Power plant elasticity aggregation method and device considering backup margin, equipment and medium

Abstract

The application discloses a power plant elasticity aggregation method and device considering backup margin, equipment and medium, relates to the technical field of power system optimization scheduling, and is used for solving the problem that the existing elasticity aggregation method lacks flexible adjustment. The method comprises the following steps: predicting a flexible resource adjustment potential probability interval through a QR-Siamese LSTM model constructed by quantile regression and twin long short-term memory; performing virtual power plant flexible resource aggregation modeling according to the flexible resource adjustment potential probability interval; and obtaining a virtual power plant backup adjustment margin elasticity interval based on error propagation solution of the virtual power plant flexible resource aggregation modeling. The application predicts the flexible resource adjustment potential probability interval, and then aggregates to obtain the adjustment margin elasticity interval.

Description

Technical Field

[0001] The present invention relates to the technical field of power system optimization and dispatching, and in particular to a QR-Siamese LSTM-based power plant elastic aggregation method, device, equipment and medium considering reserve margin. Background Art

[0002] Flexible resources are those capable of flexibly adjusting to maintain a dynamic balance between supply and demand. With the increasing availability of flexible resources, renewable energy is also rapidly developing. Currently, how to efficiently integrate these diverse resources, flexibly respond to emergency regulation needs in the power system, and improve the overall economic efficiency of the power system has become a pressing issue.

[0003] Virtual power plants (VPPs) are an innovative energy management solution that leverages advanced information technology and control strategies to aggregate vast amounts of flexible resources into a coordinated, reliable power resource cluster, enabling them to participate in grid demand response control. However, current research on VPPs focuses primarily on analyzing trading mechanisms for VPPs in various markets and studying VPP market bidding strategies. There is limited research on VPPs for grid demand response control.

[0004] Existing research on virtual power plant aggregation technology fails to fully consider the correlation between the power system's reserve regulation margin requirements and the probability and potential range of flexible resources. Instead, it sets a reserve margin range based on a fixed maximum load forecast and the required reserve margin ratio, and then aggregates flexible resources based on this. This virtual power plant aggregation technology increases system reserve regulation costs and reduces system operational economics. Therefore, a flexible and adaptable aggregation method is needed. Summary of the Invention

[0005] In order to overcome the shortcomings of the existing technology, one of the objectives of the present invention is to provide a power plant elastic aggregation method considering the reserve margin, which constructs a QR-SiameseLSTM model to obtain a more accurate and reliable probability prediction interval of flexible resource regulation potential, and then obtains the regulation margin elasticity interval through elastic aggregation.

[0006] One of the purposes of the present invention is achieved by the following technical solution:

[0007] A power plant resilience aggregation method considering reserve margin includes the following steps:

[0008] The QR-SiameseLSTM model constructed by quantile regression and twin long short-term memory is used to predict the probability interval of flexible resource adjustment potential;

[0009] Performing aggregation modeling of flexible resources of the virtual power plant according to the probability interval of the flexible resource regulation potential;

[0010] The error propagation is used to solve the aggregate modeling of the flexible resource of the virtual power plant, to obtain a reserve adjustment margin elasticity interval of the virtual power plant.

[0011] Further, in order to fully exhibit the ability of capturing the time coupling and nonlinear characteristics of the flexible resource, the QR-SiameseLSTM model comprises a first LSTM neural network, a second LSTM neural network and a quantile regression layer.

[0012] The first LSTM neural network is used to output an encoded load elasticity feature according to demand side load sequence data, time-of-use electricity price, weather parameters and time information.

[0013] The second LSTM neural network is used to output an elasticity vector according to the encoded load elasticity feature.

[0014] The quantile regression layer is used to output a confidence interval of load prediction.

[0015] The flexible resource adjustment potential probability interval is calculated according to the confidence interval, and the following formula is satisfied:

[0016] ,

[0017] ,

[0018] wherein, and are updated upper and lower boundaries of confidence, and are upper and lower boundary prediction deviations at the / 2 quantile level, is the flexible resource adjustment potential probability interval, and are prediction values corresponding to the quantile levels of / 2 and / 2.

[0019] Further, in order to obtain the total reserve capacity required by the power system, after the flexible resource adjustment potential probability interval is calculated, the following steps are further included: calculating a reserve adjustment set capacity interval, satisfying: wherein, represents a probability conversion function, represents a confidence of the flexible resource probability interval, represents a center value of the probability interval, represents a quantile level of the prediction interval at the time.

[0020] Furthermore, based on the flexible resource adjustment potential probability interval, aggregation modeling of the flexible resources of the virtual power plant is performed, including:

[0021] According to the number of flexible resources involved in regulation and the number of equipment of each type of resources in the virtual power plant, the algebraic sum of the regulation capacity of the virtual power plant is calculated;

[0022] The adjustable potential interval characteristics of flexible resources are mapped to the aggregated external adjustable characteristics of the virtual power plant for aggregation calculation to meet the following requirements:

[0023] ,

[0024] in, Before optimization The forecast load capacity at the time, for Flexible aggregation capacity at all times, Indicates the upper and lower limits of adjustable potential.

[0025] Furthermore, the constraints of the aggregate calculation include technical constraints combining linear network constraints and flexible resource models, and constraints on the aggregate regulation potential of virtual power plants;

[0026] The technical constraints are satisfied:

[0027] ,

[0028] ,

[0029] ,

[0030] in, For decision variables The vector of time, , represents the set of technical constraints, , Represents the matrix and vector form of the network constraint set, represents the output power trajectory of the virtual power plant, , Respectively represent the matrix and vector forms of the output power constant coefficient matrix;

[0031] The virtual power plant aggregate regulation potential constraint satisfy: ,in, and are the quantile levels in / 2 and / 2 corresponding to the predicted value, Indicates the center value.

