A non-intrusive modeling approach for integrated energy systems based on zero-shot learning

Abstract

The present invention belongs to the technical field of energy system modeling, and in particular, relates to a non-invasive modeling method for an integrated energy system based on zero-sample learning. The present invention utilizes zero-sample learning based on sparse coding to classify equipment in self-energy, trains the classification model with known data sets and auxiliary information of the integrated energy system, effectively transfers the knowledge learned from the known data sets to the unknown data sets, and realizes the classification of unknown equipment data sets; uses a bidirectional long short-term memory network to extract the equipment operation status, and divides the operation status of electrical equipment, power generation equipment and energy storage equipment in the integrated energy system according to the extraction results; establishes a multi-mode factor hidden Markov model based on the equipment operation status, and realizes non-invasive modeling of the integrated energy system containing unknown types of equipment. The present invention accurately identifies unknown types of equipment in the integrated energy system, effectively reduces the average error of the non-invasive modeling method, and significantly improves the model precision and accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy system modeling, and in particular relates to a non-invasive modeling method for an integrated energy system based on zero-sample learning. Background Art

[0002] The integrated energy system is a new open energy ecosystem that uses clean and renewable energy such as wind and solar as the main primary energy form, and uses energy technology and advanced control technology, intelligent optimization technology and information processing technology as implementation means. It can achieve efficient use of renewable energy and increase the proportion of renewable energy in primary energy production and consumption. Integrated energy system modeling is the basis for studying the safe and stable control of energy Internet systems and energy optimization scheduling.

[0003] At present, domestic and foreign scholars mainly focus on the study of intrusive modeling methods in the field of integrated energy system modeling. This method models the equipment included in the energy system separately, and then integrates each model according to the type of energy to represent the overall model of the energy system. However, this modeling method requires not only an exhaustive list of the equipment included in the system, but also the study of the operating mechanism of each equipment, which leads to many difficulties in energy system modeling, including:

[0004] (1) As the number of devices in the energy system increases, the model becomes more complex, and the number of model parameters and state variables will increase when optimizing and controlling such models;

[0005] (2) Due to the nonlinearity and uncertainty of the energy system, it is difficult to integrate a unified model that can reflect the state of the energy system;

[0006] (3) The system operating states are diverse, and different models need to be established and controlled.

[0007] Non-intrusive load monitoring methods provide a new solution for energy system modeling. Non-intrusive load monitoring can obtain refined user internal load categories and usage status data by decomposing and identifying total user data. However, there are many types of equipment included in the integrated energy system. With the replacement of various types of equipment and the access of new equipment, the above existing methods need to continuously update the system structure and equipment models, which greatly affects the accuracy of the integrated energy system model. Summary of the invention

[0008] In view of the shortcomings of the above-mentioned prior art, the present invention provides a non-invasive modeling method for an integrated energy system based on zero-sample learning. The purpose is to achieve high precision and high accuracy, and to be able to non-invasively model an integrated energy system containing unknown types of equipment.

[0009] The technical solution adopted by the present invention to achieve the above-mentioned purpose is:

[0010] A non-intrusive modeling method for an integrated energy system based on zero-shot learning comprises the following steps:

[0011] Step 1. Classify the known type equipment and unknown type equipment in the integrated energy system;

[0012] Step 2. Extract the operating status of each device in the integrated energy system according to the classification results;

[0013] Step 3. Divide according to the extracted running status;

[0014] Step 4: Establish a multi-modal factor hidden Markov model based on the partitioning results;

[0015] Step 5: Parameter estimation of the established multi-modal factor hidden Markov model;

[0016] Step 6: Decode the hidden states of the multi-modal factor hidden Markov model based on the parameter estimation results, and finally establish the comprehensive energy system decomposition model.

[0017] Furthermore, the known types of equipment in the integrated energy system include: photovoltaic power generation equipment, wind power generation equipment, electrical energy storage equipment, electric vehicles, rotating motors, electric boilers, micro gas turbines, electric heaters, electric ovens, refrigerators, washing machines and lighting equipment.

