Load Decomposition Method and Device for Industrial Facilities Based on Mixed-Integer Programming

Abstract

The present application relates to a load decomposition method and device for industrial facilities based on mixed-integer programming. The method includes: obtaining an equipment asset list of the industrial facilities to be decomposed, classifying industrial equipment into two categories, one being dynamic load equipment with continuously adjustable power and the other being stable load equipment with power switching between fixed states; using a certain scale of equipment-level power consumption data as a training set, processing the load curves of appropriate dynamic load equipment using digital filtering methods, and expanding its training data set by utilizing the time-invariance of the frequently fluctuating pulse part in the load; constructing a mixed-integer programming model for decomposition using equipment information and training data; introducing model correction terms or constraints by utilizing physical limitations such as the industrial process dependencies of some equipment; and solving the programming problem to obtain the decomposition result. This method can simultaneously consider the load characteristics of dynamic load equipment and stable load equipment, and can perform non-intrusive load decomposition with relatively high accuracy in industrial scenarios.

Description

Technical Field

[0001] The present application relates to the technical field of non-intrusive load decomposition, and particularly relates to a load decomposition method and device for industrial facilities based on mixed integer programming. Background Art

[0002] With the rapid development of smart grids and urbanization, the flow of massive amounts of information has become an inevitable trend. Power sensors have been widely used to monitor the load information of various electricity users. These sensors have collected a large amount of load data, laying the foundation for the development of an Internet of Things-based smart grid. On the other hand, renewable energy has received increasing attention. The popularization of a large amount of renewable energy and the gradual phasing out of coal-fired power plants have brought huge challenges to grid operation. In order to better promote energy conservation and facilitate two-way communication between the supply and demand sides of the grid, the feedback of device-level load data of electricity users has become very important, and this feedback can be achieved through two methods: invasive and non-invasive. Invasive monitoring usually involves obstacles such as sensor installation, economic costs, and data privacy. In contrast, non-intrusive load monitoring provides consumers with the lowest-cost and lowest-sensor-installation-effort solution and is considered a more promising option.

[0003] Due to the huge differences in the power consumption scale and electrical equipment types of different electricity users, load decomposition is usually divided into three types: load decomposition for residential, commercial, and industrial settings. Although the application of load decomposition in commercial and industrial scenarios has more significant potential profits (the electricity consumption of a factory may be equivalent to that of hundreds of residences), current research on load decomposition mainly focuses on residential facilities. Different from the residential scenario, there are a large number of dynamic load devices in the industrial scenario, and the electricity consumption characteristics are more complex, which brings great difficulties to load decomposition.

[0004] In the aspect of industrial NILM (non-intrusive load monitoring), in the related technology, the load decomposition in large industrial buildings was first studied. By collecting power data from a large industrial cold storage and comparing and analyzing the collected data with typical residential data, and using two benchmark models, namely CO (Combinationrial Optimization, combination optimization algorithm) and FHMM (Factorial Hidden Markov Models, multiple hidden Markov models), the industrial load was decomposed, and improvement measures for targeted training and sub-metering were proposed.

[0005] In the related technology, a decomposition model based on a deep neural network was also established, achieving better results than the classical FHMM.

[0006] In the related art, the active power and reactive power data in FHMM are also used for industrial load decomposition, which improves the results to a certain extent. At present, most of the research on industrial NILM only slightly modifies the residential NILM model to handle industrial scenarios, without designing a suitable model to utilize the characteristics of industrial data.

[0007] Industrial loads have characteristics such as a large proportion of dynamic loads and pipeline dependencies that are rare in civil loads. These characteristics will greatly affect the effects of traditional optimization model-based and FHMM model-based methods. Related research using deep learning methods shows that in industrial scenarios, deep learning methods can still achieve good results. At the same time, these models highly depend on a large amount of training data, but industrial data is often very sensitive, which becomes a major obstacle to the application of the models.

[0008] There are mainly the following two solution techniques in the related art:

[0009] (1) k-means clustering technique: This technique can divide the data set into k classes according to the distance through iteration when the number of clustering centers k is given.

[0010] (2) Mixed-integer programming problem solving technique: Solve a class of problems in mathematical programming where the independent variables are integers. Generally, an algorithm based on branch and bound is used to solve it. The specific approach of branch and bound is as follows:

[0011] Delete all integer constraints from the original mixed-integer programming to obtain the relaxation of the original programming. Generally, this relaxed programming is considered to be efficiently solvable. If the solution of the relaxed problem exactly satisfies all the restrictions of the integer constraints, then this solution is the optimal solution of the original mixed-integer programming, and the operation terminates. If the solution does not satisfy all the integer restrictions (most cases are like this), constraints need to be introduced to exclude this solution from the feasible region. At this time, the original feasible region is divided into two regions, and the original problem can be divided into two sub-problems to solve, which is called "branching". At this time, the original problem is replaced by two sub-problems with fewer integer variables.