[0032] Further, based on error propagation solving the virtual power plant flexible resource aggregation modeling, a virtual power plant reserve adjustment margin elasticity interval is obtained, including:

[0033] The virtual power plant active power flexibility is modeled as a polyhedral projection, and the expression satisfies: Wherein, And The calculation parameters of the virtual power plant active power elasticity interval are represented

[0034] The constraint condition of the aggregation calculation is optimized and modeled, and the coupling constraint is eliminated through a robust optimization algorithm, so that the uncertain parameters in the constraint condition are converted into a deterministic problem.

[0035] According to the polyhedral projection and the optimized constraint condition, an optimized aggregation calculation formula is constructed, and the aggregation calculation formula is:

[0036] Wherein, The occurrence probability conversion function is represented The virtual battery active power output power affine value is represented The virtual battery active power output power affine value is represented The virtual generator active power output power affine value is represented The virtual battery aggregation elasticity range is represented The virtual generator aggregation range is represented The virtual power plant aggregation range is represented

[0037] The aggregation calculation formula is solved to obtain a virtual power plant reserve margin elasticity interval.

[0038] Further, solving the aggregation calculation formula includes:

[0039] The aggregation calculation is described by error interval propagation to obtain an aggregation formula for solving:

[0040] Wherein, The interface shrinkage equation of the polyhedral projection is represented The noise source of the variable , , The interval center value and interval radius variable are represented

[0041] The virtual power plant reserve margin elasticity interval is obtained by solving.

[0042] The second purpose of the present application is to provide a power plant elasticity aggregation device considering reserve margin.

[0043] The second purpose of the present application is achieved by the following technical solutions:

[0044] A power plant elasticity aggregation device considering reserve margin, comprising:

[0045] A potential prediction module is configured to predict a flexible resource adjustment potential probability interval by quantile regression, a QR-SiameseLSTM model constructed by a twin long short-term memory.

[0046] An aggregation module is configured to perform virtual power plant flexible resource aggregation modeling according to the flexible resource adjustment potential probability interval.

[0047] A solving module is configured to solve the virtual power plant flexible resource aggregation modeling based on error propagation to obtain a virtual power plant reserve adjustment margin elasticity interval.

[0048] A third object of the present application is to provide an electronic device for implementing one of the objects of the present application, which comprises a processor, a storage medium, and a computer program stored in the storage medium and implemented by the processor when executed.

[0049] A fourth object of the present application is to provide a computer-readable storage medium storing one of the objects of the present application, which stores a computer program and is implemented by a processor when executed.

[0050] Compared with the prior art, the present application has the following advantages:

[0051] The present application proposes a virtual power plant elasticity interval quantification evaluation and elasticity aggregation strategy considering reserve adjustment margin and flexible resource, an accurate and reliable flexible resource adjustment potential probability prediction interval is obtained by constructing a QR-SiameseLSTM model, to predict the overall required reserve capacity of the power system, and then perform aggregation calculation of the flexible resource according to the prediction result, to complete dynamic evaluation of the overall adjustment potential after aggregation of the reserve capacity and the flexible resource, improve resource utilization efficiency, and save system reserve adjustment cost. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 is a flowchart of the power plant elasticity aggregation considering reserve margin of embodiment one;

[0053] Figure 2 is a wind power adjustment potential probability interval prediction result schematic diagram of embodiment one;

[0054] Figure 3 is a flexible load demand probability interval prediction result schematic diagram of embodiment one;

[0055] Figure 4 is a virtual power plant ordered aggregation result schematic diagram of embodiment one;

[0056] Figure 5 is a schematic diagram of optimization results of example one;

[0057] Figure 6 is a structural block diagram of a power plant elastic aggregation device considering backup margin of example two;

[0058] Figure 7 is a structural block diagram of an electronic device of example three. DETAILED DESCRIPTION

[0059] The application will be described in more detail below with reference to the drawings, it should be noted that the following description of the application with reference to the drawings is merely illustrative and not restrictive. Various different embodiments can be combined with each other to constitute other embodiments not shown in the following description.

[0060] Example one

[0061] Example one provides a power plant elastic aggregation method considering backup margin, which aims to predict the flexible resource adjustment potential elastic interval, realize virtual power plant ordered aggregation based on the center value of the probability interval of each type of flexible resource, and finally solve the virtual power plant elastic space by considering the dynamic relationship between the flexible resource output probability and the backup adjustment margin probability interval by using interval analysis method.

[0062] It should be emphasized that the power plant described in the present embodiment and the present application is specifically a virtual power plant.

[0063] The power system usually needs to consider a certain backup adjustment margin to ensure that the system is safe and stable when encountering peak load or impact disturbance. Backup adjustment margin is a part of additional capacity reserved by the power system to ensure safe and stable operation under peak load, sudden failure or disturbance. The backup adjustment margin of the power system is usually 25% to 30% of the day-ahead maximum load forecast value, which includes load backup capacity, accident backup capacity and maintenance backup capacity, respectively, to cope with load impact, generator unit accident shutdown and maintenance plan. For example: when the day-ahead maximum load forecast value is 100, the backup adjustment margin interval is [100 25%, 100 30%]. However, the day-ahead maximum load does not exist for a long time, and does not necessarily reach the predicted value. If the backup adjustment margin is planned based on the maximum load forecast value, the system backup adjustment cost will be increased and the system operation economy will be reduced.

[0064] In fact, the reserve regulation margin can be dynamically adjusted according to the load demand. The reserve condition margin demand is different from the conventional load demand, and the reserve regulation margin demand has great flexibility. The reserve regulation margin flexibility can be determined by the load occurrence probability, the load possible interval, and the reserve regulation margin demand proportion. Specifically, the load demand can be described in the form of a probability interval prediction, such as: the load prediction value has a 68% probability in [90, 120] or a 90% probability in [80, 125]; correspondingly, the reserve regulation margin has a 68% probability in [90 25%, 120 30%] and a 90% probability in [80 25%, 125 30%]. In addition, the reserve regulation margin demand proportion (conventional 25-30%) will also be adjusted according to the load occurrence probability. Therefore, the calculation of the probability interval prediction can provide the probability of the occurrence of uncertain things and the uncertain interval in which they are located, and can realize the characterization and depiction of the reserve regulation margin flexibility. Therefore, the embodiment proposes a virtual power plant reserve regulation margin quantification and flexibility aggregation method based on a QR-SiameseLSTM neural network. This method no longer sets the reserve regulation margin interval based on the fixed maximum load prediction value as the base value, but considers the load demand occurrence probability, possible interval, required reserve regulation margin proportion, and probability interval prediction results of flexible resources. Under the premise of ensuring system safety, the virtual power plant flexibility aggregation is realized.

