Multi-user-side resource cooperative regulation potential evaluation method, system and product

Abstract

The invention belongs to the field of power systems, and provides a multi-user-side resource cooperative adjustment potential evaluation method, system and product, and the technical scheme is that the upward and downward adjustment potential of a single user-side resource is evaluated based on short-term prediction data and historical data of distributed photovoltaic and user-level flexible load user-side resources; then, considering the sensitivity of different user side resources to the electricity price signal, and correcting the adjustment potential; and finally, considering the cooperative effect of the aggregator internal user side resources, and evaluating and calculating the cooperative regulation potential of the aggregator internal user side resources. According to the method, the problems that a traditional multi-user-side resource adjustment capability evaluation method depends on equipment modeling and is large in workload are solved, and a new thought is provided for evaluation of the multi-user-side resource adjustment capability.

Description

Technical Field

[0001] The present invention belongs to the field of collaborative regulation potential assessment of user-side resources in power systems, and in particular, relates to a method, system and product for collaborative regulation potential assessment of multiple user-side resources. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the development of society and the adjustment of industrial structure, the proportion of new energy sources is increasing, and the peak-to-valley difference of the power grid is increasing. User-side resources, due to their flexible adjustment capabilities, can help solve many problems facing the current power system. Therefore, evaluating and predicting the adjustment potential of user-side resources is the key to optimizing the dispatching of the power grid.

[0004] To ensure the effective use of load regulation capabilities on the user side of multiple industries such as industry, commerce, and residents, it is necessary to understand their power characteristics. Therefore, how to accurately forecast their load is the key to studying the load regulation potential of users in different industries.

[0005] The current load forecasting methods are mainly divided into three categories: traditional methods, similar day methods, and artificial intelligence model prediction. Traditional load forecasting methods mainly analyze and count historical load data and its influencing factors, establish a mathematical model of load change trends, and predict load data in the future time period. As the scale of the power system increases, traditional methods are increasingly difficult to meet the needs of modern large systems. The similar day method is based on the periodic characteristics of power load and directly uses the load data of similar historical days as the prediction result. However, affected by factors such as social economy and climate change, the characteristics of power load are becoming increasingly complex and changeable. In this context, relying solely on similar days will not provide completely accurate information. In recent years, the rise of artificial intelligence algorithms such as artificial neural networks and ensemble learning has provided new ideas for load forecasting. Artificial intelligence algorithms can make full use of the advantages of machine learning to capture the complex nonlinear relationship between power load and various influencing factors. Through continuous training and updating, these algorithms can adapt to changes in load data and environmental conditions, so they have broad application prospects in the context of current massive load data. However, in current AI-based load forecasting research, the influencing factors are usually input into the forecasting model and the load forecast results are directly output, without considering the differences in load characteristics under different weather scenarios, and also without fully utilizing the advantages of the similar day method in determining load change trends. In addition, a single forecasting model cannot show good forecasting performance in all scenarios.

[0006] Distributed photovoltaic power prediction on the user side has been one of the research hotspots in recent years. Accurate power prediction is of great significance for evaluating its regulation potential. Traditional photovoltaic power prediction methods mainly include physical models and data-driven methods. The physical model is based on the physical characteristics of the photovoltaic system, but it is highly dependent on meteorological data and has high model complexity. The data-driven method trains the model through historical data and has strong adaptability and flexibility. In recent years, with the development of machine learning and deep learning technologies, a variety of data-driven methods have been applied to distributed photovoltaic power prediction. For example, methods such as support vector machine (SVM), artificial neural network (ANN), and extreme learning machine (ELM) have achieved good results in photovoltaic power prediction. However, these methods often face problems such as large data volume and high computational complexity when dealing with large-scale distributed photovoltaic power stations.

[0007] At present, research has been carried out on the adjustable potential of multiple user-side resources. Some scholars have proposed an evaluation method for the adjustable potential of power system loads based on interruptibility, portability and product characteristics, and evaluated the adjustable potential of industrial users based on the correlation weighting method. Some scholars have constructed a residential community load response model and used the probability distribution method to predict the adjustable potential of residential community loads. Other scholars have constructed an adjustable potential evaluation index system containing multiple characteristics by exploring the adjustment potential of different types of industrial users, and used the VMD-TCN potential analysis method to carry out industrial load potential evaluation. However, the current evaluation of the adjustable potential of multiple user-side resources is mostly based on the perspective of equipment models. This method requires the establishment of a large number of equipment models. The parameters required by the equipment model, such as equipment type, quantity, and operating characteristics, are numerous and difficult to obtain, which becomes a major problem in evaluating adjustable potential. Summary of the invention

[0008] In order to solve at least one technical problem existing in the above-mentioned background technology, the present invention provides a method, system and product for evaluating the potential of coordinated regulation of resources on the side of multiple users. It breaks through the difficult problem that the traditional method for evaluating the regulation capability of resources on the side of multiple users relies on device modeling and has a large workload, and efficiently completes the evaluation of the adjustable capability of resources on the side of multiple users.

