable point detection is that a time point or a change point of the change of the condition density can be rapidly detected after the condition density in the signal is changed; it is expressed by the following formula:

d＝inf(k:g_k(i₁,...,i_k)≥h)

in the formula: d is the detection delay time, h is a threshold value, g is a statistical function, and k is the detected time or point number of the change point.

The sudden event may be load switching, through k_startThe more obvious mark can be used for displaying and alarming in the oscillogram;

as a preferable technical means: the statistical function g is:

in the formula: s_k＝x_k-μ₀-σ，μ₀Is the average value before the occurrence of the change point, which is usually assumed to be known or estimable, and σ is the environmental noise when x is_kIf the amount of change is larger than σ, it is considered that the load level may be abruptly changed.

As a preferable technical means: the calculation formula of the threshold value is as follows:

in the formula: t is t_sIs a sampling interval, n_maxΔ ch is the minimum amount of change of the sampled data, α, for the maximum allowable delay time₁，α₂Is a correction factor.

As a preferable technical means: considering that the load level of the power system may be increased or decreased, the transient load event rapid capture model adopts a bilateral CUSUM algorithm, and the specific formula is as follows:

in the formula:positive and negative shift accumulations, mu, respectively₀The average value of the variable points can be estimated and assumed through sequence process information;

the model assumes that the mean value of the sequence is μ when no change occurs in the power system₀The bilateral CUSUM algorithm is:

1) when no change has occurred in the signal or signals,andclose to 0, mu₀＝0，g_k＝0；

2) When the load is switched in or out, the sequence changes, if the forward shift is generated, the mean value after the shift rises to mu₁＞μ₀+σ，Continuously increasing; if it is negative, the shifted μ becomes μ₁＜μ₀-σ，Continuously increasing;

3)moving d +1 backwards, otherwise d is 0, when d is greater than h, a signal mutation event is detected, and the occurrence time of the mutation event in the signal is estimated to be tau-k-d; in the same way, whenAnd moving d to d +1, otherwise d to 0, and when d is larger than h, detecting the signal mutation event and estimating the occurrence time of the mutation event in the signal to be tau to k-d.

Has the advantages that: according to the technical scheme, the change of the load characteristics in the adjacent time points or time periods is calculated and compared with the set threshold, when the change exceeds the threshold, an event is judged to occur, the model is rapidly captured according to the load start-stop event, the system transient process caused by the load switch is analyzed, the rapid load event capturing and identifying process is realized by combining the characteristic database, the realization is simple, and the detection result can be rapidly and accurately obtained. The method can provide accurate identification time for realizing the transient load identification technology, and has good economic benefit and practical value.

Drawings

FIG. 1 is a flow chart of the present invention.

Detailed Description

The technical scheme of the invention is further explained in detail by combining the drawings in the specification.

The invention comprises the following steps:

s1, acquiring data; the data includes: real-time voltage and current acquisition data, real-time active power data, active power minimum variation and oscilloscope sampling frequency;

s2, bringing the acquired data into a transient load event rapid capture model for solving;

analyzing a system transient process caused by a load switch by using collected transient signals when different loads are put into and cut out and combining a transient load event rapid capture model, acquiring mean values in a sliding window and a calculation window, and calculating and comparing; when no load is cut out or put in, the two mean values are the same; when mutation occurs, the mean values are different, whether mutation occurs is judged according to the mean value condition, and when the change exceeds a set threshold value, a mutation event is judged to occur; and after the sudden change event is detected, extracting the current waveform characteristics of the power system at the moment, comparing the current waveform characteristics with a characteristic library, and quickly identifying the load type.