[0065] The reason for selecting the QR-SiameseLSTM model to construct the prediction in this embodiment is that deep learning technology has shown strong ability in time series prediction and probability modeling, and has significant advantages in handling uncertain data and correlation analysis. The QR-SiameseLSTM model can learn the probability distribution of load demand based on historical data, thereby generating a probability interval prediction of load demand. The SiameseLSTM neural network is good at capturing the similarity between different data points and is suitable for probability interval prediction and aggregation analysis of flexible resources.

[0066] According to the above principle, the embodiment proposes a power plant flexibility aggregation method considering reserve margin combining probability interval prediction and its aggregation projection, realizes the regulation potential probability interval prediction of various flexible resources, and realizes the ordered aggregation of the virtual power plant based on the probability interval center value of each type of flexible resource. Finally, by using interval analysis method, considering the dynamic relationship between the flexible resource output probability and the reserve regulation margin probability interval, the virtual power plant flexibility space is solved.

[0067] Please refer to Figure 1 The power plant flexibility aggregation method considering reserve margin includes the following steps:

[0068] S1, predicting the flexible resource adjustment potential probability interval by quantile regression, Siamese long short-term memory constructed;

[0069] Step S1 realizes the prediction of the flexible resource adjustment potential probability interval based on the QR-SiameseLSTM neural network model. First, the long short-term memory (LSTM) neural network is suitable for solving the long-term dependence problem in the training process. The LSTM network contains multiple cell units to process sequence information, and each cell unit contains the structure of the forget gate, the input gate and the output gate. This structure can effectively remember the time-varying dynamic behavior of the data and fully demonstrate the ability to capture the time coupling and nonlinear characteristics of flexible resources. SiameseLSTM is a special neural network evolved on the basis of LSTM network units, which is composed of two LSTM neural networks with shared weights and the same structure and ).

[0070] The QR-SiameseLSTM model described above includes a first LSTM neural network, a second LSTM neural network and a quantile regression layer;

[0071] The first LSTM neural network is used to output the encoded load elasticity characteristics according to the demand side load sequence data, time-of-use electricity price, weather parameters and time information;

[0072] The second LSTM neural network is used to output the elasticity vector according to the encoded load elasticity characteristics;

[0073] Wherein, the first LSTM neural network , the second LSTM neural network The hidden state of each cell unit , can be represented as:

[0074] ,

[0075] ,

[0076] Wherein and are input data of and , respectively, and are weight parameters of the two neural networks, respectively.

[0077] The similarity between the two hidden state sequences is calculated using the cosine similarity function:

[0078] .

[0079] In summary, the Siamese twin neural network structure of this embodiment includes two identical sub-networks that share weights and architecture and capture load elasticity characteristics through a dual-channel LSTM structure.

[0080] Specifically, the first neural network It is used to encode the load elasticity characteristics and analyze the reserve adjustment margin of the virtual power plant's multiple flexible resources, such as wind power generation, photovoltaic power generation, and controllable loads, in the face of the scenario of reduced grid operation flexibility. The input data includes demand-side load series data, time-of-use electricity prices, weather parameters, and time information. The second neural network It is used to estimate the elastic vector that changes with time and reflects the impact of time-varying cumulative error on prediction accuracy. A new density layer, the Sigmoid function, is introduced to transform The encoding vector is converted into an elastic vector : ,in, represents the Sigmoid activation function, represents the weight matrix, Represents the bias vector.

[0081] The output elastic vector can be used to measure the similarity between the predicted value and the observed value at each time step. It is also used as the memory information of the LSTM network in the form of prediction error and is input into the parameter update of the next recurrent unit. It aims to achieve real-time rolling parameter optimization in multi-step prediction. Through the dual-channel structure characteristics of the Siamese neural network, the similarity between the predicted value and the observed value is measured, and rolling optimization parameters in multi-step prediction are achieved to reduce prediction error. The specific formula is as follows:

[0082] ,in, is the learning rate, is the loss function, which takes into account the impact of prediction error and elastic vector, express The model parameters at time t, express The above model structure can significantly improve the prediction accuracy of elastic interval.

[0083] The quantile regression layer is used to output the confidence interval of the load forecast;

[0084] Specifically, for each training set output result, quantile regression is used to estimate the upper and lower boundaries of the prediction interval. The confidence level of the new output observation value belonging to the prediction interval is set to 1- i, when the input distribution is , the output The conditional quantile regression function under the quantile can be expressed as:

[0085] ,

[0086] is the input sample set, is the output Probability distribution function, its probability density can be estimated from all the training data set corresponding to the conditional quantile regression function. At this time, the output probability interval length is 1- , the output probability interval Can be determined by the upper and lower bounds of the two quantiles corresponding to the predicted value, as shown below:

[0087] ,

[0088] Where, and The predicted value corresponding to the quantile level is / 2 and / 2. For different input values, the interval width can be dynamically adjusted, so as to effectively adapt to the elastic change characteristics of the standby adjustment margin. Calculate The upper and lower boundary deviations of the original sequence and the new sequence under the quantile level / 2, these deviations can be regarded as And the approximate error of the actual interval

[0089] ,

[0090] Where, Indicates the new upper boundary sequence, is the predicted deviation value of Under the quantile level / 2; Indicates the quantile function under the quantile level / 2 of the calculation sequence.

[0091] According to the confidence interval, the flexible resource adjustment potential probability interval is calculated, specifically, the predicted deviation value is added to the original output interval, and the deviation of the original elastic probability interval is corrected to obtain the corresponding elastic probability interval under the set confidence level, and the calculation satisfies:

[0092] ,

[0093] ,​

[0094] wherein, and are the updated upper and lower confidence bounds, and are the upper and lower prediction bias at the quantile level of is the flexible resource regulation potential probability interval, and are the predicted values corresponding to the quantile levels of / 2 and / 2, respectively.