[0012] The solution of the above-mentioned relaxed problem is the optimal solution under a larger feasible region. The solution of the original problem will definitely not be better than this solution. Therefore, the solution of the relaxed problem is defined as a "lower bound" of the original problem. In the case of branching, sub-problems will be continuously divided. It is possible to obtain the optimal solution that satisfies all the constraints of the sub-problems. This solution is only the optimal within the local feasible region of the original problem and is not necessarily the global optimal. Therefore, this solution is defined as an "upper bound" of the original problem. When calculating each branched sub-problem, if solving a sub-problem falls outside the currently set bounds, this branch can be deleted and no further consideration is needed. Therefore, the branch and bound method is an iterative algorithm. As the sub-problems are solved, the upper and lower bounds are continuously updated. When the algorithm meets the convergence condition, the numerical solution of the original problem can be obtained.

[0013] Currently, mature commercial solvers are generally used to solve such problems. These solvers generally adopt many algorithms to optimize the solving speed, such as the cutting plane method, etc. Summary of the Invention

[0014] The present application provides a load decomposition method, device, electronic device and storage medium for industrial facilities based on mixed integer programming, so as to solve the problem that existing load decomposition technologies are difficult to effectively process the complex equipment load identification in industrial scenarios under the limitations of insufficient industrial load data and difficult data acquisition, and can simultaneously consider the load characteristics of dynamic load equipment and stable load equipment, and can perform non-intrusive load decomposition with higher accuracy in industrial scenarios.

[0015] The first aspect of the embodiments of the present application provides a load decomposition for industrial facilities based on mixed integer programming, including the following steps:

[0016] Obtain the equipment list of the industrial equipment to be decomposed and the load curve to be decomposed, and obtain the equipment-level load curve of each equipment based on the equipment list and the load curve to be decomposed;

[0017] Divide the industrial equipment to be decomposed into dynamic load equipment and stable load equipment according to the power consumption characteristics, perform digital filtering on the equipment-level load curve of the dynamic load equipment, and use time invariance to expand the frequently fluctuating pulse part in the load to obtain the load curve of the processed dynamic load equipment;

[0018] Construct a mixed integer programming model based on the load curve of the processed dynamic load equipment and the load curve of the stable load equipment. After introducing the correction of industrial process constraints and equipment operation restrictions in the mixed integer programming model, solve to obtain the optimal values of the optimization variables; and

[0019] Based on the optimization problem model, reconstruct the decomposition result for the combined signal matrix of the processed dynamic load equipment, and reconstruct the decomposition result for the stable load equipment converted into a power state sequence based on the optimization problem model.

[0020] Optionally, the performing digital filtering on the equipment-level load curve of the dynamic load equipment and using time invariance to expand the frequently fluctuating pulse part in the load to obtain the processed dynamic load equipment includes:

[0021] First, use median filtering to remove noise spikes from the equipment-level load curve of the dynamic load equipment, then detect the large slope part in the equipment-level load curve of the dynamic load equipment and separate it, and finally decompose the load of each dynamic equipment into a smooth basic part and a violently fluctuating pulse part;

[0022] Generate a column vector with all elements being 0 for a preset time period, connect the column vector to the beginning of each basis vector to be augmented, and discard a sequence of the same length at the end of the basis vector to be augmented to obtain an augmented pulse lagged by one unit. Superimpose the augmented pulse on the smoothed basic part to obtain a new basis vector;

[0023] Based on the new basis vector and the severely fluctuating pulse part, obtain the processed dynamic load equipment.

[0024] Optionally, the step of dividing the industrial equipment to be decomposed into dynamic load equipment and stable load equipment according to the power consumption characteristics includes:

[0025] Model the stable load equipment using a combined optimization method:

[0026]

[0027]

[0028]

[0029] where, is the power of equipment n at time t, t is the time, T is the length of the time series for one optimization, n is the equipment number, b n，k (t) is a variable for determining the power state of equipment n, k is an integer in [1, K n , and K n is the number of states of equipment n; Model the dynamic load equipment using a linear combination of the known equipment-level load curves of each equipment:

[0030]

[0031] s.t.: A n ≥0;

[0032] where, is the power time series of equipment n, D n is the signal matrix of equipment n; A n is the activation coefficient matrix.

[0033] Optionally, the step of constructing a mixed-integer programming model based on the load curves of the processed dynamic load equipment and the load curves of the stable load equipment further includes:

[0034] Connect the signal matrices of each dynamic load equipment as D = [D 1 , D 2 ,..., D n ; Express the load model of the stable load equipment in matrix form:

[0035]

[0036] Take the dimension that is the same as the basis vector in the sum signal matrix, and we have:

[0037]

[0038] Construct the above-mentioned mixed-integer programming model:

[0039]

[0040] s.t.: A ≥ 0

[0041] Wherein, X is the load curve read by the total smart electricity meters of users, β is the regularization term coefficient, U is the column vector composed of 1s, Q is the signal grouping matrix, A is the activation coefficient matrix, N is the number of stable load devices, X n is the power state matrix, B n is the state selection matrix, and m is the length of the time series intercepted by one optimization.

[0042] Optionally, it is characterized in that the modification of introducing industrial process constraints and equipment operation limitations in the above-mentioned mixed-integer programming model includes:

[0043] The process constraint is:

[0044]

[0045] Limit the longest duration and the shortest duration of the stable load device. Among them, the longest duration W n，k Is constrained by a set of linear inequalities:

[0046]

[0047]

[0048] The shortest duration w n，k Is constrained by a set of linear inequalities:

[0049]

[0050]

[0051] Wherein, γ is the regularization term coefficient, and R is the process limitation matrix.