[0018] Furthermore, the classification of the known type of equipment and the unknown type of equipment is to utilize zero-sample learning based on sparse coding to classify the equipment in the self-energy, train the classification model on the known data set of the integrated energy system and its auxiliary information, and effectively transfer the knowledge learned from the known data set to the unknown data set, thereby realizing the classification of the unknown device data set.

[0019] Furthermore, the classification of known type equipment and unknown type equipment in the integrated energy system described in step 1 includes:

[0020] Step 1.1: N in the integrated energy system know The operating power time series of equipment of known types constitute a known data set Where D know is a known data set, p know,i Run the power time series for a known type of equipment, y know,i is the corresponding known type device label, Y know is a set of device labels of known categories, Q know The number of devices of known type;

[0021] Step 1.2: Identify the N in the integrated energy system X The unknown data set consists of the operating power time series of unknown type equipment Where D X is the unknown data set, p X,i Run the power time series for unknown type equipment, y know,i is the corresponding unknown type device label, Y know is a set of device labels of unknown categories, Q know is the number of devices of known type, y X,i The corresponding unknown type device label;

[0022] Step 1.3: Perform semantic dictionary learning on the operating power time series of equipment of known types:

[0023]

[0024] Among them, L know is a semantic dictionary of known device types, H know For a known type of device semantic representation, ||·|| F is the Frobenius norm, is the regularization term of the semantic dictionary of a known type of device, λ controls the strength of the regularization term, P know Run power time series for known types of equipment;

[0025] Step 1.4: Perform semantic dictionary learning on the operating power time series of unknown type equipment:

[0026]

[0027] in, q j for y X,j Representation in semantic embedding space, ω ij The input time series p X,i Belongs to label y X,j The probability of H X The semantic representation of unknown type device, L X It is a semantic dictionary of unknown type device, s i is the equipment operation status, ||L X -L know || is used to limit L X With L know The regularization term of fitness, ||h i -q j || is a regular term used to limit the similarity between the representation of the power time series of unknown type equipment in the semantic embedding space and the representation of the unknown type equipment label in the semantic embedding space;

[0028] Step 1.5: Fix the H in steps 3.3 and 3.4 separately. X and L X Perform alternating iterative solutions and find the corresponding classification in the embedding space based on the results:

[0029]

[0030]

[0031] In the above formula, The optimal solution for the semantic representation of unknown type of device, The optimal solution for the semantic representation of unknown type of device,||·|| F is the Frobenius norm.

[0032] Furthermore, in step 2, the operating status of each device in the integrated energy system is extracted according to the classification results, and the operating status is extracted using a device based on a bidirectional long short-term memory network. The Bi-LSTM network used contains six layers, where the length of the input layer is the length of the time window t, the second layer is a convolutional layer, which is used to extract features from the signal, the third and fourth layers are Bi-LSTM, the fifth layer is a convolutional layer, and the sixth layer is a fully connected layer. The entire network is trained by a time forward and reverse bidirectional propagation method.

[0033] Furthermore, the division according to the extracted running status in step 3 includes the following steps:

[0034] Step 3.1: Classify the operating status of electrical equipment according to the classification results in step 1:

[0035] Among them, n d is the number of electrical equipment of type d, The operating status of the electrical equipment at time t;

[0036] Step 3.2: Classify the operating status of the power generation equipment according to the classification results in step 1:

[0037] Among them, n DG is the number of working modes of power generation equipment, s DG , t is the operating status of the power generation equipment at time t;

[0038] Step 3.3: Classify the operating status of the energy storage equipment according to the classification results in step 1:

[0039]

[0040] in, is the operation mode of the energy storage device, is the rated power of the energy storage device, ses,t is the operating status of the energy storage device at time t.

[0041] Furthermore, in step 4, a multi-modal factor hidden Markov model is established according to the division result, as shown in the following formula:

[0042]

[0043]

[0044]

[0045] In the above formula, is the operating status of the Q+1th device at the initial moment, p t is the operating power of the equipment at time t, is the operating status of the Q+1th device at time t, is the operating status of the Q+1th device at time t-1, N(μ i ,ε i ) is the expected value μ i , standard deviation is ε i The normal distribution of is the operating status of the first device at time t, is the operating state of the second device at time t, π is the initial state probability distribution, A is the state transfer matrix, and B is the observation matrix.

