Intelligent terminal for power load resource management based on storage and computing integration

Abstract

The invention discloses an intelligent terminal for power load resource management based on storage and computing integration, and relates to the field of power load resource management. A prediction model is constructed by collecting historical operation data of a distributed energy system to be managed, and a time series of operation data prediction values ​​and sampling values ​​are constructed respectively, and each data point in the series is subjected to difference processing to obtain a difference data series, and the difference data series is decomposed and reconstructed by using wavelet transform, and a Markov chain model of operation data sampling values ​​is constructed by using historical data, and a Markov chain model of operation data sampling values ​​and an observation probability matrix of operation data sampling values ​​are constructed by using historical data, and the probability that the state of the operation data sampling value is 1 at a moment is calculated, and the judgment functions when the state of the operation data sampling value is =1 and =0 are defined as and respectively, and the function values ​​of the judgment functions and are used to judge the abnormal situation of the operation data sampling value at a moment.

Description

Technical Field

[0001] The present invention relates to the field of power load resource management, and in particular to an intelligent terminal for power load resource management based on storage and computing integration. Background Art

[0002] With the continuous construction of new power systems, source-load interaction has become an important way of power grid operation, and various types of loads, distributed energy and energy storage need to be effectively collected and coordinated. As an important part of power demand side management, load management is directly related to the benefits of power supply companies on the one hand, and has a direct impact on the safety of power equipment on the load side and the working status of equipment. Power load management plays an important role in improving the load situation of the power grid and balancing the working status of the power load. It is the main means to improve the safety of power grid operation and ensure the economy of the power grid. Through the load management system, the power supply unit can comprehensively manage the load characteristics and load power demand data of the power supply object, thereby laying a data foundation for the formulation of subsequent power system control plans.

[0003] The power load management system mainly includes the main station, load management terminal and communication channel. Due to the high complexity of the power system itself, the load pressure of the power grid during peak power consumption is relatively large, and the possibility of failure of the power load management system will be greatly increased. Terminal failure refers to one of the common failures when the power load management system cannot be put online normally. For distributed energy systems, the premise of the power supply subject using the power load management system for management is to analyze the data fed back by the terminal. A large number of terminal devices are the guarantee for the operation of the power load management system. In addition to the inability of the system to go online normally, the manifestation of terminal failure also includes the inability of the terminal device to accurately process data or damage to the terminal device hardware. From the perspective of the impact of the failure, once this failure occurs, the power supply subject will be unable to grasp the power consumption status of the power user, and the lack of information related to the power consumption status will cause the prediction and analysis function of the power load management system to fail to function.

[0004] In the actual terminal fault handling, under the influence of system complexity and external factors, although the surface fault seems to be handled, the user power consumption data collected by the terminal will be very different from the normal data. In this state, terminal fault troubleshooting and processing must be carried out by manual data adjustment and terminal adaptation. Although the difficulty of terminal fault handling is relatively low, the complexity of the processing process is relatively high. At present, the terminal equipment of the existing distributed energy system does not have the ability to analyze and calculate data, and cannot realize fault self-detection. It is often necessary to manually inspect and maintain the terminal equipment regularly, which will affect the overall operation efficiency of the system during operation and maintenance. To this end, we propose an intelligent terminal for power load resource management based on storage and computing integration. Summary of the invention

[0005] The main purpose of the present invention is to provide an intelligent terminal for power load resource management based on storage and computing integration, which can effectively solve the problems in the background technology.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] Intelligent terminal for power load resource management based on storage and computing integration, including:

[0008] A data collection module for collecting operating data of the distributed energy system to be managed;

[0009] A data analysis module for analyzing the operation data and obtaining a predicted value of the operation data;

[0010] A fault self-checking module for determining whether an operation data sampling value is abnormal data according to the acquired operation data prediction value;

[0011] An early warning module used to issue a fault alarm when the operating data sampling value is determined to be abnormal data.

[0012] The process of obtaining the predicted value of the running data includes the following steps:

[0013] Constructing a time series of the running data sampling values ,in, Expressed as Class running data Sample values;

[0014] Time series Data points in As real-time operation data, a prediction model is constructed using the historical operation data sampling values ​​in the time series, and the operation data prediction value is obtained through the prediction model , where the expression of the prediction model is:

[0015]

[0016] In the formula, is a constant coefficient, and ∈(0,1); , , are all positive integers, and ∈(1, ).

