Multi-level layered aggregation regulation and control method and system for adjustable resources

Abstract

The invention belongs to the technical field of load regulation and control, and provides a multistage hierarchical aggregation regulation and control method and system for adjustable resources, and the method comprises the steps: obtaining historical power consumption data, and carrying out the clustering of buildings according to the power consumption; carrying out controllability analysis on the electric equipment of each building group, and determining the load response level of the electric equipment; building energy consumption is divided into different energy consumption modes; inputting the historical power consumption data into the prediction model to obtain power consumption load prediction results of different response levels in different energy consumption modes; and regulating and controlling the electrical load of the electrical equipment according to the prediction result. According to the invention, partition aggregation regulation and control are carried out according to the priorities of the energy consumption mode, the controllability and the response speed.

Description

Technical Field

[0001] The present disclosure relates to the technical field of load regulation, and in particular to a multi-level hierarchical aggregation regulation method and system for adjustable resources. Background Art

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

[0003] As power facilities are gradually improved, the situation of continuous power shortage has been alleviated. However, the power grid still faces great pressure to regulate the peak load during peak load. Some areas still have power shortages. The imbalance between power demand and resource distribution and seasonal supply and demand imbalances have led to a severe power situation. At the same time, with the reform of the power market and the development of smart grid technology, power users can actively change their own power consumption methods through demand response mechanisms to reduce and transfer loads, thereby achieving a balance between supply and demand of the power system through control of the load side.

[0004] The existing aggregate control scheme only considers the adjustable resources, but does not take into account the equipment response speed and the different energy consumption patterns due to different seasons. It is impossible to achieve refined aggregate control and the system stability is insufficient. Summary of the invention

[0005] In response to the above-mentioned defects, the present disclosure proposes a multi-level hierarchical aggregation control method and system for adjustable resources, which performs partition aggregation control according to the three levels of priority of energy consumption pattern, controllability and response speed. It can implement aggregation control on the load in a more refined manner, improve the adaptability and response speed of the system, and realize comprehensive, coordinated and efficient management of multi-energy systems.

[0006] In order to achieve the above objectives, the present disclosure adopts the following technical solutions:

[0007] The first aspect of the present disclosure provides a multi-level hierarchical aggregation control method for adjustable resources, comprising the following steps:

[0008] Obtain historical electricity consumption data and cluster buildings based on electricity consumption;

[0009] Conduct controllability analysis on the electrical equipment in each building complex and determine the load response level of the electrical equipment;

[0010] Classify building energy consumption into different energy consumption modes;

[0011] Input historical electricity consumption data into the prediction model to obtain electricity load prediction results of different response levels under different energy consumption modes;

[0012] According to the prediction results, the power load of the electrical equipment is regulated.

[0013] As a further implementation method, controllability analysis is the response and controllability of the equipment to the power system.

[0014] As a further implementation, the load response level includes instantaneous response, short-time response and medium- and long-time response.

[0015] As a further implementation method, the energy consumption modes include cooling, heating and electricity.

[0016] As a further implementation, the three energy consumption modes have different peak power consumption periods and power consumption characteristics.

[0017] As a further implementation method, the power load of the power equipment is regulated according to the prediction results, specifically:

[0018] Analyze the trend and change pattern of the predicted load. When the actual load is higher or lower than the predicted load and the range exceeds the set threshold, the controllable equipment will be adjusted to reduce or increase the load.

[0019] As a further implementation method, the historical electricity consumption data includes the electricity consumption, energy consumption time distribution and load curve of each building.

[0020] A second aspect of the present disclosure provides a multi-level hierarchical aggregation control system for adjustable resources, including:

[0021] The clustering module is configured to: obtain historical electricity consumption data and cluster the buildings according to the electricity consumption;

[0022] The load response level determination module is configured to: perform controllability analysis on the electrical equipment of each building complex and determine the load response level of the electrical equipment;

[0023] The energy consumption mode classification module is configured to: classify the building energy consumption into different energy consumption modes;

[0024] The model prediction module is configured to: input historical electricity consumption data into the prediction model to obtain electricity load prediction results of different response levels under different energy consumption modes;

[0025] The control module is configured to: control the power load of the electrical equipment according to the prediction result.

[0026] The third aspect of the present disclosure provides a medium on which a program is stored. When the program is executed by a processor, the steps of the multi-level hierarchical aggregation control method for adjustable resources described in the first aspect of the present disclosure are implemented.

[0027] The fourth aspect of the present disclosure provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of a multi-level hierarchical aggregation control method for adjustable resources described in the first aspect of the present disclosure are implemented.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The multi-level hierarchical aggregation control method and system for adjustable resources proposed in the present disclosure performs partition aggregation control according to the three-level priority of energy consumption mode, controllability and response speed, and can implement aggregation control on loads in a more refined manner. This makes the energy system more flexible, able to adapt to changes in different times, seasons and business needs, and improves the adaptability and response speed of the system.