[0046] Furthermore, the step 5 of estimating parameters of the established multi-modal factor hidden Markov model includes the following steps:

[0047] Step 5.1: Establish the parameter estimation equation of the multi-modal factor hidden Markov model as follows:

[0048]

[0049] In the above formula, θ * is the multi-mode factor hidden Markov model parameter with the maximum probability of the observed equipment operating power time series, P is the equipment operating power, and p t is the operating power of the equipment at time t, S is the operating status of the equipment, s t is the operating state of the device at time t, θ is the multi-mode factor hidden Markov model parameter, |P 1:T | is the operating power time series of the equipment from the initial moment to moment t;

[0050] Step 5.2: Introducing forward auxiliary variables Represents at time t , operating power time series of type q equipment in operating state i The joint probability of, given the initial parameter θ0, It is expressed as: Under the initial conditions, Expressed as

[0051] In the above formula, is the operating power time series of type q equipment at time t and operating state i The joint probability of θ0 is the initial parameter of the multi-modal factor hidden Markov model. is the operating status of the q-type equipment at the initial moment, and i is the operating status of the equipment at moment t;

[0052] Step 5.3: Based on The forward recursive principle obtains

[0053]

[0054] In the above formula, is the operating power time series of type q equipment at time t and operating state i The joint probability of is the state transition probability of type q equipment, is the operating power of type q equipment at time t+1, is the operating status of the q type equipment at time t+1, j is the operating status at time t+1, is the operating power time series of type q equipment at time t+1 and operating state j The joint probability of

[0055] Step 5.4: Introducing backward auxiliary variables Represents the operating power time series of type q equipment observed at time t and operating state i The probability of, given the initial parameter θ0, It is expressed as:

[0056]

[0057] Step 5.5: Use the recursive formula to calculate the backward variables Perform the calculation:

[0058] Among them, the initial value

[0059] In the above formula, is the state transition probability of type q equipment, is the operating status of the q-type device at time t+1, is the operating power of type q equipment at time t+1;

[0060] Step 5.6: Given the initial parameters θ0 and the observation sequence The multi-modal factor hidden Markov model calculates the type of device q from the state Transfer to state Probability and the state at time t Probability

[0061]

[0062]

[0063] In the above formula, is the state transition probability of type q equipment, is the backward auxiliary variable at time t+1, is the backward auxiliary variable at time t, The state presented at time t The probability of For a device of type q, the slave state Transfer to state probability;

[0064] Step 5.7: Recalculate the model parameters of the q hidden Markov chains:

[0065]

[0066]

[0067]

[0068]

[0069] In the above formula, is the estimated value of the initial state probability of the q-type device, is the estimated value of the state transition probability of type q equipment, is the estimated value of the expected value of the observation matrix of type q equipment, is the estimated value of the standard deviation of the observation matrix of type q equipment;

[0070] Step 5.8: Iterate and calculate the forward variables according to the new parameters Backward Variable and Until convergence.

[0071] Furthermore, the step 6 of decoding the hidden state of the multi-mode factor hidden Markov model from the parameter estimation result and finally establishing the comprehensive energy system decomposition model is to decode the hidden state of the multi-mode factor hidden Markov model by applying the Viterbi algorithm to the aggregated power consumption sequence parameter estimation result, and finally establishing the comprehensive energy system decomposition model; comprising the following steps:

[0072] Step 6.1: Introduce variable δ t (i) represents all states {s1,s1,...,s t} to the observation sequence {p1,p2,...,p T The maximum probability of} is: Where N is the number of hidden layer states of the device, t = 1, 2, ..., T;

[0073] Step 6.2: Input the value obtained in step 5.7 And the integrated energy system output power time series P 1:T ={p1,p2,...,p T};

[0074] Step 6.3: Initialization

[0075] Step 6.4: Recursion

[0076] Step 6.5: Optimal path backtracking:

[0077] A computer storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of a non-invasive modeling method for an integrated energy system based on zero-sample learning are implemented.