[0017] The process of determining whether the running data sampling value is abnormal data comprises the following steps:

[0018] According to the sampling values ​​of historical operation data, the time series are calculated using the constructed prediction model The predicted value of each data point in is used as the element point to construct the time series of the running data prediction value, which is recorded as , Expressed as Class running data The predicted value of the data point;

[0019] The timing sequence With timing sequence The data points in the equation are processed by difference to obtain the difference data sequence, which is recorded as ,in, , determine the number of wavelet decomposition layers , wavelet transform is used to decompose and reconstruct the difference data sequence. The specific process is as follows:

[0020] The steps of wavelet decomposition are as follows:

[0021] 1) Select wavelet basis function. Wavelet basis function is the basis of wavelet transform. Different wavelet basis functions are suitable for different types of data. Commonly used wavelet basis functions include Danbechics wavelet, Haar wavelet, etc. Selecting a suitable wavelet basis function can better extract the characteristics of the data.

[0022] 2) Perform multi-scale decomposition. Perform multi-scale decomposition on time series data to obtain wavelet coefficients of different scales. Multi-scale decomposition can be achieved through continuous wavelet transform or discrete wavelet transform. Continuous wavelet transform is applicable to continuous signals, while discrete wavelet transform is applicable to discrete signals.

[0023] 3) Select the number of decomposition layers N. Selecting the appropriate number of decomposition layers can level the resolution of time and frequency. The more decomposition layers there are, the higher the time resolution and the lower the frequency resolution; the fewer decomposition layers there are, the lower the time resolution and the higher the frequency resolution. Select the appropriate number of decomposition layers according to the specific situation.

[0024] Perform wavelet reconstruction. Wavelet reconstruction is the process of reconstructing wavelet coefficients into original data. The steps are as follows;

[0025] 1) Select the number of reconstruction layers. According to the wavelet coefficients and decomposition number obtained by wavelet decomposition, select the appropriate number of reconstruction layers. The number of reconstruction layers should be equal to the number of decomposition layers to ensure the integrity of the data;

[0026] 2) Perform wavelet reconstruction, use the selected wavelet basis function and reconstruction layer number, perform inverse wavelet operation on the wavelet coefficients to obtain the reconstructed data.

[0027] Define the running data sampling value state set as When the running data sampling value is normal data, the state value is 1, and when the running data sampling value is abnormal data, the state value is 0. , The state of the running data sampling value at the moment is , ∈ , Indicates that the running data sampling value is determined by the status Transfer to state The probability value of is constructed based on the definition content and historical data to build a Markov chain model of the running data sampling value, then the state transfer matrix is , where the state transition probabilities in the state transition matrix include , , , Four categories;

[0028] Construct the observation probability matrix of running data sampling values ,in, =0,1; when =0, expressed as The probability that the data sampling value is 0 at the moment; when =1, it is expressed as The probability that the state of the data sampling value at the moment is 1;

[0029] Calculate according to the following formula The probability that the state of the data sampling value at the moment is 1 , the calculation formula is: , where Expressed as Wavelet transform coefficients of the time difference data sequence; It is expressed as the average value of the wavelet coefficients when the running data sampling value is normal data;

[0030] Define the running data sampling value status separately =1 and = 0 when the judgment function is and , using the decision function , The function value pair The abnormal situation of the data sampling value is judged at all times, where the calculation formula of the judgment function value is as follows: = × ; = × (1- ).

[0031] Using the decision function , The function value pair The principles for judging abnormal conditions of the momentary running data sampling values ​​are as follows:

[0032] when ≥ hour, =1, the running data sampling value is judged to be normal data;

[0033] when < hour, =0, the running data sampling value is judged to be abnormal data.

[0034] State transition matrix The state transition probability in , , , The calculation formulas are:

[0035]

[0036] In the formula, The frequency of the transition from abnormal state to abnormal state in the historical statistical operation data sampling values; The frequency of the abnormal state being transferred to the normal state in the historical statistical operation data sampling values; The frequency of the transition from normal state to abnormal state in the historical statistical operation data sampling values; It is the frequency of transition from normal state to abnormal state in the historical statistical operation data sampling values.

[0037] The intelligent terminal further includes a data output module for correcting abnormal data of the operation data sampling value, wherein the abnormal data correction process includes the following steps:

[0038] Step S1: Obtain abnormal data in the data sequence ,in, ∈ ;

[0039] Step S2: Extract the timing sequence of the running data sampling value In the case of abnormal data points Before Normal data are recorded in sequence according to their sampling order. , ,..., ;

[0040] Step S3: Extract Normal data is brought into the prediction model, and abnormal data is calculated using the prediction model The predicted value at ;

[0041] Step S4: Get the predicted value Whether it is abnormal data is determined, and the predicted value is determined When the data is normal, the predicted value is obtained As abnormal data The correction value is output;

[0042] Step S5: When the predicted value is determined When the data is abnormal, the number of normal data extracted in step S2 is increased to + Then, repeat steps S3 to S5 until the predicted value is obtained. is normal data, where is an integer greater than or equal to 1.