[0052] The second aspect embodiment of this application provides a load decomposition device for industrial facilities based on mixed-integer programming, including:

[0053] An acquisition module, configured to acquire an equipment list of an industrial equipment to be decomposed, a load curve to be decomposed, and obtain an equipment-level load curve of each equipment based on the equipment list and the load curve to be decomposed;

[0054] A processing module, configured to divide the industrial equipment to be decomposed into dynamic load equipment and stable load equipment according to power consumption characteristics, perform digital filtering on the equipment-level load curve of the dynamic load equipment, and utilize time invariance to expand the frequently fluctuating pulse part in the load to obtain a processed load curve of the dynamic load equipment;

[0055] A first calculation module, configured to construct a mixed integer programming model based on the processed load curve of the dynamic load equipment and the load curve of the stable load equipment, and after introducing corrections of industrial process constraints and equipment operation restrictions in the mixed integer programming model, solve to obtain the optimal value of the optimization variable; and

[0056] A second calculation module, configured to reconstruct a decomposition result for the combined signal matrix of the processed dynamic load equipment based on the optimization problem model, and reconstruct a decomposition result for converting the stable load equipment into a power state sequence based on the optimization problem model.

[0057] Optionally, the processing module is specifically configured to:

[0058] First, use median filtering to remove noise spikes from the equipment-level load curve of the dynamic load equipment, then detect and separate the large slope part in the equipment-level load curve of the dynamic load equipment, and finally decompose the load of each dynamic equipment into a smooth basic part and a violently fluctuating pulse part;

[0059] Generate a column vector with all elements being 0 in a preset time period, connect the column vector to the beginning of each base vector to be expanded, and discard the sequence of the same length at the end of the base vector to be expanded to obtain an expanded pulse lagging by one unit, and superimpose the expanded pulse on the smooth basic part to obtain a new base vector;

[0060] Based on the new base vector and the violently fluctuating pulse part, obtain the processed dynamic load equipment.

[0061] Optionally, the processing module is used for:

[0062] Model the stable load equipment by using a combinatorial optimization method:

[0063]

[0064]

[0065]

[0066] Among them, is the power of device n at time t, t is the time, T is the length of the time series for one optimization, n is the device number, and b n，k (t) is a variable for determining the power state where device n is located, and k is an integer in [1, K n , and K n is the number of states of device n;

[0067] Model the dynamic load devices using a linear combination of the known device-level load curves of each device:

[0068]

[0069] s.t.: A n ≥ 0;

[0070] Among them, is the power time series of device n, and D n is the signal matrix of device n; A n is the activation coefficient matrix.

[0071] Optionally, the first calculation module is further configured to:

[0072] Connect the signal matrices of each of the dynamic load devices as D = [D 1 , D 2 ,..., D n ; Express the load model of the stable load device in matrix form:

[0073]

[0074] Take the dimension that is the same as the basis vector in the sum signal matrix, and there is:

[0075]

[0076] Construct the mixed-integer programming model:

[0077]

[0078] s.t.: A ≥ 0

[0079] Among them, X is the load curve read by the user's total smart meter, β is the regularization term coefficient, U is a column vector composed of 1s, Q is the signal grouping matrix, A is the activation coefficient matrix, N is the number of stable load devices, and X n is the power state matrix, B n is the state selection matrix, and m is the length of the intercepted time series for one optimization.

[0080] Optionally, the first calculation module is further configured to:

[0081] The process constraints are as follows:

[0082]

[0083] The maximum and minimum durations of the stable load equipment are restricted, where the maximum duration is W n，k Constrained by a set of linear inequalities:

[0084]

[0085]

[0086] The minimum duration is w n，k Constrained by a set of linear inequalities:

[0087]

[0088]

[0089] where γ is the regularization term coefficient and R is the process constraint matrix.

[0090] The third aspect of the embodiments of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the program to implement the load decomposition method for industrial facilities based on mixed integer programming as described in the above embodiments.

[0091] The fourth aspect of the embodiments of the present application provides a computer-readable storage medium, on which a computer program is stored, and the program is executed by a processor to implement the load decomposition method for industrial facilities based on mixed integer programming as described above.

[0092] Thus, by obtaining the equipment asset list of the industrial facilities to be decomposed, classifying industrial equipment into two categories: one is dynamic load equipment with continuously adjustable power, and the other is stable load equipment with power switching between fixed states; using a certain scale of equipment-level power consumption data as the training set, processing the load curves of appropriate dynamic load equipment using digital filtering methods, and expanding its training data set using the time-invariance of the frequently fluctuating pulse parts in the load; constructing a mixed integer programming model for decomposition using equipment information and training data; introducing model correction terms or constraints using physical constraints such as the industrial process dependencies of some equipment; and solving the planning problem to obtain the decomposition result. Thus, the problem that existing load decomposition technologies are difficult to effectively handle the complex equipment load identification in industrial scenarios under the limitations of insufficient industrial load data and difficult data acquisition is solved, and the load characteristics of both dynamic load equipment and stable load equipment can be considered simultaneously, enabling non-intrusive load decomposition with relatively high accuracy in industrial scenarios.