[0009] In order to achieve the above object, the present invention adopts the following technical solution: A first aspect of the present invention provides a method for evaluating the potential of collaborative regulation of multiple user-side resources, comprising the following steps: Based on the acquired power load characteristic data of various users, similar daily load data and the constructed multi-user integrated learning load forecasting model, the load forecast results for each type of user are obtained; Distributed photovoltaic power stations are clustered, and the source domain and target domain are determined based on the clustering results. A short-term photovoltaic power prediction model based on the source domain data is established and trained. The features learned from the short-term photovoltaic power prediction model based on the source domain data are transferred to the target domain to obtain a short-term photovoltaic power prediction model suitable for the target domain, and the distributed photovoltaic power data is further predicted. The regulation potential of a single user-side resource is established by combining multiple user load data and distributed photovoltaic data. The sensitivity of different user-side resources to electricity price signals is considered to correct the regulation potential. The synergy of user-side resources within the aggregator is considered, and the coordinated regulation potential of user-side resources within the aggregator is calculated to obtain the coordinated regulation potential.

[0010] Furthermore, the load forecasting result for each type of user is obtained based on the acquired power load characteristic data of each type of user, similar daily load data and the constructed multi-user integrated learning load forecasting model, including: Based on the acquired power load characteristic data of various users and similar daily load data, a training set and a prediction set are constructed, multiple base learners are trained based on the training set, and then the trained base learners are predicted based on the prediction set to obtain a prediction result data set; The prediction result dataset is used as the secondary training dataset and the prediction dataset. The meta-learner is trained based on the secondary training dataset. The load prediction results of various users are obtained based on the prediction dataset and the trained meta-learner.

[0011] Furthermore, the distributed photovoltaic power stations are clustered and the source domain and the target domain are determined based on the clustering results, including clustering the distributed photovoltaic power stations using an AP clustering algorithm, defining the cluster center of each cluster as the source domain according to the clustering results, and classifying the remaining power stations as the target domain.

[0012] Furthermore, the short-term photovoltaic power prediction model adopts Transformer neural network.

[0013] Furthermore, when the features learned by the short-term photovoltaic power prediction model based on the source domain data are transferred to the target domain, the model parameters are adjusted with the minimum RMSE and MAPE as the goal.

[0014] Furthermore, the regulation potential of single user side resources is established by combining multiple user load data and distributed photovoltaic data, including: Calculate the baseline output of each user-side resource in different seasons; Satisfaction Maximum value in principle and minimum value As t Maximum baseline output and minimum baseline output at the moment; Based on the future t The predicted values ​​of different types of user-side resources and the maximum baseline output at each moment are used to calculate the flexibility increase potential of different types of user-side resources. t The predicted values ​​and minimum baseline outputs of different categories of user-side resources at each moment are used to obtain the flexibility reduction potential of different categories of user-side resources.

[0015] Furthermore, the adjustment potential after correction is: , , in, and is the revised upward and downward potential, and Before correction t The potential for upward and downward adjustment of resource flexibility for different categories of users at all times. is the load demand at time t before demand response, is the load demand change before and after participating in demand response at time t.

[0016] Furthermore, the synergistic regulation potential is: , , in, User-side resources i The revised upside potential, User-side resources i The revised downside potential, For distributed photovoltaic, For load.