[0095] The prediction results of the above process can be referred to as shown in Figure 2 and Figure 3 Figure 2 shows the prediction results of the flexible load demand probability interval, Figure 3 shows the prediction results of the flexible resource regulation potential probability interval.

[0096] In summary, the embodiment considers the correlation between the reserve regulation margin and the flexibility of the resource, dynamically adjusts the reserve regulation margin demand proportion based on the probability of the flexible resource, and sets the reserve regulation margin interval based on the probability interval prediction results of the flexible resource.

[0097] In order to obtain the overall required reserve capacity of the power system, based on the probability interval prediction results of the flexible controllable load, the application will adopt a dynamic adjustment strategy for the reserve regulation margin demand proportion. Ignoring the time coupling factor, considering at time, the confidence of the prediction interval is 1- , and the load demand prediction value is the center value of the probability interval, that is , wherein and respectively represent the upper and lower confidence bounds at time. Therefore, after calculating the predicted flexible resource regulation potential probability interval, it also includes: calculating the reserve regulation set capacity interval, satisfying: , wherein, represents a probability occurrence conversion function that maps the confidence of the probability interval to the reserve regulation margin elasticity, the probability occurrence conversion function is an empirical formula, which is represented as a nonlinear function based on the quantile of the standard normal distribution, and the historical data is used to fit the key parameters; represents the confidence of the flexible resource probability interval, represents the center value of the probability interval, represents the quantile level of the prediction interval at time, ​The reserve regulation margin is set based on the flexible resource occurrence probability interval at the current time step. The higher the occurrence probability, the higher the reserve regulation margin demand proportion.

[0098] S2, according to the flexible resource regulation potential probability interval, the virtual power plant flexible resource aggregation modeling is carried out;

[0099] Step S2 realizes the ordered aggregation of virtual power plant flexible resources based on the characteristics of flexible resource probability interval, such as the aforementioned center value. Facing the reserve regulation demand of power grid, virtual power plant can provide upward power support for the system in real time. As an equivalent power plant, the reserve regulation capacity of virtual power plant follows the definition of power supply, that is, by reducing the discharge of its load or energy storage device to provide additional power to respond to frequency reduction. The power regulation capacity of virtual power plant is composed of the regulation capacity of each distributed energy resource inside it, and can be realized in two ways: it can allow multiple distributed power sources to participate in power regulation at the same time point, and this synchronous regulation can quickly respond to the demand changes of power grid; secondly, virtual power plant can also arrange distributed power sources to participate in regulation in a certain order, which allows virtual power plant to gradually adjust power output at different time points according to the changes of power grid conditions. In S2, according to the flexible resource regulation potential probability interval, the virtual power plant flexible resource aggregation modeling is carried out, including:

[0100] According to the number of flexible resources participating in regulation and the number of devices of each type of resource in the virtual power plant, the regulation capacity algebra sum of the virtual power plant is calculated.

[0101] This embodiment ignores the coupling factors in time sequence, and the regulation capacity algebra sum of the virtual power plant is as follows:

[0102]

[0103] In the formula, and respectively, the number of flexible resources participating in regulation and the number of devices of each type of resource in the virtual power plant. is the maximum upward regulation margin that the i-th device of the j-th type of resource can provide according to the scheduling instruction and its own condition constraints, represents the starting time, represents the time span, represents the upward regulation direction, represents the total regulation capacity of the virtual power plant at t time.

[0104] ​​​In virtual power plants, various resources are distributed over a wide area, and individual regulation margins are limited. Therefore, a certain scale of flexible resources needs to be aggregated and collectively regulated to optimize the elasticity of the power system, achieving more efficient resource utilization and more stable grid operation. This embodiment proposes a new method for evaluating the aggregate elasticity interval of virtual power plants. This method models the aggregate elasticity interval evaluation problem as a high-dimensional polyhedron projection problem. A dimensionality-reducing polyhedron of a specific shape is used to approximate the aggregate elastic power interval of the virtual power plant.

[0105] In this embodiment, the adjustable potential of each type of flexible resource is mapped as a central value to the aggregated external adjustable characteristics of the virtual power plant, participating in the elastic optimization of the virtual power plant and transforming the uncertainty optimization model into a deterministic optimization model. Specifically, the adjustable potential interval characteristics of the flexible resource are mapped to the aggregated external adjustable characteristics of the virtual power plant for aggregation calculation to meet the following requirements:

[0106] ,

[0107] in, Before optimization The predicted load capacity at each moment can be obtained by traditional time series forecasting methods, that is, the capacity of the optimization period is predicted based on the trend of historical load data. for The moment-to-moment elastic aggregation capacity, i.e. the regulation potential predicted by S1, Indicates the upper and lower limits of adjustable potential. Figure 4 The diagram below shows the results of the ordered aggregation of virtual power plants. Based on the daily load probability forecast, the flexible aggregation capacity of virtual power plants is determined across 24 time periods with the goal of smoothing the daily load curve. This aggregation scheme, based on the center value of the flexible resource probability interval, can achieve the ordered aggregation of virtual power plants and optimize resource allocation.

[0108] The externally adjustable feature mentioned above refers to aggregating the results of S1 based on the central value, and roughly obtaining the capacity of the virtual power plant regulation.

[0109] The constraints of aggregate calculation include technical constraints combining linear network constraints with flexible resource models, and constraints on the aggregate regulation potential of virtual power plants;

[0110] The linear network constraints and the flexible resource model are combined to form the technical constraints of the virtual power plant, which meet the following requirements:

[0111] ,

[0112] ,

[0113] ,

[0114] wherein, is the vector of decision variables at time instant, the decision variables refer to the adjustment potential of various flexible resources, , denotes the matrix and vector form of the set of technical constraints, , denotes the matrix and vector form of the set of network constraints, denotes the virtual power plant output power trajectory, , denotes the matrix and vector form of the constant matrix of output power,

[0115] Based on the flexible resource output probability model, the central value of the virtual power plant aggregated adjustment potential is calculated, so that the virtual power plant aggregated adjustment potential constraint satisfies: wherein, and are the predicted values corresponding to the quantile levels of / 2 and / 2 respectively, denotes the central value.