[0093] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of embodiments in conjunction with the accompanying drawings, in which:

[0095] Figure 1 is a flowchart of a load decomposition method for industrial facilities based on mixed-integer programming according to an embodiment of the present application;

[0096] Figure 2 is a flowchart of a load decomposition method for industrial facilities based on mixed-integer programming according to an embodiment of the present application;

[0097] Figure 3 is a block diagram of a load decomposition device for industrial facilities based on mixed-integer programming according to an embodiment of the present application;

[0098] Figure 4 is a schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0099] Embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present application and should not be construed as limiting the present application.

[0100] The load decomposition method, device, electronic device and storage medium for industrial facilities based on mixed integer programming according to the embodiments of the present application will be described below with reference to the accompanying drawings. Aiming at the problem that the existing load decomposition technology mentioned in the above background technology is difficult to effectively handle the complex equipment load identification in the industrial scenario under the limitations of insufficient industrial load data and difficult data acquisition, the present application provides a load decomposition method for industrial facilities based on mixed integer programming. In this method, by obtaining the equipment asset list of the industrial facilities to be decomposed, the industrial equipment is classified into two categories: one is the dynamic load equipment with continuously adjustable power, and the other is the stable load equipment with power switching between fixed states; and a certain scale of equipment-level power consumption data is used as the training set. The load curve of the appropriate dynamic load equipment is processed by the digital filtering method, and the training data set is expanded by using the time invariance of the frequently fluctuating pulse part in the load; a mixed integer programming model for decomposition is constructed by using the equipment information and the training data; some physical limitations such as the industrial process dependence of the equipment are utilized to introduce model correction terms or constraints; the decomposition result is obtained by solving the programming problem. Thus, the problem that the existing load decomposition technology is difficult to effectively handle the complex equipment load identification in the industrial scenario under the limitations of insufficient industrial load data and difficult data acquisition is solved, and the load characteristics of both dynamic load equipment and stable load equipment can be considered simultaneously, and non-intrusive load decomposition with higher accuracy can be performed in the industrial scenario.

[0101] Specifically, Figure 1 FIG. is a schematic flow chart of a load decomposition method for industrial facilities based on mixed integer programming provided by an embodiment of the present application.

[0102] As Figure 1 shown, the load decomposition method for industrial facilities based on mixed integer programming includes the following steps:

[0103] In step S101, obtain the equipment list of the industrial equipment to be decomposed, the load curve to be decomposed, and obtain the equipment-level load curve of each equipment based on the equipment list and the load curve to be decomposed.

[0104] Among them, define the power reading sequence recorded by the intelligent meter of the user's low-voltage circuit in a day as the "total load curve", and record it as a column vector; define the power reading sequence collected by installing sensors separately for the electrical equipment in a day as the "equipment-level load curve", and record it as a column vector of the same dimension.

[0105] Specifically, obtain the equipment asset list of the industrial facilities to be decomposed; obtain the load curve to be decomposed from the intelligent meter of the industrial facilities to be decomposed, and obtain the equipment-level load curve of each equipment from the equipment manufacturer or the equipment load database.

[0106] In step S102, the industrial equipment to be decomposed is divided into dynamic load equipment and stable load equipment according to the power consumption characteristics. Digital filtering is performed on the equipment-level load curve of the dynamic load equipment, and the frequently fluctuating pulse part in the load is expanded using time invariance to obtain the load curve of the processed dynamic load equipment.

[0107] Optionally, in some embodiments, dividing the industrial equipment to be decomposed into dynamic load equipment and stable load equipment includes: modeling the stable load equipment using a combined optimization method:

[0108]

[0109]

[0110]

[0111] where, is the power of device n at time t, t is the time, T is the time series length of one optimization, n is the device number, and the power magnitudes of each state of device n are denoted as The power values of each state are determined using the k-means clustering technique, b n，k (t) is the variable for determining the power state where device n is located, k is an integer in [1, K n , and K n is the number of states of device n;

[0112] Model the dynamic load equipment using a linear combination of the known device-level load curves of each device:

[0113]

[0114] s.t.: A n ≥0;

[0115] where, is the power time series of device n, D n is the signal matrix of device n, and each column vector of it is the known device-level load curve of this device, called a basis vector; A n is the activation coefficient matrix, that is, the coefficient for linearly combining the basis vectors, and generally requires the elements of A to be non-negative.

[0116] Optionally, in some embodiments, digital filtering is performed on the device-level load curve of the dynamic load device, and the frequently fluctuating pulse part in the load is augmented using time invariance to obtain a processed dynamic load device, including: first using median filtering to remove noise spikes from the device-level load curve of the dynamic load device, then detecting and separating the large slope parts in the device-level load curve of the dynamic load device, and finally decomposing the load of each dynamic device into a smooth basic part and a violently fluctuating pulse part; generating a column vector with all elements being 0 for a preset time period, connecting the column vector to the beginning of each base vector to be augmented, and discarding the sequence of the same length at the end of the base vector to be augmented to obtain an augmented pulse lagged by one unit, and superimposing the augmented pulse on the smooth basic part to obtain a new base vector; obtaining the processed dynamic load device based on the new base vector and the violently fluctuating pulse part.