[0017] A second aspect of the present invention provides a multi-user side resource collaborative adjustment potential evaluation system, comprising: The load forecasting module is used to obtain the load forecasting results for each type of user based on the acquired power load characteristic data of each type of user, similar daily load data and the constructed multi-user integrated learning load forecasting model; The photovoltaic power prediction module is used to cluster distributed photovoltaic power stations, determine the source domain and target domain based on the clustering results, establish and train a short-term photovoltaic power prediction model based on the source domain data; transfer the features learned by the short-term photovoltaic power prediction model based on the source domain data to the target domain, obtain a short-term photovoltaic power prediction model suitable for the target domain, and further predict the distributed photovoltaic power data; The regulation potential assessment module is used to establish the regulation potential of a single user-side resource by combining multiple user load data and distributed photovoltaic data, consider the sensitivity of different user-side resources to electricity price signals, correct the regulation potential, consider the synergy of user-side resources within the aggregator, and calculate the coordinated regulation potential of user-side resources within the aggregator to obtain the coordinated regulation potential.

[0018] A third aspect of the present invention provides a program product.

[0019] A program product, which is a computer program product, includes a computer program, and when the computer program is executed by a processor, it implements the steps in the method for evaluating the potential for collaborative regulation of multiple user-side resources as described above.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention evaluates the upward and downward regulation potential of a single user-side resource based on the short-term forecast curves and historical data of distributed photovoltaics, user-level flexible loads and other user-side resources; then, the sensitivity of different user-side resources to electricity price signals is considered to correct the regulation potential; finally, the synergy of user-side resources within the aggregator is considered to evaluate and calculate the synergistic regulation potential of user-side resources within the aggregator. This method breaks through the difficulty of traditional multi-user-side resource regulation capacity evaluation methods that rely on equipment modeling and have a large workload, and provides a new idea for the evaluation of the adjustable capacity of multi-user-side resources.

[0021] 2. Taking into account that a single model cannot achieve good prediction performance in various prediction scenarios, the present invention establishes an integrated learning load prediction model for each type of user, including industrial, commercial, and residential users, and fuses multiple base learners to obtain a strong learning model, thereby improving the prediction accuracy.

[0022] 3. In order to solve the problem that each site in the distributed photovoltaic power prediction model has large calculation requirements for individual prediction, the present invention proposes a distributed photovoltaic power prediction method based on AP clustering transfer learning, constructs a short-term power prediction model based on Transformer neural network, and avoids redundant training costs through AP clustering-based transfer learning, thereby effectively improving the prediction accuracy of distributed photovoltaic power prediction in complex prediction scenarios.

[0023] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0025] Figure 1 It is a flow chart of a method for evaluating the potential of collaborative regulation of multiple user-side resources provided by an embodiment of the present invention; Figure 2 It is a flow chart of load forecasting based on multi-model stacking ensemble learning provided by an embodiment of the present invention; Figure 3 The present invention provides a transfer learning flow chart. DETAILED DESCRIPTION

[0026] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0027] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0028] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0029] As mentioned in the background technology, the current evaluation of the adjustable potential of multiple user-side resources is mostly based on the perspective of equipment models. This method requires the establishment of a large number of equipment models. The parameters required for the equipment model, such as equipment type, quantity, and operating characteristics, are numerous and difficult to obtain, which becomes a major problem when evaluating adjustable potential. The present invention evaluates the upward and downward adjustment potential of a single user-side resource based on the short-term prediction curves and historical data of user-side resources such as distributed photovoltaics and user-level flexible loads; then, the sensitivity of different user-side resources to electricity price signals is considered to correct the adjustment potential; finally, the synergy of user-side resources within the aggregator is considered, and the coordinated adjustment potential of user-side resources within the aggregator is evaluated and calculated. This method breaks through the problem that the traditional method for evaluating the adjustment capacity of multiple user-side resources relies on equipment modeling and has a large workload, and provides a new idea for the evaluation of the adjustable capacity of multiple user-side resources.

[0030] Embodiment 1 like Figure 1 As shown, this embodiment provides a method for evaluating the potential of collaborative regulation of multiple user-side resources, including the following steps: Step 1: Obtain the load forecast results for each type of user based on the acquired power load characteristic data and the constructed multi-user integrated learning load forecasting model; like Figure 2 As shown, the specific steps include: Step 101: Obtaining the power load characteristic data and similar day load data of various users In this embodiment, the power load characteristic data includes meteorological characteristic data affecting the power load, such as perceived temperature, humidity, irradiance, instantaneous precipitation, and wind speed; Considering the periodic characteristics of power load, the input features also include the time label and the day of the week. In addition, since the power loads on adjacent days are similar, The load of the day arrive Daily load data are also used as input features to better predict the changing trend of daily load and thus improve accuracy.