[0078] The present invention has the following beneficial effects and advantages:

[0079] The present invention proposes a non-invasive modeling method for an integrated energy system based on zero-sample learning, which can be used to non-invasively model an integrated energy system containing unknown types of equipment. First, the operating characteristics of known types of equipment and unknown types of equipment connected to the integrated energy system are extracted through dictionary learning; further, the equipment in the integrated energy system is classified using zero-sample learning based on sparse coding; on this basis, the bidirectional long short-term memory network is used to extract the equipment operation status, and the integrated energy system is modeled based on a multi-modal factor hidden Markov model. The method of the present invention accurately identifies unknown types of equipment in the integrated energy system, and the established model effectively reduces the average error of the non-invasive modeling method, so that the precision and accuracy of the model are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0081] Figure 1 It is a flow chart of the non-intrusive modeling method of the integrated energy system based on zero-sample learning of the present invention;

[0082] Figure 2 is a schematic diagram of the integrated energy system of the present invention;

[0083] Figure 3 It is a schematic diagram of the classification of the integrated energy system based on zero-sample learning of the present invention;

[0084] Figure 4 It is a bidirectional long short-term memory network structure diagram of the present invention;

[0085] Figure 5 It is a diagram for dividing the operation states of the multi-modal device of the present invention;

[0086] Figure 6 It is the multi-modal factor hidden Markov model of the present invention. DETAILED DESCRIPTION

[0087] In order to more clearly understand the above-mentioned objectives, features and advantages of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0088] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0089] Refer to the following Figure 1-Figure 6 The technical solutions of some embodiments of the present invention are described.

[0090] Example 1

[0091] The present invention provides an embodiment, which is a non-intrusive modeling method for an integrated energy system based on zero-sample learning, which is implemented by a non-intrusive modeling method for an integrated energy system based on zero-sample learning, such as Figure 1 As shown, Figure 1 It is a flow chart of the non-invasive modeling method of the integrated energy system based on zero-sample learning of the present invention.

[0092] The present invention specifically comprises the following steps:

[0093] Step 1. Classify the known type equipment and unknown type equipment in the integrated energy system;

[0094] The known types of equipment in the integrated energy system based on the non-invasive modeling of zero-sample learning in the present invention include: photovoltaic power generation equipment, wind power generation equipment, electrical energy storage equipment, electric vehicles, rotating motors, electric boilers, micro gas turbines, electric heaters, electric ovens, refrigerators, washing machines and lighting equipment.

[0095] The steps include:

[0096] Step 1.1: Figure 2 N in the comprehensive energy system shown know The operating power time series of equipment of known types constitute a known data set Where D know is a known data set, p know,i Run the power time series for a known type of equipment, y know,i is the corresponding known type device label, Y know is a set of device labels of known categories, Q know The number of devices of known type;

[0097] Step 1.2: Figure 2 The N to be identified in the integrated energy system shown X The unknown data set consists of the operating power time series of unknown type equipment Where D X is the unknown data set, p X,i Run the power time series for unknown type equipment, y know,i is the corresponding unknown type device label, Y know is a set of device labels of unknown categories, Q know is the number of devices of known type, y X,i The corresponding unknown type device label.

[0098] Step 1.3: Perform semantic dictionary learning on the operating power time series of equipment of known types:

[0099]

[0100] Among them, L know is a semantic dictionary of known device types, H know For a known type of device semantic representation, ||·|| F is the Frobenius norm, is the regularization term of the semantic dictionary of a known type of device, λ controls the strength of the regularization term, P know Run a power time series for a known type of device.

[0101] Step 1.4: Perform semantic dictionary learning on the operating power time series of unknown type equipment:

[0102]

[0103] in, q j for y X,j Representation in semantic embedding space, ω ij The input time series p X,i Belongs to label y X,j The probability of H X The semantic representation of unknown type device, L X It is a semantic dictionary of unknown type device, s i = Equipment operation status; ||L X -L know || is used to limit L X With L know The regularization term of fitness, ||h i -q j || is a regular term used to limit the similarity between the representation of the power time series of the unknown type of equipment in the semantic embedding space and the representation of the unknown type of equipment label in the semantic embedding space.