[0117] Specifically, industrial devices are classified according to their power consumption characteristics during operation, and the devices are divided into dynamic load devices and stable load devices; dynamic load devices generally include adjustable-speed motors, and their load curves change within a continuous range; the load curves of stable load devices switch between a series of fixed power states; these two types of devices are modeled in different ways. For the device-level load curve of the dynamic load device, the curve vector corresponding to the appropriate device is processed using digital filtering, and the time invariance of the frequently fluctuating pulse part in the load is used to construct a new realistic vector, thereby augmenting the training set; for the device-level load curve of the dynamic load device, digital filtering is first performed; during filtering, median filtering is first used to remove noise spikes, then the large slope parts in the curve are detected and separated, and a limiter is additionally added to assist in the identification of pulses. Finally, the load of each device is decomposed into a smooth basic part and a violently fluctuating pulse part. The pulse part is augmented using time invariance; a fixed time length is taken, and a column vector with all elements being 0 corresponding to this time length is generated. The column vector is connected to the beginning of each base vector to be augmented, and the sequence of the same length at the end of the base vector is discarded to obtain an augmented pulse lagged by one unit; the augmented pulse is superimposed on the basic part to obtain a new base vector for the device.

[0118] In step S103, a mixed-integer programming model is constructed based on the processed dynamic load device load curve and the stable load device load curve. After introducing the corrections of industrial process constraints and device operation limitations into the mixed-integer programming model, the optimal values of the optimization variables are obtained by solving.

[0119] Optionally, in some embodiments, constructing a mixed-integer programming model based on the processed dynamic load device load curve and the stable load device load curve further includes: connecting the signal matrices of each dynamic load device as D = [D 1 , D 2 ,..., D n; Express the load model of the stable load equipment in the form of a matrix:

[0120]

[0121] Take the dimension that is the same as the basis vector in the sum signal matrix, and we have:

[0122]

[0123] Construct a mixed-integer programming model:

[0124]

[0125] s.t.: A ≥ 0

[0126] Among them, X is the load curve read by the total smart meters of users, is the load to be decomposed, the second term is the regularization term of the basis vector grouping, β is the regularization coefficient, U is a column vector composed of 1s, Q is the signal grouping matrix, where the diagonal blocks of Q are 1s, and the dimension corresponds to the number of grouped basis vectors, and the elements in other areas are 0, A is the activation coefficient matrix, N is the number of stable load equipment units, X n is the power state matrix, B n is the state selection matrix, and m is the length of the time series intercepted in one optimization.

[0127] Optionally, in some embodiments, introduce corrections for industrial process constraints and equipment operation limitations in the mixed-integer programming model, including: the process constraint is to add an additional penalty term in the optimization objective function (mixed-integer programming model):

[0128]

[0129] Limit the longest duration and the shortest duration for the stable load equipment. Among them, the longest duration W n，k Use a set of linear inequalities to constrain:

[0130]

[0131]

[0132] The shortest duration w n，k Use a set of linear inequalities to constrain:

[0133]

[0134]

[0135] Among them, γ is the regularization term coefficient, and R is the process constraint matrix. Among them, R is composed of a series of identity matrices connected. For two devices with a process relationship, the corresponding blocks are a positive identity matrix and a negative identity matrix respectively.

[0136] In step S104, based on the optimization problem model, the decomposition result is reconstructed from the processed dynamic load equipment combination signal matrix, and based on the optimization problem model, the decomposition result is reconstructed from the stable load equipment converted into a power state sequence.

[0137] Specifically, the hybrid integer programming solution technique is used to solve the above optimization problem to obtain A and B n , the load of the dynamic load equipment is reconstructed through the load model DA, and through B n X n reconstruct the load of the stylistic load equipment to obtain the decomposition result.

[0138] Thus, this application combines the advantages of the classical combinatorial optimization model in dealing with stable load equipment and the advantages of the matrix decomposition model in dealing with dynamic load equipment, and solves them unifiedly in an optimization planning problem. Compared with other algorithms based on probability models and optimization models, it is more suitable for complex industrial scenarios with mixed different load types. On the other hand, this application also considers the data dependence problem. Compared with the widely used deep learning model, this application can achieve the same level of decomposition effect as the general deep learning model with less training data. This application can also make full use of the existing data in the industrial actual application scenario to obtain a more accurate load decomposition result, which has important significance and good application prospects.

[0139] To enable those skilled in the art to further understand the load decomposition method for industrial facilities based on hybrid integer programming, the following will be described in combination with specific embodiments.

[0140] Figure 2 is the flowchart of the load decomposition method for industrial facilities based on hybrid integer programming.

[0141] S201, obtain the equipment list; obtain the load curve to be decomposed and the equipment-level load curve.

[0142] S202, classify the equipment into dynamic load equipment, denoted by D n A n denote, and stable load equipment, the load is denoted by denote.

[0143] S203, perform digital filtering on the equipment-level load curve of the dynamic load equipment, use time invariance to expand the impulse part, and realize the expansion of the basis vector after reconstruction.

[0144] S204. Consider different types of devices and establish a mixed-integer programming model.

[0145] S205. Consider the physical limitations of industrial production and introduce corrections for industrial process constraints and equipment operation limitations into the model.

[0146] S206. Use mixed programming solution techniques to solve the optimization problem and reconstruct the load decomposition results.