[0031] Step 102: construct a training set and a prediction set based on the acquired power load characteristic data of various users and similar daily load data, train the constructed multiple base learners based on the training set, and then predict the trained base learners based on the prediction set to obtain a prediction result data set; In this embodiment, the base learner uses a support vector machine (SVM) and a deep neural network (DNN); Among them, the support vector machine (SVM) maps the nonlinear problem in the low-dimensional space to the high-dimensional space by introducing the kernel method, so that it can be solved using linear methods. The optimization goal of SVM is to find a function , so that the prediction error is The smallest in the range, That is, it is a nonlinear mapping function, and the RBF kernel function is adopted in this embodiment.

[0032] The optimization objective and constraints of SVM are shown as follows: , In the formula, and b denote the weight vector and bias term respectively, and is a slack variable used to handle -Error outside the insensitive loss, C is the penalty parameter, For the i input features, For thei prediction results; When training SVM, SVM uses -Insensitive loss function, its purpose is to The loss is not calculated within the error range to improve the robustness of the model. The expression of the -insensitive loss function is as follows: , This loss function indicates that when the predicted value and actual value y The error between The loss is not included when Losses are only included when

[0033] Among them, the Deep Neural Network (DNN) consists of an input layer, multiple hidden layers, and an output layer. The input layer receives a feature vector consisting of input features. After that, the hidden layer passes through the weight matrix and the bias vector Process the input and calculate the linear transformation result : , , In the formula, is the activation function; It is l The output of the layer represents the result after the activation function.

[0034] The output layer generates the final prediction result. For regression problems, the output layer does not use an activation function, but directly outputs a continuous real value: , In the formula, is the predicted value, and are the weight matrix and bias vector of the output layer respectively, is the output of the penultimate layer.

[0035] During training, DNN uses the gradient descent algorithm to optimize weights and biases to minimize the loss function. The above units work together to enable DNN to learn the complex mapping relationship between input features and target outputs, thereby achieving accurate predictions.

[0036] Step 103: Prediction result dataset As the secondary training data set and prediction data set, the meta-learner is trained based on the secondary training data set, and the load prediction results of various users are obtained based on the prediction data set and the trained meta-learner; In this embodiment, the meta-learner adopts the extreme gradient booster (eXtreme Gradient Boosting, XGBoost), the basic idea of ​​which is to correct the errors of the previous round of learners by gradually adding new decision trees, thereby improving the prediction performance of the overall model.

[0037] The objective function of XGBoost consists of a loss function and a regular term: , In the formula, Used to represent a set of parameters; is the loss function, is a regular term, and the two can be obtained by the following two formulas: , , Where M is the number of leaf nodes in the tree, It is j The weight of a leaf node, The table is t tree; and is the regularization parameter, , Respectively i The true value and the predicted value.

[0038] The prediction values ​​of each tree are weighted to get the final prediction value: , In the formula, T is the total number of trees.

[0039] Using XGBoost as the final meta-model has the following advantages: First, it is good at processing large-scale data sets, and the use of parallel processing and cache optimization can significantly improve the training speed and efficiency; second, by introducing regularization terms, it can effectively prevent overfitting and improve the generalization ability of the model; finally, XGBoost supports undersampling and multiple loss functions, which enhances its flexibility and adaptability on different tasks and data sets.

[0040] Taking into account that a single model cannot achieve good prediction performance in all prediction scenarios, the above scheme establishes an integrated learning load prediction model for each type of user, namely industrial, commercial, and residential users, and fuses multiple base learners to obtain a strong learning model, thereby improving the prediction accuracy.

[0041] Step 2: Cluster the distributed photovoltaic power stations, define the cluster center of each cluster as the source domain according to the clustering results, and classify the remaining power stations as the target domain; establish and train a short-term photovoltaic power prediction model based on source domain data; transfer the features learned by the short-term photovoltaic power prediction model based on source domain data to the target domain, and obtain a short-term photovoltaic power prediction model suitable for the target domain; Aiming at the problem that the calculation requirements for individual prediction of each site in the distributed photovoltaic power prediction model are large, a short-term photovoltaic power prediction model based on Transformer neural network is constructed. The redundant training cost is avoided through transfer learning based on AP clustering, which effectively improves the prediction accuracy of distributed photovoltaic power prediction in complex prediction scenarios.