[0104] Step 1.5: Fix the H in steps 3.3 and 3.4 separately. X and L X Perform alternating iterations to solve the problem, and then Figure 3 Find the corresponding classification in the embedding space shown:

[0105]

[0106]

[0107] In the above formula, The optimal solution for the semantic representation of unknown type of device, The optimal solution for the semantic representation of unknown type of device,||·|| F is the Frobenius norm.

[0108] Step 2. Extract the operating status of each device in the integrated energy system according to the classification results;

[0109] Utilize based Figure 4 The bidirectional long short-term memory network (Bi-directional Long Short-Term Memory, Bi-LSTM) device shown in the figure performs running status extraction. The Bi-LSTM network used contains six layers in total, where the length of the input layer is the length of the time window t, the second layer is a convolutional layer, which is used to extract features from the signal, the third and fourth layers are Bi-LSTM, the fifth layer is a convolutional layer, and the sixth layer is a fully connected layer. The entire network is trained by the time forward and reverse bidirectional propagation method.

[0110] Step 3: Divide according to the extracted running status.

[0111] According to the equipment operation status extracted by Bi-LSTM, the operation status of the power-consuming equipment, power generation equipment and energy storage equipment in the integrated energy system is divided, and the 0-1 switch operation status division problem of the equipment is transformed into a multi-modal division problem of the same type of equipment according to the equipment type, such as Figure 5 As shown, Figure 5 The multi-modal device operation state division diagram of the present invention includes the following steps:

[0112] Step 3.1: Classify the operating status of electrical equipment according to the classification results in step 1:

[0113] Among them, n d is the number of electrical equipment of type d, The operating status of the electrical equipment at time t.

[0114] Step 3.2: Classify the operating status of the power generation equipment according to the classification results in step 1:

[0115] Among them, n DG is the number of working modes of power generation equipment, s DG,t is the operating status of the power generation equipment at time t.

[0116] Step 3.3: Classify the operating status of the energy storage equipment according to the classification results in step 1:

[0117]

[0118] in, is the operation mode of the energy storage device, is the rated power of the energy storage device, s es,t is the operating status of the energy storage device at time t.

[0119] Step 4: Establish a multi-modal factor hidden Markov model based on the partition results, such as Figure 6 As shown:

[0120]

[0121]

[0122]

[0123] In the above formula, is the operating status of the Q+1th device at the initial moment, p t is the operating power of the equipment at time t, is the operating status of the Q+1th device at time t, is the operating status of the Q+1th device at time t-1, N(μ i ,ε i ) is the expected value μ i , standard deviation is ε i The normal distribution of is the operating status of the first device at time t, is the operating state of the second device at time t, π is the initial state probability distribution, A is the state transfer matrix, and B is the observation matrix.

[0124] Step 5: Parameter estimation of the established multi-modal factor hidden Markov model includes the following steps:

[0125] Step 5.1: Establish the parameter estimation equation of the multi-modal factor hidden Markov model as follows:

[0126]

[0127] In the above formula, θ * is the multi-mode factor hidden Markov model parameter with the maximum probability of the observed equipment operating power time series, P is the equipment operating power, and p t is the operating power of the equipment at time t, S is the operating status of the equipment, s t is the operating state of the device at time t, θ is the multi-mode factor hidden Markov model parameter, |P 1:T | is the operating power time series of the equipment from the initial moment to moment t.

[0128] Step 5.2: Introducing forward auxiliary variables Represents the operating power time series of type q equipment at time t and operating state i The joint probability of, given the initial parameter θ0, It is expressed as: Under the initial conditions, Expressed as

[0129] In the above formula, is the operating power time series of type q equipment at time t and operating state i The joint probability of θ0 is the initial parameter of the multi-modal factor hidden Markov model. is the operating status of the q-type equipment at the initial moment, and i is the operating status of the equipment at moment t.