[0043] Preferably, it also includes a memory, a processor, and a computer program for an intelligent terminal for power load resource management based on storage and computing integration, which is stored in the memory and can be run on the processor.

[0044] The present invention has the following beneficial effects:

[0045] Compared with the prior art, the prediction model is constructed by collecting the historical operation data of the distributed energy system to be managed, respectively constructing the time series of the operation data prediction value and the sampling value, and performing difference processing on each data point in the sequence to obtain the difference data sequence, using wavelet transform to decompose and reconstruct the difference data sequence, using historical data to construct the Markov chain model of the operation data sampling value, using historical data to construct the Markov chain model of the operation data sampling value and the observation probability matrix of the operation data sampling value, and calculating The probability that the state of the data sampling value at the moment is 1 , respectively define the running data sampling value status =1 and = 0 when the judgment function is and , using the decision function , The function value pair It can judge the abnormal situation of the operating data sampling value at any time, analyze and calculate the collected operating data of the distributed energy system, timely discover the fault data points in the operating data, and realize the intelligent workflow of data collection and fault self-detection of terminal equipment, thereby improving the overall operation efficiency of the power load resource management system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1This is an implementation flow chart of the intelligent terminal for power load resource management based on storage and computing integration of the present invention;

[0047] Figure 2 This is a structural block diagram of the intelligent terminal for power load resource management based on storage and computing integration of the present invention. DETAILED DESCRIPTION

[0048] The present invention will be further described below in conjunction with specific implementation methods, wherein the accompanying drawings are only used for exemplary descriptions and represent only schematic diagrams rather than actual drawings, and should not be understood as limiting the present invention. In order to better illustrate the specific implementation methods of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product.

[0049] Intelligent terminal for power load resource management based on storage and computing integration, including:

[0050] A data collection module for collecting operating data of the distributed energy system to be managed;

[0051] A data analysis module for analyzing operation data and obtaining predicted values ​​of operation data;

[0052] A fault self-checking module for determining whether a sampling value of operating data is abnormal data according to the acquired predicted value of operating data;

[0053] An early warning module used to issue a fault alarm when the operating data sampling value is determined to be abnormal data.

[0054] The workflow of the intelligent terminal includes the following steps:

[0055] Step 1: Collect the operating data of the distributed energy system to be managed and construct a time series of operating data sampling values ,in, Expressed as Class running data Sample values;

[0056] Step 2: Sequence in time Data points in As real-time operation data, the prediction model is constructed using the historical operation data sampling values ​​in the time series, and the operation data prediction value is obtained through the prediction model. , where the expression of the prediction model is:

[0057]

[0058] In the formula, is a constant coefficient, and ∈(0,1); , , are all positive integers, and ∈(1, ).

[0059] Step 3: Based on the historical operation data sampling values, use the constructed prediction model to calculate the time series The predicted value of each data point in is used as the element point to construct the time series of the running data prediction value, which is recorded as , Expressed as Class running data The predicted value of the data point;

[0060] Step 4: Sequence the timing With timing sequence The data points in the equation are processed by difference to obtain the difference data sequence, which is recorded as ,in, , determine the number of wavelet decomposition layers , wavelet transform is used to decompose and reconstruct the difference data sequence. The specific process is as follows:

[0061] The steps of wavelet decomposition are as follows:

[0062] 1) Select wavelet basis function. Wavelet basis function is the basis of wavelet transform. Different wavelet basis functions are suitable for different types of data. Commonly used wavelet basis functions include Danbechics wavelet, Haar wavelet, etc. Selecting a suitable wavelet basis function can better extract the characteristics of the data.

[0063] 2) Perform multi-scale decomposition. Perform multi-scale decomposition on time series data to obtain wavelet coefficients of different scales. Multi-scale decomposition can be achieved through continuous wavelet transform or discrete wavelet transform. Continuous wavelet transform is applicable to continuous signals, while discrete wavelet transform is applicable to discrete signals.

[0064] 3) Select the number of decomposition layers N. Selecting the appropriate number of decomposition layers can level the resolution of time and frequency. The more decomposition layers there are, the higher the time resolution and the lower the frequency resolution; the fewer decomposition layers there are, the lower the time resolution and the higher the frequency resolution. Select the appropriate number of decomposition layers according to the specific situation.