[0042] like Figure 3 As shown, the specific steps include: Step 201: cluster the distributed photovoltaic power stations, define the cluster center of each cluster as the source domain according to the clustering result, and classify the remaining power stations as the target domain; In this embodiment, the AP clustering algorithm is used to cluster distributed photovoltaic power stations. AP clustering is an unsupervised clustering algorithm that performs cluster division based on information transmission. It can adapt to various types of data distribution and is not affected by the selection of the initial center. When faced with complex and nonlinearly distributed photovoltaic power station data in reality, it can effectively identify power stations with "consistent output", and the clustering results can be more stable. The core idea of ​​the AP clustering algorithm is that there is no need to pre-specify the number of clusters. All data points are regarded as potential cluster centers. Each data point sends two types of information to other points: attraction information and belonging information. The cluster center is determined by iterating the two types of information. Attraction information Indicates power station Compared with other candidate cluster centers, it is more suitable as a power station The degree of cluster center, information from the power station Transfer to the candidate cluster center station ; Attribution information Indicates power station Select a power station As the degree of suitability of the cluster center, the candidate cluster center station Send to power station , for each data point , whose cluster center is The biggest point.

[0043] The specific steps include: Step 211: After preprocessing the historical power data of the existing distributed photovoltaic power stations, calculate the initial similarity between the power stations. , get the similarity matrix ,In this embodiment, the DTW distance is selected as the initial similarity between two photovoltaic power stations.

[0044] Step 2012: Set the initial reference degree and the number of iterations. The diagonal elements of Characterizes the reference degree of a power station to become a cluster center. The larger the value, the higher the possibility that the data point is a cluster center, and the number of cluster centers will increase. Its value is 0 when calculated by DTW distance, but the degree of a power station as a cluster center cannot be 0 in the AP clustering algorithm. Therefore, before the iteration starts, it is assumed that all points have the same ability to become cluster centers, and the reference degree is generally set to the minimum or median of all values ​​in the similarity matrix.

[0045] Step 2013: Initially, the algorithm needs to initialize the attraction matrix and the attribution matrix to zero matrices, and then calculate the attraction matrix elements and the attribution matrix elements The calculation formula is as follows: , , , In the formula, is the similarity matrix Except the first All elements outside the column; Indicates the first Except the first All elements outside the column; Represents the attraction matrix Excluding The sum of all elements greater than zero outside the row; self-attribution is equal to the sum of the positive attraction obtained from other points, is the attraction matrix k Line k Column elements, is the similarity matrix i Line k Column elements; Step 2014, Update and In order to ensure the convergence speed and stability during the iteration process, it is necessary to introduce a damping coefficient Adjustment, expressed as: , in, Indicates the number of iterations.

[0046] Step 2015: After each update of the attraction information and the attribution information, select a cluster center for each sample point. Repeat the update of the attraction degree matrix and the attribution degree matrix until the matrix changes tend to be stable or the predetermined number of iterations is reached. Calculate the silhouette coefficient based on the clustering results. Change the reference degree, recalculate, confirm the best clustering result, and complete the division of the photovoltaic power station group.

[0047] Step 2016: In order to determine the best clustering result for the division of the photovoltaic power station and ensure the effectiveness of the clustering result, the silhouette coefficient (SC) is selected as the evaluation function of the clustering result. The formula is as follows: , In the formula, After clustering Silhouette coefficient of sample points; For power station The average distance between a sample point and other power station sample points in the same cluster is called cohesion, which indicates the closeness of the point to the cluster; For power station The average distance between the sample point and all samples in other clusters is called separation, which indicates the degree of alienation between the point and other clusters. Each sample point has a corresponding silhouette coefficient value. The average value of the silhouette coefficients of all sample points is calculated as the total silhouette coefficient. The value range is [-1, 1]. The closer the value is to 1, the better the clustering effect. By comparing the silhouette coefficients under different cluster numbers, the best photovoltaic power station clustering result is selected. The station group divided according to this method can basically be equivalent to a small-scale photovoltaic power station group with consistent output.

[0048] Step 202: Establish and train a short-term photovoltaic power prediction model based on source domain data; In this embodiment, the short-term photovoltaic power prediction model uses a Transformer neural network; Transformer completely abandons network structures such as RNN and CNN, and only uses the Attention mechanism (i.e. attention mechanism) for machine translation tasks, and has achieved excellent results.

[0049] The Attention mechanism enables the neural network model to give different attention to the input data through weight setting, thereby capturing the temporal dependency characteristics of the input data. The Attention mechanism can usually be expressed as follows: query (Q) and key-value pairs Mapped to the output, where the query, each key, and each value are vectors, the output is The weight of all values ​​in , where the weight is calculated from the query and each key.