The specific step of S2 is as shown in fig. 1:

s201, initializing the minimum variation of the transient event and the environmental noise;

s202, calculating a threshold h;

s203, reading a data sequence;

s204, initializing sampling time;

s205, judging whether k is<n-n_m-n_dIs there a If yes, go to step 207); if not, entering the next step; k is the number of collected signal data points, n is the number of collected signal points, n_mIs the average calculation window length, n_dThe number of sample data of an event detection window;

s206, reading the next data length and returning to the step S205;

s207, calculating a sequence mean value;

s208, counting k_process<k+n_m(ii) a In the formula: k is a radical of_processIs the number of times an event is processed;

s209, judging whether k is_process<n+n_m+n_d(ii) a If not, returning to the step S207; if yes, entering the next step;

s210, calculatingAnd d;positive and negative shift accumulation amounts at point k, respectively;

s211, judging whether g is present⁺>h or g^->h; if not, k_process＝k_process+1, and return to step S209; if yes, entering the next step;

S212:k_start＝k_process-d; in the formula: k is a radical of_startIs the time at which the mutation occurred; when sudden change of state occurs, such as load switching, pass k_startCan display alarm with more obvious mark in the oscillogram);

S213:k＝k_process+ 1; and returns to step S205.

The establishment of the transient load event rapid capture model is explained in one step as follows:

the event detection problem can be summarized as a variable point detection problem, when the active power of the power system changes suddenly at a certain moment, and after sampling data (voltage, current, active power, reactive power and the like) are given, the transient process of the system caused by a load switch and the like can be accurately detected, in order to better understand a load start-stop event rapid detection model, a mathematical model of the variable point problem is given to further understand event detection, and the variable point model is as follows:

in the formula: p is a radical of₀,…,p_nAnd N represents a model with a parameter of a vector theta for an active power observed value of the power system.

In the formula: i is a current signal sequence, p_θIs the conditional density of the sequence component, A_kAnd B_kD is the variation of the mean, τ is the time at which the change occurs, M is the indicator function, μ is the mean of I.

The essence of the change point detection is that the time point or change point of the change of the condition density can be rapidly detected after the condition density in the signal is changed. This can be expressed by the following formula:

d＝inf(k:g_k(i₁,...,i_k)≥h)

in the formula: d is the moment or change point of density change, h is the threshold, g is the statistical function, k is the detected moment or point of change point.

Defining a statistical function g in a non-parametric sliding window based bilateral accumulation and transient event detection algorithm (CUSUM) as follows:

in the formula: s_k＝x_k-μ₀-σ，μ₀Is the average value before the occurrence of the change point, which is usually assumed to be known or estimable, and σ is the environmental noise when x is_kIf the amount of change is larger than σ, it is considered that the load level may be abruptly changed.

Recording the detection delay time as d, defining a threshold formula as follows:

in the formula: t is t_sIs the sampling interval, d_maxΔ ch is the minimum amount of change of the sampled data, α, for the maximum allowable delay time₁，α₂Is a correction factor.

Considering that the load level of the power system may be increased or decreased, a bilateral CUSUM algorithm needs to be adopted, and the specific formula is as follows:

in the formula: g⁺,g^-Positive and negative shift accumulations, mu, respectively₀Which is the average of the change points, can be estimated and assumed by the sequence process information.

The model assumes that the mean value of the sequence is μ when no change occurs in the power system₀The detailed usage method of the bilateral CUSUM algorithm is as follows:

1) when no change has occurred in the signal or signals,andclose to 0, mu₀＝0，g_k＝0；

2) When the load is switched in or out, the sequence changes, if the forward shift is generated, the mean value after the shift rises to mu₁＞μ₀+σ，Continuously increasing; if it is negative, the shifted μ becomes μ₁＜μ₀-σ，Continuously increasing;

3)moving d +1 backwards, otherwise d is 0, when d is greater than h, a signal mutation event is detected, and the occurrence time of the mutation event in the signal is estimated to be tau-k-d; in the same way, whenAnd moving d to d +1, otherwise d to 0, and when d is larger than h, detecting the signal mutation event and estimating the occurrence time of the mutation event in the signal to be tau to k-d.

The method for rapidly capturing the transient load start-stop event based on the sliding window technology shown in fig. 1 is a specific embodiment of the present invention, already embodies the substantial features and the progress of the present invention, and can be modified equivalently according to the practical use requirements and under the teaching of the present invention, and is within the protection scope of the present invention.