[0030] The present invention discloses a multi-level hierarchical aggregation control method and system for adjustable resources, proposes a TAM-CNN-LSTM multi-task learning method for extracting features according to seasons, and adds SARIMA and VAE models according to the different characteristics of instantaneous short-term and medium- and long-term responses, which can improve the model's representation and generalization capabilities for load data, thereby improving the accuracy and stability of load forecasting.

[0031] 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

[0032] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.

[0033] Figure 1 It is a flowchart of a multi-level hierarchical aggregation control method for adjustable resources disclosed in the present invention. DETAILED DESCRIPTION

[0034] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0035] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. 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 disclosure belongs.

[0036] In the absence of conflict, the embodiments in the present disclosure and the features in the embodiments may be combined with each other.

[0037] Embodiment 1

[0038] like Figure 1 As shown, the first embodiment of the present disclosure provides a multi-level hierarchical aggregation control method for adjustable resources, including the following steps:

[0039] S1 obtains historical electricity consumption data and clusters buildings according to their electricity consumption;

[0040] S2 conducts controllability analysis on the electrical equipment in each building complex and determines the load response level of the electrical equipment;

[0041] S3 divides building energy consumption into different energy consumption modes;

[0042] S4 inputs historical electricity consumption data into the prediction model to obtain electricity load prediction results of different response levels under different energy consumption modes;

[0043] S5 adjusts the power load of the electrical equipment according to the prediction result.

[0044] Partition aggregation control is carried out according to three levels of priority: energy consumption pattern, controllability and response speed.

[0045] Energy consumption patterns are considered the first priority factor because they can directly reflect the importance of actual demand: energy consumption patterns determine whether the electricity demand of a region or system is regular and predictable. Areas with long-term stable and high-load demand (such as hospitals and communication base stations) are naturally given higher priority because they are more dependent on energy supply and power outages will have a serious impact. Therefore, regulation must first ensure the power supply of these areas.

[0046] Ability to optimize resource allocation: By prioritizing different types of energy consumption patterns, energy regulation resources can be allocated more reasonably. When regulation capacity is limited, high-priority and power-demand scenarios are prioritized to ensure maximum energy efficiency.

[0047] Controllability is considered as the second priority factor because the flexibility of controllability affects the dispatching strategy: controllability determines whether a load can flexibly participate in energy dispatching. In energy dispatching, the easier the load is to control, the greater its contribution to grid stability. Therefore, loads with high controllability can better cope with dispatching requirements such as power supply fluctuations and peak shaving. Although the energy consumption pattern determines the rigidity of demand, if the load is highly controllable, it can be effectively adjusted while meeting stable demand and reducing grid pressure.

[0048] It can reduce energy waste: loads with high controllability can adjust power according to the real-time situation of the power grid, thus avoiding unnecessary energy waste. Under the dual carbon goals, the dispatching system must give priority to those controllable loads to maximize energy utilization efficiency.

[0049] Response speed is the third priority factor because it can affect the emergency response capability of the system: response speed measures how quickly the load can be adjusted after receiving the dispatch instruction. Although response speed is very important for emergency dispatch or peak load shifting, it is based on the first two factors. Because if an area or load has a fast response capability, but the energy consumption pattern is not important or cannot be adjusted, its dispatch value will also decrease. Therefore, although response speed is important, it only becomes an indicator for optimizing regulation after the energy consumption pattern and controllable load are reasonably selected.

[0050] Improve system flexibility: Loads with fast response speed can respond quickly in emergencies and help the power grid maintain stability when loads fluctuate drastically. However, such loads are usually designed to cope with extreme situations or short-term fluctuations and are not a top priority in daily operations.

[0051] First, S1 obtains historical electricity consumption data, which includes the electricity consumption, energy consumption time distribution and load curve of each building. The k-means clustering algorithm is used to cluster buildings according to electricity consumption. The basic idea of ​​the k-means clustering algorithm is to find a partitioning scheme of k clusters through iteration. The first premise of k-means is to determine the value of k. This paper finally determines the value of k by the silhouette coefficient method. The loss function corresponding to the clustering result is minimized. The loss function formula is shown in formula (1):

[0052]

[0053] where x i represents the i-th sample, c i is x i The cluster to which it belongs, Represents the center point corresponding to the cluster, and M is the total number of samples. The silhouette coefficient is an indicator used to evaluate the quality of clustering. It can be used to measure whether the cluster assigned to each data point is appropriate and the degree of separation between clusters. The formula of the silhouette coefficient is shown in formula (2):

[0054]

[0055] Where a(i) represents the cohesion of the sample point, and the calculation method is shown in formula (3):

[0056]

[0057] Here, j represents other sample points in the same class as sample i, and distance represents the distance between i and j. Therefore, the smaller a(i), the closer the class is. The calculation method of b(i) is similar to a(i).