[0050] The Transformer model uses an encoder-decoder structure, where the encoder converts the input sequence Mapping to continuous representation , then the decoder generates an output sequence , output a result at each moment.

[0051] When training the short-term photovoltaic power prediction model, the processed key meteorological factors are used as input, and the neural network is used for the first training, and the historical power data at the corresponding time is used as output; the reliability of the model is verified by the test set of the source domain, and the model parameters are adjusted with the minimum RMSE and MAPE as the goal until the prediction accuracy requirements are met. The completed network model is recorded as .

[0052] Step 203: transferring the features learned by the short-term photovoltaic power prediction model based on the source domain data to the target domain to obtain a short-term photovoltaic power prediction model suitable for the target domain; Considering that the feature extraction of neural networks at high levels is more obvious, while the information at the bottom level is relatively basic, the parameters of all layers are kept basically unchanged, and only these parameters are fine-tuned, while the parameters of the fully connected layer are retrained. The data division results of the source domain and the target domain are the same. We continue to minimize the RMSE and MAPE of the prediction error as the optimization goal, and finally construct a prediction model suitable for the target domain, denoted as , In the migration process, the source domain and the target domain are two core concepts of transfer learning. The source domain is a domain that already has a large amount of labeled data. The model is usually pre-trained in the source domain with sufficient data. The target domain is a domain where the model is expected to perform well, but usually has less labeled data. Generally, the pre-trained model is only fine-tuned to make full use of the source domain data to improve the performance of the model on the target data. The data space of the source domain and the target domain are expressed by the formulas: , , in, and Represent the data space of the source domain and the target domain respectively, and Represent the characteristics of the source domain and target domain data space respectively, and Represent the corresponding labels respectively.

[0053] The task of the source domain and the target domain is to use the appropriate mapping function and Find the optimal source domain parameters of the corresponding mapping function in and the optimal target domain parameters , so that the learning task and As close to the label as possible and , while transfer learning is to find the optimal source domain parameters Based on fine-tuning, the target domain parameters Reach the optimum as quickly as possible. and It is expressed as follows: , , Step 204: Obtain distributed photovoltaic power data based on the short-term photovoltaic power prediction model applicable to the target domain.

[0054] The present invention proposes a distributed photovoltaic power prediction method based on AP clustering transfer learning, constructs a short-term power prediction model based on Transformer neural network, and avoids redundant training costs through AP clustering-based transfer learning, thereby effectively improving the prediction accuracy of distributed photovoltaic power prediction in complex prediction scenarios.

[0055] Step 3: Combine the multi-user load data, distributed photovoltaic data and historical data to calculate the multi-user side resource coordinated regulation potential and obtain the multi-user side resource coordinated regulation potential.

[0056] The specific steps include: Step 301: Based on the load forecast curves of different types of users such as industrial, commercial, and residential users and the power forecast curves of distributed photovoltaics, the regulation potential of different types of user-side resources is evaluated and calculated; The specific steps include: Step 3011: Calculate the baseline output of each user-side resource in different seasons; Statistical forecasting of a specified time range before the day, such as the output value distribution of different types of user-side resources at the same historical moment in a month, and calculating the mean value of different types of user-side resources based on the output value distribution and variance .

[0057] Step 3012: Get the content Maximum value in principle and minimum value As t The maximum and minimum baseline output at the moment.

[0058] Step 3013: Future-based tThe predicted values, maximum baseline output, and minimum baseline output of different types of user-side resources at all times are used to calculate the flexibility adjustment potential of different types of user-side resources; The calculation formula for the potential increase in flexibility of resources on the user side of different categories is as follows: , In the formula, for t The potential for increasing the flexibility of resources on the user side of different categories at all times, for t The maximum baseline output of different types of user-side resources at all times, for t Predicted values ​​of different types of user-side resources at each moment.

[0059] The calculation formula for the downward adjustment potential of different types of user-side resource flexibility is as follows: , In the formula, for t The flexibility of resource reduction at different user-side categories at all times, for t The minimum baseline output of different types of user-side resources at all times.