[0116] S3, based on error propagation, solves the aggregation modeling of the virtual power plant flexible resource, and obtains the virtual power plant reserve adjustment margin elastic interval.

[0117] S3, based on the virtual power plant aggregation target, models the multi-type and heterogeneous distributed resource aggregation problem as a high-dimensional polyhedron projection problem, and solves the multi-probability interval superposition quantization problem based on the error propagation law, in order to solve the virtual power plant elastic space.

[0118] Specifically, based on error propagation, the aggregation modeling of the virtual power plant flexible resource is solved to obtain the virtual power plant reserve adjustment margin elastic interval, including:

[0119] In a single time period, the virtual power plant active power flexibility can be represented as a polyhedron in the feasible region, and the virtual power plant active power flexibility is modeled as a multi-face projection, and its expression satisfies: wherein, and denote the calculation parameters of the virtual power plant active elastic interval, is the expression of the projection polyhedron at time period, contains the calculation parameters of the VPP such as the active and reactive power levels of the device, the ramp rate, etc. denotes the constant term on the right side, which defines the limit value that the output power of the VPP can reach after considering various technical constraints.

[0120] ​Considering the dynamic relationship between reserve regulation margin and flexible resources, the aggregation interval of virtual power plant needs to be flexibly adjusted, and the adjustable characteristics outside the virtual power plant are mapped to more flexible elastic space. Based on this, the embodiment classifies according to the characteristics of the constraints on the flexible resource cluster technology, and realizes the elastic aggregation of the virtual power plant by using two types of approximate models of virtual generators and virtual batteries. When the flexible resource regulation potential probability is low, the virtual battery model is mainly used for aggregation; when the flexible resource regulation potential probability is high, the reliable power supply needs to be increased to improve the safety and stability of the power system, and the virtual generator model is mainly used for aggregation.

[0121] The constraint conditions of the aggregation calculation are optimized and modeled, and the coupling constraints are eliminated by a robust optimization algorithm, so that the uncertain parameters in the constraint conditions are converted into a deterministic problem.

[0122] Specifically, the same cluster composed of isomorphic resources has similar characteristics in technical constraints, and the same method can be used for modeling, and the coupling constraints between the two types of resources are eliminated by using a robust optimization algorithm. Specifically, by introducing balance constraints and corresponding auxiliary variables, the constraint conditions containing uncertain parameters are converted into explicit linear or quadratic constraint conditions. The problem after the conversion can be solved by using conventional optimization techniques.

[0123] In order to simplify the model, the decision variable is divided into three sub-variables, namely the active power of the virtual generator, the active power of the virtual battery, and the reactive power of all distributed resources, and the specific formula is as follows:

[0124] , wherein, and represent the constraint matrix and the constraint vector at time , respectively.

[0125] Firstly, the is taken as an uncertain parameter in robust optimization, and the constraint conditions related to it are separated, so that the above formula is rearranged as:

[0126] ,

[0127] By using the robust counterpart method, the robust optimization problem can be converted into a deterministic problem that is easy to handle, and the specific formula is as follows:

[0128] ,

[0129] ,

[0130] wherein, denote the different types of constraint matrices after total constraint separation, denote the different types of constraint vectors.

[0131] The decoupled network constraints of the virtual generator at the last time period can be simplified as:

[0132] where, denotes the time-decoupled network constraint matrix, denotes the virtual generator output power, denotes the time-decoupled network constraint vector.

[0133] Similarly, the decoupled network constraints of the virtual battery can be expressed as:

[0134] where, denotes the decoupled network constraint matrix and constraint vector of the virtual battery.

[0135] Thus, the construction of the decoupled network constraints of the two types of homogeneous distributed power sources is completed.

[0136] Then, considering all the technical constraints of the virtual generator:

[0137] where, denotes the constraint matrix, denotes the constraint vector.

[0138] and the active output power of the virtual generator is expressed in the following affine form:

[0139] where, denotes the active output power trajectory, denotes the active power conversion matrix, denotes the active power bias vector.

[0140] Thus, the origin polytope and the corresponding projection polytope are:

[0141]

[0142]

[0143] where, denotes the virtual generator output power feasible region, denotes the virtual generator output power feasible region after dimensionality reduction projection, denotes the virtual generator output power,​​​​​​ constraint matrix of the projection polyhedron, constraint vector of the projection polyhedron.

[0144] This embodiment uses an inscribed polyhedron similar to the target polyhedron for approximation quantization to simplify the complex calculation process, and uses interface shrinkage technology to gradually optimize the polyhedron boundary. The advantage of this method is that it can retain the key characteristics of the generator model, such as the power output of the generator in a single time period , whose elastic characteristics can be described by other key parameters. These parameters include the ramping limit and in each time period and the upper and lower limits of the power output. In this way, the aggregated range of the virtual generator can be described and analyzed more effectively.

[0145]

[0146] Further simplified to the tight form:

[0147] Here is a constant matrix, is a key definition parameter vector, and the specific formula is as follows:

[0148]

[0149]

[0150]

[0151] Then, the aggregated elastic range of the virtual battery can be solved by using a similar method, including the energy constraints in each time period and and the power upper and lower limits and The formula is as follows:

[0152]

[0153] Further simplified to the tight form:

[0154]

[0155] Here is a constant matrix, is a key definition parameter vector, and the specific formula is as follows: ​​​​​​

[0156]

[0157] ,

[0158] .

[0159] Based on the time period and the optimized constraint condition, an optimized aggregation calculation formula is constructed, and the aggregation calculation formula is:

[0160] wherein, represents a probability conversion function, represents an affine value of virtual power plant output power, represents an affine value of virtual battery output power, represents an affine value of active output power of virtual generator, represents an aggregated flexibility range of virtual battery, represents an aggregated range of virtual generator, represents an aggregated range of virtual power plant;

[0161] The aggregation calculation formula is solved to obtain a virtual power plant reserve margin flexibility interval.