[0065] Perform wavelet reconstruction. Wavelet reconstruction is the process of reconstructing wavelet coefficients into original data. The steps are as follows;

[0066] 1) Select the number of reconstruction layers. According to the wavelet coefficients and decomposition number obtained by wavelet decomposition, select the appropriate number of reconstruction layers. The number of reconstruction layers should be equal to the number of decomposition layers to ensure the integrity of the data;

[0067] 2) Perform wavelet reconstruction, use the selected wavelet basis function and reconstruction layer number, perform inverse wavelet operation on the wavelet coefficients to obtain the reconstructed data.

[0068] Step 4: Define the running data sampling value state set as When the running data sampling value is normal data, the state value is 1, and when the running data sampling value is abnormal data, the state value is 0. , The state of the running data sampling value at the moment is , ∈ , Indicates that the running data sampling value is determined by the status Transfer to state The probability value of is constructed based on the definition content and historical data to build a Markov chain model of the running data sampling value, then the state transfer matrix is , where the state transition probability in the state transition matrix includes , , , There are four categories, and the calculation formulas are:

[0069]

[0070] In the formula, The frequency of the transition from abnormal state to abnormal state in the historical statistical operation data sampling values; The frequency of the abnormal state being transferred to the normal state in the historical statistical operation data sampling values; The frequency of the transition from normal state to abnormal state in the historical statistical operation data sampling values; The frequency of the transition from the normal state to the normal state in the historical statistical operation data sampling values;

[0071] Step 5: Construct the observation probability matrix of the running data sampling values ,in, =0,1; when =0, it is expressed as The probability that the data sampling value is 0 at the moment; when =1, it is expressed as The probability that the state of the data sampling value at the moment is 1;

[0072] Step 6: Calculate according to the following formula The probability that the state of the data sampling value at the moment is 1 , the calculation formula is: , where Expressed as Wavelet transform coefficients of the time difference data sequence; It is expressed as the average value of the wavelet coefficients when the running data sampling value is normal data;

[0073] Step 7: Define the running data sampling value status separately =1 and = 0 when the judgment function is and , using the decision function , The function value pair The abnormal situation of the data sampling value is judged at all times, where the calculation formula of the judgment function value is as follows: = × ; = × (1- ), the judgment principle is:

[0074] when ≥ hour, =1, the running data sampling value is judged to be normal data;

[0075] when < hour, =0, the running data sampling value is judged to be abnormal data.

[0076] State transition matrix The state transition probability in , , , of.

[0077] Step 8: Acquire abnormal data from the operating data sampling values, correct the abnormal data of the acquired operating data sampling values, and output the corrected values ​​of the abnormal data. The abnormal data correction process includes the following steps:

[0078] Step 81: Get the abnormal data in the data sequence ,in, ∈ ;

[0079] Step 82: Extract the time series of running data sampling values In the case of abnormal data points Before Normal data are recorded in sequence according to their sampling order. , ,..., ;

[0080] Step 83: Extract Normal data is brought into the prediction model, and the abnormal data is calculated using the prediction model. The predicted value at ;

[0081] Step 84: Obtain the predicted value Whether it is abnormal data is determined, and the predicted value is determined When the data is normal, the predicted value is obtained As abnormal data The correction value is output;

[0082] Step 85: When determining the predicted value When the data is abnormal, the number of normal data extracted in step S2 is increased to + Then, repeat steps S3 to S5 until the predicted value is obtained. is normal data, where is an integer greater than or equal to 1.

[0083] The intelligent terminal for power load resource management based on storage and computing integration of the present invention also includes a memory, a processor, and a computer program for the intelligent terminal for power load resource management based on storage and computing integration stored in the memory and executable on the processor.

[0084] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An intelligent terminal for power load resource management based on storage and computing integration, characterized in that: include: A data collection module for collecting operating data of the distributed energy system to be managed; A data analysis module for analyzing the operation data and obtaining a predicted value of the operation data; A fault self-checking module for determining whether an operation data sampling value is abnormal data according to the acquired operation data prediction value; An early warning module used to issue a fault alarm when the running data sampling value is determined to be abnormal data. The process of obtaining the predicted value of the running data includes the following steps: Construct the time series L of the running data sampling value = (x i1 ,x i2 ,...,x i(j-1) ,x ij ), where x ij It is represented as the jth sampling value of the i-th type of running data; Take the data point x in the time series L ij As real-time operation data, a prediction model is constructed using the historical operation data sampling values ​​in the time series, and the operation data prediction value x is obtained through the prediction model. i(j+1) , where the expression of the prediction model is: Wherein, a is a constant coefficient, and a∈(0,1); n, m, M are all positive integers, and M∈(1,j).