[0060] Step 302: Consider the sensitivity of different user-side resources to electricity price signals and correct the regulation potential; Time-of-use electricity price is a system that divides daily electricity consumption into peak, valley and normal periods according to the power supply and demand status of the power grid and the seasonal load characteristics, and reasonably determines the price difference of time-of-use electricity price and the electricity price of the corresponding period. Time-of-use electricity price provides economic incentives for users through the difference in electricity prices in different time periods. During peak hours, the electricity price is higher, while during off-peak hours, the electricity price is lower. This encourages users to use electricity during off-peak hours, thereby saving electricity bills. Time-of-use electricity price affects the regulation potential of user-side resources by affecting the user's load value.

[0061] Price-based DR is achieved with the help of demand elasticity, which defines the price elasticity of electricity volume: , In the formula, Indicates time period s For the time period t price elasticity; and Pre-DR period s Power load and time period t The price of electricity; and Post-DR period s Load variation and time period t The amount of price change.

[0062] The load change of user participation in price-based DR is: , The load demand after users participate in demand response becomes , In the formula, is the load demand at time t before demand response.

[0063] Considering the impact of time-of-use electricity prices on the resource regulation potential on the user side, the calculated up and down regulation potential is corrected, and the corrected regulation potential is: , , in, is the load demand change before and after participating in demand response at time t.

[0064] Step 303: Based on the corrected regulation potential, considering the synergy of the user-side resources within the aggregator, the synergy regulation potential of the user-side resources within the aggregator is evaluated and calculated to obtain a potential evaluation calculation result; Specifically, based on the prediction of the adjustable potential of single user-side resources such as flexible loads and distributed photovoltaics, the synergy of multiple user-side resources is considered, and the collaborative adjustment potential of users who aggregate multiple user-side resources is evaluated and calculated: , , in, User-side resources i The revised upside potential, User-side resources i The revised downside potential, For distributed photovoltaic, For load.

[0065] The above solution breaks through the problem that the traditional method of evaluating the resource adjustment capability of multiple user sides relies on device modeling and has a large workload, and provides a new idea for evaluating the adjustable capability of multiple user side resources.

[0066] Embodiment 2 This embodiment provides a multi-user side resource collaborative adjustment potential assessment system, including: The load forecasting module is used to obtain the load forecasting results for each type of user based on the acquired power load characteristic data of each type of user, similar daily load data and the constructed multi-user integrated learning load forecasting model; The photovoltaic power prediction module is used to cluster distributed photovoltaic power stations, determine the source domain and target domain based on the clustering results, establish and train a short-term photovoltaic power prediction model based on the source domain data; transfer the features learned by the short-term photovoltaic power prediction model based on the source domain data to the target domain, obtain a short-term photovoltaic power prediction model suitable for the target domain, and further predict the distributed photovoltaic power data; The regulation potential assessment module is used to establish the regulation potential of a single user-side resource by combining multiple user load data and distributed photovoltaic data, consider the sensitivity of different user-side resources to electricity price signals, correct the regulation potential, consider the synergy of user-side resources within the aggregator, and calculate the coordinated regulation potential of user-side resources within the aggregator to obtain the coordinated regulation potential.

[0067] It should be noted that the specific implementation method of a system for evaluating the potential for collaborative regulation of resources on the multi-user side in an embodiment of the present invention is similar to the specific implementation method of a method for evaluating the potential for collaborative regulation of resources on the multi-user side in embodiment one of the present invention. Please refer to the description of the method part for details. In order to reduce redundancy, it will not be repeated here.

[0068] Embodiment 3 This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the method for evaluating the potential for collaborative regulation of multiple user-side resources as described above are implemented.

[0069] Embodiment 4 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for evaluating the potential for collaborative regulation of multiple user-side resources as described above are implemented.

[0070] Embodiment 5 This embodiment provides a program product, which is a computer program product, including a computer program. When the computer program is executed by a processor, the steps in the method for evaluating the potential for collaborative regulation of multiple user-side resources as described above are implemented.

[0071] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating the potential of collaborative regulation of multiple user-side resources, characterized in that: The steps include: Based on the acquired power load characteristic data of various users, similar daily load data and the constructed multi-user integrated learning load forecasting model, the load forecast results for each type of user are obtained; Distributed photovoltaic power stations are clustered, and the source domain and target domain are determined based on the clustering results. A short-term photovoltaic power prediction model based on the source domain data is established and trained. The features learned from the short-term photovoltaic power prediction model based on the source domain data are transferred to the target domain to obtain a short-term photovoltaic power prediction model suitable for the target domain, and the distributed photovoltaic power data is further predicted. The regulation potential of a single user-side resource is established by combining multiple user load data and distributed photovoltaic data. The sensitivity of different user-side resources to electricity price signals is considered to correct the regulation potential. The synergy of user-side resources within the aggregator is considered, and the coordinated regulation potential of user-side resources within the aggregator is calculated to obtain the coordinated regulation potential.