[0162] Based on the probability prediction result of flexible resources, the variables in the above-mentioned flexibility space are described in the form of an interval, and the deterministic aggregation problem of flexible resources is converted into a probability interval superposition solving problem. The above-mentioned model is solved based on the error propagation law. Since the output probability prediction of flexible resources has an error interval, the flexibility space mapping also contains an error interval, which can be explained by the error propagation law. If the prediction error is defined as , the function error is defined as , and the error propagation law is:

[0163] ,

[0164] In the formula, is a transfer function from the prediction error to the function error, , respectively represent the observation value and the prediction value.

[0165] Suppose that the confidence interval of the output prediction of flexible resources is According to the interval analysis method, it can be expressed in the form of an affine:

[0166] wherein, and represent the interval center value and the interval radius.

[0167] wherein, ; denotes the independent variable of the noise source, in the embodiment .

[0168] The power variable is represented by interval center value and interval radius in the elastic space mapping, and the model is solved. The superposition problem of flexible resource probability interval can be simplified and decomposed into two core steps: the first step is to participate in the elastic optimization of the virtual power plant with the interval center value as the main body to meet the rigid load demand in the day-ahead prediction; the second step is to solve the elastic interval of the standby adjustment margin, and consider the error propagation influence caused by the potential uncertainty in the interval radius. The interval analysis method based on the error propagation law used in the application can effectively solve the interval expansion problem and is suitable for analyzing the result interval in the interval aggregation problem. The center value variable of the flexible resource output prediction participates in the solution of the virtual power plant aggregation feasible region model, and the standby adjustable margin center value under the set confidence level 1- is obtained.

[0169] In summary, the aggregation calculation formula is solved, including:

[0170] The aggregation calculation is described by error interval propagation, and the aggregation formula for solving is obtained:

[0171] wherein, denotes the interface shrinkage equation of the polyhedral projection, and the shrinkage equation is a quantitative description of the boundary after the dimension reduction projection of the high-dimensional aggregation range, i.e. the feasible region of the variable, denotes the derivative of , which represents the rate of change of the polyhedral projection interface with the variable, denotes the noise element of the variable, , denotes the virtual power plant center value variable and radius variable;

[0172] The virtual power plant standby margin elastic interval is solved.

[0173] Based on the ordered aggregation result of the central value variable of the flexible resource output model in step S2, the rigid load capacity of the power system after optimization of flexibility can be obtained, and the reserve regulation margin capacity considering the occurrence probability of the flexible resource is set. Due to the interval characteristics of the flexible load, the embodiment adopts an interval analysis method to depict the flexibility of the reserve regulation margin. According to the occurrence probability and possible interval of the flexible resource, considering the cost investment benefit of the reserve regulation of the power system, the interval lower bound of the reserve regulation margin is determined at a lower output probability (such as 90%). Considering the safety and stability of the power system, the interval upper bound of the reserve regulation margin is determined at a higher output probability (such as 95%), that is, a more abundant interval value. Through the dynamic relationship between the flexible resource and the reserve regulation margin probability interval, the virtual power plant flexibility space is solved, and more accurate and efficient flexible optimization scheduling is realized.

[0174] Please refer to the optimization result diagram shown in Figure 5 , which is the horizontal line of the confidence level set to 90% and 95%. Based on this, the virtual power plant probability interval is solved.

[0175] Embodiment Two

[0176] Embodiment Two discloses a device corresponding to the power plant flexibility aggregation method considering the reserve margin of the above-mentioned embodiment. For the virtual device structure of the above-mentioned embodiment, please refer to Figure 6 , which includes:

[0177] The potential prediction module 210 is used to predict the flexible resource regulation potential probability interval through the quantile regression QR-SiameseLSTM model constructed by the quantile regression and the twin long short-term memory.

[0178] The aggregation module 220 is used to aggregate and model the virtual power plant flexible resource according to the flexible resource regulation potential probability interval.

[0179] The solving module 230 is used to solve the aggregation modeling of the virtual power plant flexible resource based on error propagation to obtain the reserve regulation margin flexibility interval of the virtual power plant.

[0180] Preferably, the QR-SiameseLSTM model includes a first LSTM neural network, a second LSTM neural network, and a quantile regression layer.

[0181] The first LSTM neural network is used to output the coded load flexibility characteristics according to the demand side load sequence data, the time-of-use electricity price, the weather parameter, and the time information.

[0182] The second LSTM neural network is used to output the flexibility vector according to the coded load flexibility characteristics.

[0183] The quantile regression layer is used to output the confidence interval of the load forecast;

[0184] The probability interval of flexible resource adjustment potential is calculated based on the confidence interval, which satisfies:

[0185] ,

[0186] ,

[0187] in, and are the updated upper and lower bounds of confidence, and They are / 2 quantile level upper and lower boundary prediction deviations, is the probability interval of flexible resource adjustment potential, and are the quantile levels in / 2 and / 2 corresponding predicted value.

[0188] Preferably, after calculating the probability interval of the flexible resource adjustment potential, the method further includes: calculating the reserve adjustment set capacity interval to meet the following requirements: ,in, represents the occurrence probability conversion function, represents the confidence level of the probability interval of flexible resources, represents the center value of the probability interval, Indicates The quantile level of the prediction interval at time .

[0189] Preferably, the aggregation modeling of the flexible resources of the virtual power plant is performed according to the flexible resource adjustment potential probability interval, including:

[0190] According to the number of flexible resources involved in regulation and the number of equipment of each type of resources in the virtual power plant, the algebraic sum of the regulation capacity of the virtual power plant is calculated;

[0191] The adjustable potential interval characteristics of flexible resources are mapped to the aggregated external adjustable characteristics of the virtual power plant for aggregation calculation to meet the following requirements:

[0192] ,

[0193] in, Before optimization The forecast load capacity at the time, for Flexible aggregation capacity at all times, Indicates the upper and lower limits of adjustable potential.