[0130] Step 5.3: Based on The forward recursive principle obtains

[0131]

[0132] In the above formula, is the operating power time series of type q equipment at time t and operating state i The joint probability of is the state transition probability of type q equipment, is the operating power of type q equipment at time t+1, is the operating status of the q type equipment at time t+1, j is the operating status at time t+1, is the operating power time series of type q equipment at time t+1 and operating state j The joint probability of .

[0133] Step 5.4: Introducing backward auxiliary variables Represents the operating power time series of type q equipment observed at time t and operating state i The probability of, given the initial parameter θ0, It is expressed as:

[0134]

[0135] Step 5.5: Use the recursive formula to calculate the backward variables Perform the calculation:

[0136] Among them, the initial value

[0137] In the above formula, is the state transition probability of type q equipment, is the operating status of the q-type device at time t+1, is the operating power of type q equipment at time t+1.

[0138] Step 5.6: Given the initial parameters θ0 and the observation sequence The multi-modal factor hidden Markov model calculates the type of device q from the state Transfer to state Probability and the state at time t Probability

[0139]

[0140]

[0141] In the above formula, is the state transition probability of type q equipment, is the backward auxiliary variable at time t+1, is the backward auxiliary variable at time t, The state presented at time t The probability of For a device of type q, the slave state Transfer to state probability.

[0142] Step 5.7: Recalculate the model parameters of the q hidden Markov chains:

[0143]

[0144]

[0145]

[0146]

[0147] In the above formula, is the estimated value of the initial state probability of the q-type device, is the estimated value of the state transition probability of type q equipment, is the estimated value of the expected value of the observation matrix of type q equipment, is the estimated value of the standard deviation of the observation matrix of type q equipment.

[0148] Step 5.8: Iterate and calculate the forward variables according to the new parameters Backward Variable and Until convergence.

[0149] Step 6: Apply the Viterbi algorithm to the summarized power consumption sequence parameter estimation results to decode the hidden states of the multi-mode factor hidden Markov model, and finally establish the comprehensive energy system decomposition model.

[0150] The specific steps include:

[0151] Step 6.1: Introduce variable δ t (i) represents all states {s1,s1,...,s t} to the observation sequence {p1,p2,...,p T The maximum probability of} is: Where N is the number of hidden layer states of the device, t=1,2,...,T.

[0152] Step 6.2: Input the value obtained in step 5.7 And the integrated energy system output power time series P 1:T ={p1,p2,...,p T};

[0153] Step 6.3: Initialization

[0154] Step 6.4: Recursion

[0155] Step 6.5: Optimal path backtracking:

[0156] This embodiment achieves the following technical effects:

[0157] (1) It can accurately identify the types of unknown devices connected to the integrated energy system;

[0158] (2) It can accurately extract the operating status of electrical equipment, energy storage equipment, and power generation equipment in the integrated energy system, and classify their operating status according to the equipment type;

[0159] (3) Based on equipment type identification and equipment operating status extraction, the factor hidden Markov model can accurately reflect the operating status of each device in the integrated energy system and the overall energy output of the integrated energy system.

[0160] (4) The method of the present invention effectively reduces the computational complexity and time of modeling the integrated energy system by dividing the operating status of the equipment in the integrated energy system. At the same time, it uses zero-sample learning to classify unknown equipment, thereby improving the precision and accuracy of the model.

[0161] Example 2

[0162] Based on the same inventive concept, an embodiment of the present invention also provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a non-invasive modeling method of an integrated energy system based on zero-sample learning as described in Example 1 are implemented.

[0163] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0164] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0165] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0166] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A non-intrusive modeling method for integrated energy systems based on zero-shot learning, characterized by: The following steps are involved: Step 1. Classify the known type equipment and unknown type equipment in the integrated energy system, use zero-shot learning based on sparse coding to classify the equipment in the energy source, train the classification model with the known data set of the integrated energy system and its auxiliary information, and effectively transfer the knowledge learned from the known data set to the unknown data set, so as to realize the classification of the unknown equipment data set; including: Step 1.