[0147] According to the load decomposition method for industrial facilities based on mixed-integer programming proposed in the embodiments of the present application, by obtaining the equipment asset list of the industrial facilities to be decomposed, classifying industrial equipment into two categories: one is dynamic load equipment with continuously adjustable power, and the other is stable load equipment with power switching between fixed states; using the power consumption data at the equipment level of a certain scale as the training set, processing the load curves of appropriate dynamic load equipment using digital filtering methods, and expanding its training data set using the time-invariance of the frequently fluctuating pulse part in the load; constructing a mixed-integer programming model for decomposition using equipment information and training data; introducing model correction terms or constraints using physical limitations such as the industrial process dependencies of some equipment; solving the programming problem to obtain the decomposition results. Thus, the problem that existing load decomposition technologies are difficult to effectively handle the complex equipment load identification in industrial scenarios under the limitations of insufficient industrial load data and difficult data acquisition is solved, and the load characteristics of both dynamic load equipment and stable load equipment can be considered simultaneously, enabling high-accuracy non-intrusive load decomposition in industrial scenarios.

[0148] Next, a load decomposition device for industrial facilities based on mixed-integer programming proposed in the embodiments of the present application will be described with reference to the accompanying drawings.

[0149] Figure 3 It is a block diagram of the load decomposition device for industrial facilities based on mixed-integer programming according to the embodiments of the present application.

[0150] As Figure 3 shown, the load decomposition device 10 for industrial facilities based on mixed-integer programming includes: an acquisition module 100, a processing module 200, a first calculation module 300, and a second calculation module 400.

[0151] Among them, the acquisition module 100 is used to obtain the equipment list of the industrial equipment to be decomposed, the load curve to be decomposed, and obtain the equipment-level load curve of each equipment based on the equipment list and the load curve to be decomposed;

[0152] The processing module 200 is used to divide the industrial equipment to be decomposed into dynamic load equipment and stable load equipment according to the power consumption characteristics, perform digital filtering on the equipment-level load curve of the dynamic load equipment, and expand the frequently fluctuating pulse part in the load using time-invariance to obtain the processed load curve of the dynamic load equipment;

[0153] The first calculation module 300 is configured to construct a mixed-integer programming model based on the processed dynamic load device load curve and the stable load device load curve. After introducing the industrial process constraints and the corrected device operation limits into the mixed-integer programming model, the optimal values of the optimization variables are obtained by solving; and

[0154] The second calculation module 400 is configured to reconstruct the decomposition result of the processed dynamic load device combination signal matrix based on the optimization problem model, and reconstruct the decomposition result of the stable load device converted into a power state sequence based on the optimization problem model.

[0155] Optionally, in some embodiments, the processing module 200 is specifically configured to:

[0156] For the device-level load curve of the dynamic load device, first use median filtering to remove the noise spikes, then detect the large slope part in the device-level load curve of the dynamic load device and separate it. Finally, decompose the load of each dynamic device into a smooth basic part and a violently fluctuating pulse part;

[0157] Generate a column vector with all elements being 0 in a preset time period, connect the column vector to the beginning of each basis vector to be extended, and discard the sequence of the same length at the end of the basis vector to be extended, obtaining an extended pulse lagging by one unit. Superimpose the extended pulse on the smooth basic part to obtain a new basis vector;

[0158] Based on the new basis vector and the violently fluctuating pulse part, obtain the processed dynamic load device.

[0159] Optionally, the processing module 200 is further configured to:

[0160] Model the stable load device by using a combinatorial optimization method:

[0161]

[0162]

[0163]

[0164] Wherein, is the power of device n at time t, t is the time, T is the length of the time series for one optimization, n is the device number, and b n，k (t) is the variable for determining the power state where device n is located, k is an integer in [1, K n , and K n is the number of states of device n;

[0165] Model the dynamic load device by using a linear combination of the known device-level load curves of each device:

[0166]

[0167] s.t.: A n ≥0;

[0168] Wherein, is the power time series of device n, and D n is the signal matrix of device n; A n is the activation coefficient matrix.

[0169] Optionally, the first calculation module 300 is further configured to:

[0170] Connect the signal matrices of each dynamic load device to obtain D = [D 1 , D 2 ,..., D n ; Express the load model of the stable load device in the form of a matrix:

[0171]

[0172] Take the dimension with the same basis vector in the sum signal matrix, and we have:

[0173]

[0174] Construct a mixed-integer programming model:

[0175]

[0176] s.t.: A≥0

[0177] Wherein, X is the load curve read by the user's total smart meter, β is the regularization term coefficient, U is a column vector composed of 1s, Q is the signal grouping matrix, A is the activation coefficient matrix, N is the number of stable load devices, X n is the power state matrix, B n is the state selection matrix, and m is the length of the time series intercepted in one optimization.

[0178] Optionally, in some embodiments, the first calculation module 300 is further configured to:

[0179] The process constraint is:

[0180]

[0181] Limit the longest duration and the shortest duration for the stable load device. Among them, the longest duration W n，k is constrained by a set of linear inequalities:

[0182]

[0183]

[0184] Shortest duration w n，k Constrained by a set of linear inequalities:

[0185]

[0186]

[0187] where γ is the regularization term coefficient and R is the process constraint matrix.

[0188] It should be noted that the foregoing explanation of the embodiment of the load decomposition method for industrial facilities based on mixed-integer programming also applies to the load decomposition device for industrial facilities based on mixed-integer programming in this embodiment, and will not be elaborated here.