[0194] Preferably, the constraint conditions of the aggregated calculation include linear network constraints and flexible resource model combined technical constraints, and virtual power plant aggregated regulation potential constraints.

[0195] The technical constraints satisfy:

[0196] ,

[0197] ,

[0198] ,

[0199] wherein, is a vector of decision variables at time, , , represents a matrix and a vector form of a set of technical constraints, , represents a matrix and a vector form of a set of network constraints, represents a virtual power plant output power trajectory, , represents a matrix and a vector form of an output power constant coefficient matrix.

[0200] The virtual power plant aggregated regulation potential constraint satisfies: wherein, and are prediction values corresponding to quantile levels of / 2 and / 2 respectively, represents a central value.

[0201] Preferably, the aggregated modeling of the virtual power plant flexible resource is solved based on error propagation, to obtain a virtual power plant standby regulation margin elasticity interval, including:

[0202] The virtual power plant active power flexibility is modeled as a multi-face projection, and the expression satisfies: wherein, and represent calculation parameters of a virtual power plant active elasticity interval

[0203] The constraint conditions of the aggregated calculation are optimized and modeled, and the coupling constraints are eliminated through a robust optimization algorithm, so that the uncertain parameters in the constraint conditions are converted into a deterministic problem.

[0204] According to the multi-face projection and the optimized constraint conditions, an optimized aggregated calculation formula is constructed, and the aggregated calculation formula is:

[0205] wherein, representing the probability conversion function, representing the virtual power plant active power output affine value, representing the virtual battery active power output affine value, representing the virtual generator active power output affine value, representing the virtual battery aggregated elasticity range, representing the virtual generator aggregated range, representing the virtual power plant aggregated range;

[0206] solving the aggregated calculation formula to obtain the virtual power plant reserve margin elasticity interval.

[0207] Preferably, solving the aggregated calculation formula comprises:

[0208] describing the aggregated calculation by error interval propagation to obtain the aggregated formula for solving:

[0209] wherein, representing the interface shrink equation of the polyhedral projection, which is a boundary quantitative description after the high-dimensional aggregated range is projected in a lower dimension, representing the noise source of the variable , , representing the interval center value and interval radius variable;

[0210] solving to obtain the virtual power plant reserve margin elasticity interval.

[0211] Embodiment three

[0212] Figure 7 A structural schematic diagram of an electronic device provided in the embodiment three of the present application, as shown in the figure, the electronic device comprises a processor 310, a memory 320, an input device 330 and an output device 340; the number of processors 310 in the computer device can be one or more, Figure 7 wherein the processor 310 is taken as an example; the processor 310, the memory 320, the input device 330 and the output device 340 in the electronic device can be connected through a bus or other ways, Figure 7 wherein the connection through the bus is taken as an example. Figure 7

[0213] ​The memory 320, as a computer readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the power plant flexibility aggregation method considering reserve margin in the embodiments of the present application. The processor 310 performs various function applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 320, that is, implements the power plant flexibility aggregation method considering reserve margin in the above embodiment one.

[0214] The memory 320 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system and application programs required by at least one function; the data storage area can store data created according to the use of the terminal and the like. In addition, the memory 320 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state memory device. In some examples, the memory 320 can further include a memory remotely arranged with respect to the processor 310, which can be connected to the electronic device through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0215] The input device 330 can be used to receive input user identity information, virtual power plant data and the like. The output device 340 can include a display device such as a display screen.

[0216] Embodiment four

[0217] The embodiment four of the present application also provides a storage medium containing computer executable instructions, which can be used for a computer to execute the power plant flexibility aggregation method considering reserve margin, and the method comprises:

[0218] The QR-SiameseLSTM model constructed by quantile regression and twin long short-term memory is used to predict the flexible resource adjustment potential probability interval;

[0219] According to the flexible resource adjustment potential probability interval, the virtual power plant flexible resource aggregation modeling is performed;

[0220] The error propagation is used to solve the virtual power plant flexible resource aggregation modeling, and a virtual power plant reserve adjustment margin flexibility interval is obtained.

[0221] Of course, the computer executable instructions of the storage medium containing computer executable instructions provided by the embodiment of the present application are not limited to the method operations as described above, but can also perform the related operations in the power plant flexibility aggregation method considering reserve margin provided by any embodiment of the present application.

[0222] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary universal hardware, and of course can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH memory, a hard disk, or an optical disk, etc., and includes a number of instructions to make an electronic device (which can be a mobile phone, a personal computer, a server, or a network device, etc.) execute the methods described in various embodiments of the present application.

[0223] It is worth noting that in the above embodiments of the power plant elasticity aggregation method and device based on consideration of backup margin, each unit and module included is only divided according to functional logic, but is not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional unit are only for easy mutual differentiation, and do not limit the protection scope of the present application.

[0224] For those skilled in the art, various corresponding changes and modifications can be made to the above described technical solutions and concepts, and all these changes and modifications should belong to the protection scope of the claims of the present application.