2. A method for evaluating the potential of collaborative regulation of multiple user-side resources according to claim 1, characterized in that: The load forecasting results for each type of user are obtained based on the acquired power load characteristic data of each type of user, similar daily load data and the constructed multi-user integrated learning load forecasting model, including: Based on the acquired power load characteristic data of various users and similar daily load data, a training set and a prediction set are constructed, multiple base learners are trained based on the training set, and then the trained base learners are predicted based on the prediction set to obtain a prediction result data set; The prediction result dataset is used as the secondary training dataset and the prediction dataset. The meta-learner is trained based on the secondary training dataset. The load prediction results of various users are obtained based on the prediction dataset and the trained meta-learner.

3. A method for evaluating the potential of collaborative regulation of multiple user-side resources according to claim 1, characterized in that: The distributed photovoltaic power stations are clustered, and the source domain and the target domain are determined based on the clustering results, including clustering the distributed photovoltaic power stations using an AP clustering algorithm, defining the cluster center of each cluster as the source domain according to the clustering results, and classifying the remaining power stations as the target domain.

4. A method for evaluating the potential of collaborative regulation of multiple user-side resources according to claim 1, characterized in that: The short-term photovoltaic power prediction model uses Transformer neural network.

5. The method for evaluating the potential of collaborative regulation of multiple user-side resources according to claim 1, characterized in that: When the features learned by the short-term photovoltaic power prediction model based on the source domain data are transferred to the target domain, the model parameters are adjusted with the minimum RMSE and MAPE as the goal.

6. A method for evaluating the potential of collaborative regulation of multiple user-side resources according to claim 1, characterized in that: Combine multiple user load data and distributed photovoltaic data to establish the regulation potential of single user side resources, including: Calculate the baseline output of each user-side resource in different seasons; Satisfaction Maximum value in principle and minimum value As t Maximum baseline output and minimum baseline output at the moment; Based on the future t The predicted values ​​of different types of user-side resources and the maximum baseline output at each moment are used to calculate the flexibility increase potential of different types of user-side resources. t The predicted values ​​and minimum baseline outputs of different categories of user-side resources at each moment are used to obtain the flexibility reduction potential of different categories of user-side resources.

7. A method for evaluating the potential of collaborative regulation of multiple user-side resources according to claim 1, characterized in that: The adjustment potential after correction is: ， ， in, and is the revised upward and downward potential, and Before correction t The potential for upward and downward adjustment of resource flexibility for different categories of users at all times. is the load demand at time t before demand response, is the load demand change before and after participating in demand response at time t.

8. A method for evaluating the potential of collaborative regulation of multiple user-side resources according to claim 1, characterized in that: The potential for synergistic regulation is: ， ， in, User-side resources i The revised upside potential, User-side resources i The revised downside potential, For distributed photovoltaic, For load.

9. A multi-user side resource collaborative regulation potential assessment system, characterized in that: include: The load forecasting module is used to obtain the load forecasting results for each type of user based on the acquired power load characteristic data of each type of user, similar daily load data and the constructed multi-user integrated learning load forecasting model; The photovoltaic power prediction module is used to cluster distributed photovoltaic power stations, determine the source domain and target domain based on the clustering results, establish and train a short-term photovoltaic power prediction model based on the source domain data; transfer the features learned by the short-term photovoltaic power prediction model based on the source domain data to the target domain, obtain a short-term photovoltaic power prediction model suitable for the target domain, and further predict the distributed photovoltaic power data; The regulation potential assessment module is used to establish the regulation potential of a single user-side resource by combining multiple user load data and distributed photovoltaic data, consider the sensitivity of different user-side resources to electricity price signals, correct the regulation potential, consider the synergy of user-side resources within the aggregator, and calculate the coordinated regulation potential of user-side resources within the aggregator to obtain the coordinated regulation potential.

10. A program product, the program product being a computer program product, comprising a computer program, characterized in that: When the computer program is executed by a processor, the steps in a method for evaluating the potential for collaborative regulation of multiple user-side resources as described in any one of claims 1-8 are implemented.

Images