[0189] The load decomposition device for industrial facilities based on mixed-integer programming proposed according to the embodiments of the present application obtains the equipment asset list of the industrial facilities to be decomposed, classifies industrial equipment into two categories: one is dynamic load equipment with continuously adjustable power, and the other is stable load equipment with power switching between fixed states; uses a certain scale of equipment-level power consumption data as the training set, processes the load curves of appropriate dynamic load equipment using digital filtering methods, and expands its training data set using the time-invariance of the frequently fluctuating pulse part in the load; constructs a mixed-integer programming model for decomposition using equipment information and training data; introduces model correction terms or constraints using physical limitations such as industrial process dependencies of some equipment; and solves the programming problem to obtain the decomposition result. Thus, it solves the problem that existing load decomposition technologies are difficult to effectively handle complex equipment load identification in industrial scenarios under the limitations of insufficient industrial load data and difficult data acquisition, can consider the load characteristics of both dynamic load equipment and stable load equipment simultaneously, and can perform non-intrusive load decomposition with relatively high accuracy in industrial scenarios.

[0190] Figure 4 It is a schematic structural diagram of the electronic device provided by the embodiment of the present application. The electronic device may include:

[0191] A memory 401, a processor 402, and a computer program stored on the memory 401 and executable on the processor 402.

[0192] When the processor 402 executes the program, it implements the load decomposition method for industrial facilities based on mixed-integer programming provided in the above embodiment.

[0193] Further, the electronic device further includes:

[0194] A communication interface 403 for communication between the memory 401 and the processor 402.

[0195] A memory 401 for storing a computer program that can run on a processor 402.

[0196] The memory 401 may include a high-speed RAM memory and may also include non-volatile memory, such as at least one disk memory.

[0197] If the memory 401, the processor 402, and the communication interface 403 are implemented independently, the communication interface 403, the memory 401, and the processor 402 can be interconnected through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 4 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0198] Optionally, in a specific implementation, if the memory 401, the processor 402, and the communication interface 403 are integrated on a chip, the memory 401, the processor 402, and the communication interface 403 can communicate with each other through an internal interface.

[0199] The processor 402 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0200] The embodiments of the present application also provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned load decomposition method for industrial facilities based on mixed integer programming is implemented.

[0201] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples", etc., mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0202] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0203] Any process or method description shown in a flowchart or described in other ways herein can be understood to represent a module, segment, or part of code including one or more N executable instructions for implementing a customized logical function or process, and the scope of the preferred embodiments of this application includes additional implementations, where the functions can be executed in a manner that is not in the order shown or discussed, including in a substantially simultaneous manner according to the functions involved or in a reverse order, which should be understood by those skilled in the art to which the embodiments of this application pertain.

[0204] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part (electronic device) having one or N wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then storing it in a computer memory.

[0205] It should be understood that various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0206] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by a program instructing relevant hardware, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0207] In addition, each functional unit in various embodiments of the present application may be integrated into one processing module, or each unit may exist physically alone, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0208] The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present application.

Claims

1. A load decomposition method for industrial facilities based on mixed - integer programming, characterized in that, it includes the following steps: Obtain the equipment list of the industrial equipment to be decomposed, the load curve to be decomposed, and obtain the equipment - level load curve of each equipment based on the equipment list and the load curve to be decomposed; Divide the industrial equipment to be decomposed into dynamic load equipment and stable load equipment according to the power consumption characteristics, perform digital filtering on the equipment - level load curve of the dynamic load equipment, and use time - invariance to expand the frequently fluctuating pulse part in the load to obtain the load curve of the processed dynamic load equipment; Construct a mixed - integer programming model based on the load curve of the processed dynamic load equipment and the load curve of the stable load equipment. After introducing the correction of industrial process constraints and equipment operation restrictions into the mixed - integer programming model, solve to obtain the optimal values of the optimization variables; and Based on the optimization problem model, reconstruct the decomposition result for the combined signal matrix of the processed dynamic load equipment, and reconstruct the decomposition result for converting the stable load equipment into a power state sequence based on the optimization problem model.

2. The method according to claim 1, characterized in that, The performing digital filtering on the equipment - level load curve of the dynamic load equipment and using time - invariance to expand the frequently fluctuating pulse part in the load to obtain the processed dynamic load equipment includes: First, use median filtering to remove noise spikes from the equipment - level load curve of the dynamic load equipment, then detect the large - slope parts in the equipment - level load curve of the dynamic load equipment and separate them, and finally decompose the load of each dynamic equipment into a smooth basic part and a violently fluctuating pulse part; Generate a column vector with all elements being 0 in a preset time period, connect the column vector to the beginning of each base vector to be expanded, and discard the sequence of the same length at the end of the base vector to be expanded to obtain an expanded pulse lagging by one unit. Superimpose the expanded pulse on the smooth basic part to obtain a new base vector; Based on the new base vector and the violently fluctuating pulse part, obtain the processed dynamic load equipment.