Claims

1. A power plant elastic aggregation method considering reserve margin, characterized in that: The following steps are involved: A QR-Siamese LSTM model constructed using quantile regression and twin long short-term memory is used to predict the probability interval of flexible resource regulation potential. The QR-Siamese LSTM model includes a first LSTM neural network, a second LSTM neural network, and a quantile regression layer. The first LSTM neural network is used to output coded load elasticity characteristics based on demand-side load sequence data, time-of-use electricity prices, weather parameters, and time information. The second LSTM neural network is used to output an elasticity vector based on the coded load elasticity characteristics. The quantile regression layer is used to output a confidence interval for the load forecast. The probability interval of the flexible resource regulation potential is calculated based on the confidence interval to meet the following requirements: ， ， in, and are the updated upper and lower bounds of confidence, and They are / 2 quantile level upper and lower boundary prediction deviations, is the probability interval of flexible resource adjustment potential, and are the quantile levels in / 2 and / 2 corresponding predicted value; After calculating the probability interval of flexible resource adjustment potential, the following is also included: calculating the reserve adjustment set capacity interval to meet the following requirements: ,in, represents the occurrence probability conversion function, represents the confidence level of the probability interval of flexible resources, represents the center value of the probability interval, Indicates The quantile level of the prediction interval at time t; Based on the flexible resource regulation potential probability interval, aggregate modeling of the virtual power plant flexible resources is performed, including: According to the number of flexible resources involved in regulation and the number of equipment of each type of resources in the virtual power plant, the algebraic sum of the regulation capacity of the virtual power plant is calculated; The adjustable potential interval characteristics of flexible resources are mapped to the aggregated external adjustable characteristics of the virtual power plant for aggregation calculation to meet the following requirements: ， in, Before optimization The forecast load capacity at the time, for Flexible aggregation capacity at all times, Indicates the upper and lower limits of adjustable potential; The constraints of the aggregate calculation include technical constraints combining linear network constraints and flexible resource models, and constraints on the aggregate regulation potential of virtual power plants; The technical constraints satisfy: ， ， ， in, For decision variables The vector of time, , The matrix and vector forms of the technical constraint set, , Represents the matrix and vector form of the network constraint set, represents the output power trajectory of the virtual power plant, , Represent the matrix and vector form of the output power constant coefficient matrix; The virtual power plant aggregate regulation potential constraint satisfy: ,in, and are the quantile levels in / 2 and / 2 corresponding to the predicted value, represents the central value; The aggregation modeling of the flexible resources of the virtual power plant is solved based on error propagation to obtain the elastic range of the standby regulation margin of the virtual power plant.

2. The power plant elastic aggregation method considering reserve margin according to claim 1, characterized in that: Based on error propagation, the aggregation modeling of the flexible resources of the virtual power plant is solved to obtain the elastic range of the reserve regulation margin of the virtual power plant, including: The active power flexibility of the virtual power plant is modeled as a multi-faceted projection, and its expression satisfies: ,in, and Calculation parameters representing the active power flexibility range of the virtual power plant; Optimize and model the constraints of aggregate computing, eliminate coupling constraints through robust optimization algorithms, and transform uncertain parameters in the constraints into deterministic problems; According to the multi-faceted projection and the optimized constraint conditions, an optimized aggregation calculation formula is constructed, and the aggregation calculation formula is: ,in, represents the occurrence probability conversion function, represents the affine value of the active output power of the virtual power plant, represents the affine value of the virtual battery active output power, represents the affine value of the active output power of the virtual generator, represents the aggregate elastic range of the virtual battery, represents the aggregation scope of the virtual generator, represents the aggregation scope of the virtual power plant; The aggregation calculation formula is solved to obtain the virtual power plant reserve margin elasticity range.

3. The power plant elastic aggregation method considering reserve margin according to claim 2, characterized in that: Solving the aggregation calculation formula includes: The aggregation calculation is described by error interval propagation, and an aggregation formula that is convenient for solution is obtained: ,in, represents the interface contraction equation of the polyhedron projection, Representing variables The noise source, 、 Represents the interval center value and interval radius variables; The elastic range of the virtual power plant reserve margin is obtained by solving the problem.

4. A power plant elastic aggregation device considering reserve margin, characterized in that: It includes: A potential prediction module is configured to predict the probability interval of flexible resource regulation potential using a QR-Siamese LSTM model constructed using quantile regression and twin long short-term memory. The QR-Siamese LSTM model includes a first LSTM neural network, a second LSTM neural network, and a quantile regression layer. The first LSTM neural network is configured to output coded load elasticity features based on demand-side load sequence data, time-of-use electricity prices, weather parameters, and time information. The second LSTM neural network is configured to output an elasticity vector based on the coded load elasticity features. The quantile regression layer is configured to output a confidence interval for the load forecast. The probability interval of flexible resource regulation potential is calculated based on the confidence interval to satisfy the following requirements: ， ， in, and are the updated upper and lower bounds of confidence, and They are / 2 quantile level upper and lower boundary prediction deviations, is the probability interval of flexible resource adjustment potential, and are the quantile levels in / 2 and / 2 corresponding predicted value; After calculating the probability interval of flexible resource adjustment potential, the following is also included: calculating the reserve adjustment set capacity interval to meet the following requirements: ,in, represents the occurrence probability conversion function, represents the confidence level of the probability interval of flexible resources, represents the center value of the probability interval, Indicates The quantile level of the prediction interval at time t; An aggregation module is used to perform aggregation modeling of flexible resources of a virtual power plant according to the probability interval of the flexible resource adjustment potential; and includes: According to the number of flexible resources involved in regulation and the number of equipment of each type of resources in the virtual power plant, the algebraic sum of the regulation capacity of the virtual power plant is calculated; The adjustable potential interval characteristics of flexible resources are mapped to the aggregated external adjustable characteristics of the virtual power plant for aggregation calculation to meet the following requirements: ， in, Before optimization The forecast load capacity at the time, for Flexible aggregation capacity at all times, Indicates the upper and lower limits of adjustable potential; The constraints of the aggregate calculation include technical constraints combining linear network constraints and flexible resource models, and constraints on the aggregate regulation potential of virtual power plants; The technical constraints satisfy: ， ， ， in, For decision variables The vector of time, , The matrix and vector forms of the technical constraint set, , Represents the matrix and vector form of the network constraint set, represents the output power trajectory of the virtual power plant, , Represent the matrix and vector form of the output power constant coefficient matrix; The virtual power plant aggregate regulation potential constraint satisfy: ,in, and are the quantile levels in / 2 and / 2 corresponding to the predicted value, represents the central value; A solution module is used to solve the aggregation modeling of the flexible resources of the virtual power plant based on error propagation to obtain the elastic range of the standby regulation margin of the virtual power plant.

5. An electronic device comprising a processor, a storage medium, and a computer program, wherein the computer program is stored in the storage medium, When the computer program is executed by a processor, the method for power plant resilience aggregation considering reserve margin as described in any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for power plant resilience aggregation considering reserve margin as described in any one of claims 1 to 3 is implemented.

Images