1. Classify the N in the integrated energy system know The operating power time series of equipment of known types constitute a known data set Where D know is a known data set, p know,i Run the power time series for a known type of equipment, y know,i is the corresponding known type device label, Y know is a set of device labels of known categories, Q know is the number of known types of equipment; Step 1.

2. N to be identified in the integrated energy system X The unknown data set consists of the operating power time series of unknown type equipment Where D X is the unknown data set, p X,i Run the power time series for unknown type equipment, y know,i is the corresponding unknown type device label, Y know is a set of device labels of unknown categories, Q know is the number of devices of known type, y X,i is the corresponding unknown type equipment label; Step 1.

3. Perform semantic dictionary learning on the operating power time series of known type equipment: Among them, L know is a semantic dictionary of known device types, H know For a semantic representation of a device of known type, ||·|| F is the Frobenius norm, is the regularization term of the semantic dictionary of a known type of device, λ controls the strength of the regularization term, P know Run the power time series for the known type of equipment; Step 1.

4. Perform semantic dictionary learning on the power time series of the unknown type of equipment: in, q j for y X,j Representation in semantic embedding space, ω ij The input time series p X,i Belongs to label y X,j The probability of H X The semantic representation of unknown type device, L X It is a semantic dictionary of unknown type device, s i is the equipment operation status, ||L X -L know || is used to limit L X With L know The regularization term of fitness, ||h i -q j || is a regular term used to limit the similarity between the representation of the power time series of unknown type equipment in the semantic embedding space and the representation of the unknown type equipment label in the semantic embedding space; Step 1.

5. Fix H in steps 1.3 and 1.4 respectively X and L X Perform alternating iterative solutions and find the corresponding classification in the embedding space based on the results: In the above formula, The optimal solution for the semantic representation of unknown type of device, The optimal solution for the semantic representation of unknown type of device,||·|| F is the Frobenius norm; Step 2. Extract the operating status of each device in the integrated energy system according to the classification results, and use the bidirectional long short-term memory network device to extract the operating status. The Bi-LSTM network used contains six layers, in which the length of the input layer is the length of the time window t, the second layer is a convolutional layer, which is used to extract features from the signal, the third and fourth layers are Bi-LSTM, the fifth layer is a convolutional layer, and the sixth layer is a fully connected layer. The entire network is trained by the time forward and reverse bidirectional propagation method; Step 3. Divide according to the extracted running status; Step 4. Establish a multi-modal factor hidden Markov model based on the partition results, as follows: In the above formula, is the operating status of the Q+1th device at the initial moment, p t is the operating power of the equipment at time t, is the operating status of the Q+1th device at time t, is the operating status of the Q+1th device at time t-1, N(μ i ,ε i ) is the expected value μ i , standard deviation is ε i The normal distribution of is the operating status of the first device at time t, is the operating state of the second device at time t, π is the initial state probability distribution, A is the state transfer matrix, and B is the observation matrix; Step 5. Estimating parameters of the established multi-modal factor hidden Markov model; Step 6. Decode the parameter estimation results into hidden states of the multi-modal factor hidden Markov model and finally establish the integrated energy system decomposition model.

2. The non-intrusive modeling method for an integrated energy system based on zero-sample learning according to claim 1 is characterized by: Known types of equipment in the integrated energy system include: photovoltaic power generation equipment, wind power generation equipment, electrical energy storage equipment, electric vehicles, rotating motors, electric boilers, micro gas turbines, electric heaters, electric ovens, refrigerators, washing machines and lighting equipment.

3. The non-intrusive modeling method for an integrated energy system based on zero-sample learning according to claim 1 is characterized by: Step 3, dividing according to the extracted running status, includes the following steps: Step 3.

1. Classify the operating status of electrical equipment according to the classification results in step 1: Among them, n d is the number of electrical equipment of type d, The operating status of the electrical equipment at time t; Step 3.

2. Classify the operating status of the power generation equipment according to the classification results in step 1: Among them, n DG is the number of working modes of power generation equipment, s DG,t is the operating status of the power generation equipment at time t; Step 3.