3. The method according to claim 2, characterized in that, The dividing the industrial equipment to be decomposed into dynamic load equipment and stable load equipment according to the power consumption characteristics includes: Model the stable load equipment using a combinatorial optimization method: Among them, is the power of device n at time t, t is the time, T is the length of the time series for one optimization, n is the device number, b n，k (t) is a variable for determining the power state where device n is located, k is an integer in [1, K n , and K n is the number of states of device n; Model the dynamic load equipment using a linear combination of the known equipment - level load curves of each equipment: s.t.: A n ≥ 0; Among them, is the power time series of device n, D n is the signal matrix of device n; A n is the activation coefficient matrix.

4. The method according to claim 3, characterized in that, The constructing a mixed - integer programming model based on the load curve of the processed dynamic load equipment and the load curve of the stable load equipment further includes: Connect the signal matrices of each of the dynamic load devices as D = [D 1 , D 2 ,..., D n ; Express the load model of the stable load device in the form of a matrix: Take the same dimension as the base vector in the sum signal matrix, and there is: Construct the mixed - integer programming model: s.t.: A≥0 Among them, X is the load curve read from the user's total smart meter, β is the regularization term coefficient, U is a column vector composed of 1s, Q is the signal grouping matrix, A is the activation coefficient matrix, N is the number of stable compliance devices, and X n is the power status matrix, B n is the status selection matrix, and m is the length of the time series intercepted by one optimization.

5. The method according to claim 1, characterized in that, The introducing the correction of industrial process constraints and equipment operation restrictions into the mixed - integer programming model includes: The process constraint is: Limit the longest duration and the shortest duration of the stable load equipment, where the longest duration is W n，k Use a set of linear inequalities to constrain: The shortest duration w n，k is constrained by a set of linear inequalities: where γ is the regularization term coefficient and R is the process constraint matrix.

6. A load decomposition device for industrial facilities based on mixed integer programming, characterized in that, it includes: An acquisition module, configured to acquire a list of devices of the industrial equipment to be decomposed, a load curve to be decomposed, and obtain an equipment-level load curve of each device based on the device list and the load curve to be decomposed; A processing module, configured to divide the industrial equipment to be decomposed into dynamic load devices and stable load devices according to power consumption characteristics, perform digital filtering on the equipment-level load curve of the dynamic load devices, and utilize time invariance to expand the frequently fluctuating pulse part in the load to obtain a processed load curve of the dynamic load devices; A first calculation module, configured to construct a mixed integer programming model based on the processed load curve of the dynamic load devices and the load curve of the stable load devices, and after introducing corrections of industrial process constraints and equipment operation restrictions in the mixed integer programming model, solve to obtain the optimal values of the optimization variables; and A second calculation module, configured to reconstruct a decomposition result for the combined signal matrix of the processed dynamic load devices based on the optimization problem model, and reconstruct a decomposition result for converting the stable load devices into a power state sequence based on the optimization problem model.

7. The device according to claim 6, characterized in that, the processing module is specifically configured to: First, use median filtering to remove noise spikes from the equipment-level load curve of the dynamic load devices, then detect large slope parts in the equipment-level load curve of the dynamic load devices and separate them, and finally decompose the load of each dynamic device into a smooth basic part and a violently fluctuating pulse part; Generate a column vector with all elements being 0 in a preset time period, connect the column vector to the beginning of each base vector to be expanded, and discard the sequence of the same length at the end of the base vector to be expanded to obtain an expanded pulse lagging by one unit, and superimpose the expanded pulse on the smooth basic part to obtain a new base vector; Obtain the processed dynamic load devices based on the new base vector and the violently fluctuating pulse part.

8. The device according to claim 7, characterized in that, the processing module is used for: Model the stable load devices using a combinatorial optimization method: Among them, is the power of device n at time t, where t is the time, T is the length of the time series for one optimization, n is the device number, and b n，k (t) is a variable for determining the power state where device n is located, and k is an integer in [1, K n , and K n is the number of states of device n; Model the dynamic load devices using a linear combination of the known equipment-level load curves of each device: s.t.: A n ≥ 0; Among them, is the power time series of device n, D n is the signal matrix of device n; A n is the activation coefficient matrix.

9. The device according to claim 8, characterized in that, the first calculation module is further used for: Connect the signal matrices of each of the dynamic load devices as D = [D 1 , D 2 ,..., D n ; Express the load model of the stable load device in the form of a matrix: Take the same dimension as the base vector in the sum signal matrix, and there is: Construct the mixed integer programming model: s.t.: A≥0 Among them, X is the load curve read by the user's total smart meter, β is the regularization term coefficient, U is a column vector composed of 1s, Q is the signal grouping matrix, A is the activation coefficient matrix, N is the number of stable compliance devices, and X n is the power state matrix, B n is the state selection matrix, and m is the length of the time series intercepted by one optimization.

10. The device according to claim 6, characterized in that, the first calculation module is used for: The process constraint is: Limit the longest duration and the shortest duration of the stable load equipment, where the longest duration is W n，k Use a set of linear inequalities to constrain: the shortest duration w n，k is constrained by a set of linear inequalities: where γ is a regularization term coefficient and R is a process limit matrix.

11. An electronic device, characterized in that, it includes: A memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the program to implement the load decomposition method for industrial facilities based on mixed integer programming according to any one of claims 1-5.

12. A computer-readable storage medium, on which a computer program is stored, It is characterized in that The program is executed by a processor to implement the load decomposition method for industrial facilities based on mixed integer programming according to any one of claims 1-5.

Images