2. The intelligent terminal for power load resource management based on storage and computing integration according to claim 1 is characterized in that: The process of determining whether the running data sampling value is abnormal data comprises the following steps: According to the historical operation data sampling value, the prediction model is used to calculate the prediction value of each data point in the time series L, and the obtained prediction value is used as the element point to construct the time series of the operation data prediction value, which is recorded as L'=(x' i1 ,x' i2 ,...,x' i(j-1) ,x' ij ), x' ij It is represented as the predicted value of the jth data point of the i-th type of running data; Difference processing is performed between the time series L' and each data point in the time series L to obtain the difference data sequence, which is recorded as in, Determine the wavelet decomposition layer number N, and use wavelet transform to decompose and reconstruct the difference data sequence; Define the state set of running data sampling value as S. When the running data sampling value is normal data, the state value is 1. When the running data sampling value is abnormal data, the state value is 0. Then S = {0, 1}. The state of the running data sampling value at time t is s t , t∈j,a uv It represents the probability value of the running data sampling value transferring from state u to state v. According to the definition content and historical data, the Markov chain model of the running data sampling value is constructed, and the state transfer matrix is ​​A = (a uv ) 2×2 , where the state transition probabilities in the state transition matrix include a 00 、a 01 、a 10 、a 11 Four categories; Construct the observation probability matrix P of the running data sampling value = (p tk ) 1×2 , where k = 0, 1; when k = 0, it represents the probability that the state of the running data sampling value at time t is 0; when k = 1, it represents the probability that the state of the running data sampling value at time t is 1; The probability p of the running data sampling value being 1 at time t is calculated according to the following formula t1 , the calculation formula is: Where W(t,f) represents the wavelet transform coefficient of the difference data sequence at time t; W a It is expressed as the average value of the wavelet coefficients when the running data sampling value is normal data; Define the running data sampling value status s respectively t =1 and s t = 0 when the judgment function is and Using the decision function The function value of is used to judge the abnormality of the running data sampling value at time t, where the calculation formula of the judgment function value is as follows:

3. The intelligent terminal for power load resource management based on storage and computing integration according to claim 2 is characterized in that: Using the decision function The principle for judging the abnormality of the running data sampling value at time t is as follows: when When t =1, the running data sampling value is judged to be normal data; when When t =0, the running data sampling value is judged to be abnormal data.

4. The intelligent terminal for power load resource management based on storage and computing integration according to claim 2 is characterized in that: State transfer matrix A = (a uv ) 2×2 The state transition probability a in 00 、a 01 、a 10 、a 11 The calculation formulas are: In the formula, n(a 00 ) is the frequency of the transition from abnormal state to abnormal state in the historical statistical operation data sampling value; n(a 01 ) is the frequency of the transition from abnormal state to normal state in the historical statistical operation data sampling value; n(a 10 ) is the frequency of the transition from normal state to abnormal state in the historical statistical operation data sampling value; n(a 11 ) is the frequency of transition from normal state to abnormal state in the historical statistical operation data sampling values.

5. The intelligent terminal for power load resource management based on storage and computing integration according to claim 3 or 4, characterized in that: The intelligent terminal further includes a data output module for correcting abnormal data of the operation data sampling value, wherein the abnormal data correction process includes the following steps: Step S1: Get the abnormal data x in the data sequence iq , where q∈j; Step S2: Extract the time series L of the running data sampling value = (x i1 ,x i2 ,...,x i(j-1) ,x ij ), the abnormal data point x iq The first r normal data are recorded as x according to their sampling order. i1 、x i2 , ..., x ir ; Step S3: Bring the extracted r normal data into the prediction model, and use the prediction model to calculate the abnormal data x iq The predicted value x at iq '; Step S4: Get the predicted value x iq 'Determine whether it is abnormal data. When the predicted value x is determined iq 'When it is normal data, the predicted value x is obtained iq 'As abnormal data x iq The correction value is output; Step S5: When the predicted value x is determined iq 'When it is abnormal data, increase the number of normal data extracted in step S2 to r+λ, and repeat steps S3-S5 until the predicted value x is obtained. iq ' is normal data, where λ is an integer greater than or equal to 1.

6. The intelligent terminal for power load resource management based on storage and computing integration according to claim 1 is characterized in that: It also includes a memory, a processor, and a computer program for an intelligent terminal for power load resource management based on storage and computing integration, which is stored in the memory and can be run on the processor.

Images