3. Classify the operating status of the energy storage equipment according to the classification results in step 1: in, is the operation mode of the energy storage device, is the rated power of the energy storage device, s es,t is the operating status of the energy storage device at time t.

4. The non-intrusive modeling method for an integrated energy system based on zero-sample learning according to claim 1 is characterized by: The step 5 of performing parameter estimation on the established multi-modal factor hidden Markov model comprises the following steps: Step 5.

1. Establish the parameter estimation equation of the multi-modal factor hidden Markov model as follows: In the above formula, θ * is the multi-mode factor hidden Markov model parameter with the maximum probability of the observed equipment operating power time series, P is the equipment operating power, and p t is the operating power of the equipment at time t, S is the operating status of the equipment, s t is the operating state of the device at time t, θ is the multi-mode factor hidden Markov model parameter, |P 1:T | is the operating power time series of the equipment from the initial time to time t; Step 5.

2. Introducing forward auxiliary variables Represents at time t , operating power time series of type q equipment in operating state i The joint probability of, given the initial parameter θ0, It is expressed as: Under the initial conditions, Expressed as In the above formula, is the operating power time series of type q equipment at time t and operating state i The joint probability of θ0 is the initial parameter of the multi-modal factor hidden Markov model. is the operating status of the q-type equipment at the initial moment, and i is the operating status of the equipment at moment t; Step 5.

3. Based on The forward recursive principle obtains In the above formula, is the operating power time series of type q equipment at time t and operating state i The joint probability of is the state transition probability of type q equipment, is the operating power of type q equipment at time t+1, is the operating status of the q type equipment at time t+1, j is the operating status at time t+1, is the operating power time series of type q equipment at time t+1 and operating state j The joint probability of Step 5.

4. Introducing backward auxiliary variables Represents the operating power time series of type q equipment observed at time t and operating state i The probability of, given the initial parameter θ0, It is expressed as: Step 5.

5. Use the recursive formula to calculate the backward variables Perform the calculation: Among them, the initial value In the above formula, is the state transition probability of type q equipment, is the operating status of the q-type device at time t+1, is the operating power of type q equipment at time t+1; Step 5.

6. Given the initial parameters θ0 and the observation sequence The multi-modal factor hidden Markov model calculates the type of device q from the state Transfer to state Probability and the state at time t Probability In the above formula, is the state transition probability of type q equipment, is the backward auxiliary variable at time t+1, is the backward auxiliary variable at time t, The state presented at time t The probability of For a device of type q, the slave state Transfer to state probability; Step 5.

7. Recalculate the model parameters of the q hidden Markov chains: In the above formula, is the estimated value of the initial state probability of the q-type device, is the estimated value of the state transition probability of type q equipment, is the estimated value of the expected value of the observation matrix of type q equipment, is the estimated value of the standard deviation of the observation matrix of type q equipment; Step 5.

8. Iterate and calculate the forward variables according to the new parameters Backward Variable and Until convergence.

5. The non-intrusive modeling method for an integrated energy system based on zero-sample learning according to claim 1 is characterized by: The step 6 of decoding the hidden state of the multi-mode factor hidden Markov model from the parameter estimation result and finally establishing the comprehensive energy system decomposition model is to decode the hidden state of the multi-mode factor hidden Markov model by applying the Viterbi algorithm to the aggregated power consumption sequence parameter estimation result, and finally establishing the comprehensive energy system decomposition model; comprising the following steps: Step 6.

1. Introduce variable δ t (i) represents all states {s1,s1,...,s t } to the observation sequence {p1,p2,...,p T The maximum probability of} is: Where N is the number of hidden layer states of the device, t = 1, 2, ..., T; Step 6.

2. Input the value obtained in step 5.7 And the integrated energy system output power time series P 1:T ={p1,p2,...,p T }; Step 6.

3. Initialization Step 6.

4. Recursion Step 6.

5. Optimal path backtracking: t=T-1,T-2,…,1.

6. A computer storage medium, characterized in that: The computer storage medium stores a computer program, which, when executed by a processor, implements the steps of a non-invasive modeling method for an integrated energy system based on zero-sample learning as described in any one of claims 1 to 5.

Images