[0058] By considering factors such as geographical location and building attributes for each clustered building complex, similarity is measured for individual buildings in adjacent areas but not in the same clustered building complex, and individual buildings far away from the building complex can be divided into adjacent large building complexes.

[0059] Then, S2 conducts a controllability analysis on the electrical equipment of each building complex (the response and control capability of users or equipment to the power system) to determine the load response level of the electrical equipment. Users or equipment with higher controllability can participate in control activities more flexibly, improving the stability and response capability of the system.

[0060] The load response levels of equipment include instantaneous response, short-term response, and medium- and long-term response. The specific classification standards are as follows:

[0061] (1) Transient response

[0062] Time Required: Response time is within a few seconds to a minute.

[0063] Applicable scenarios: Suitable for loads that are sensitive to power changes and have the ability to quickly adjust the equipment, such as lighting, instantaneous adjustment of certain production equipment, etc.

[0064] (2) Short-term response

[0065] Time required: Response time is from a few minutes to half an hour.

[0066] Applicable scenarios: Suitable for loads that require faster adjustment but do not have extremely high requirements for response speed, such as the start and stop of some production lines, air conditioning temperature adjustment, etc.

[0067] (3) Medium and long-term response

[0068] Time Required: Response time ranges from a few hours to a day.

[0069] Applicable scenarios: Suitable for loads with relatively slow power changes and relatively long production processes, such as the switching of some industrial production lines and the overall adjustment of building energy consumption.

[0070] S3 divides building energy consumption into different energy consumption modes. Energy consumption modes include cold, hot and electric. The three energy consumption modes have different peak hours and electricity consumption characteristics. There are multiple benefits to dividing building energy consumption into cold, hot and electric. First, this detailed division helps to achieve refined management, make energy consumption data clearer, and then gain a deeper understanding of the energy consumption of each part of the building system. Secondly, by monitoring and analyzing cold, hot and electric energy consumption separately, the energy consumption bottleneck in the system can be more accurately identified, and energy-saving measures can be proposed and implemented in a targeted manner to minimize energy waste. In addition, such a division helps to carry out comprehensive system optimization and improve the energy efficiency of the entire building system. Through scientific decision-making, building managers can better understand the usage of various energy fields and make more scientific and reasonable decisions and plans.

[0071] S4 inputs historical electricity consumption data into the prediction model to obtain electricity load prediction results with different response levels under different energy consumption modes.

[0072] Since cold, hot and electric loads have different peak hours and electricity consumption characteristics, they are zoned and aggregated according to the different energy consumption habits of the cold, hot and electric loads and the response levels of each load (instantaneous response, short-time response, medium and long-time response).

[0073] First, the historical electricity consumption data of each partition is analyzed for the characteristics of cooling, heating and electricity loads. The historical load data is collected and analyzed, including load curves, load peaks, load fluctuations and other information in different time periods. Load curves of different load partitions are drawn to show the daily, weekly and monthly load changes. The complex load curves are decomposed and analyzed by Fourier analysis, and the periodic changes in the load curves are accurately identified and quantitatively analyzed to gain a deeper understanding of the daily changes in cooling, heating and electricity loads.

[0074] In a year, the electricity and cooling loads show a trend of being high in the middle and low on both sides, with strong local randomness and volatility, and the peaks are mainly concentrated in summer. Compared with electricity and cooling loads, the heating load has low demand and large fluctuations throughout the year, and the peak is mainly concentrated in winter. Obviously, climate change has a direct impact on user load demand, and the changes of each load within a year have obvious time correlation. Multi-task learning is used to predict them simultaneously. The features of the electricity load are extracted normally. For the cooling and heating loads, their features are extracted according to the importance of the cooling and heating use of the building complex in different seasons. For example, in residential areas in summer, only the electricity and cooling loads are further extracted. In terms of prediction methods, a multi-step feature extraction method based on seasons is proposed, and the prediction model uses the TAM-CNN-LSTM model.

[0075] TAM allows the model to focus on the most relevant time points or regions in the time series, thereby improving the model's perception of important features. TAM enables the model to dynamically focus on information at different time points by learning the weights of each time step in the time series, thereby more effectively capturing the changing trends and correlations in the time series. CNN can effectively capture local features in time series data, such as sudden changes or periodic patterns. Through convolution operations, CNN can extract features at different scales and combine these features to better represent the structural information of the time series. LSTM networks can capture long-term dependencies in time series data and have better processing capabilities for memories in sequences.

[0076] The TAM-CNN-LSTM model first receives time series data, which is usually a three-dimensional tensor with the number of samples, time steps, and number of features. Then, according to different seasons, further feature extraction is performed on data with high energy consumption (such as electricity and cooling loads in summer). The data is first processed by the temporal attention mechanism (TAM) layer. The role of the TAM layer is to establish time dependencies in the time series data to help the model better focus on important time steps. The TAM layer is usually composed of multiple attention mechanisms, which are used to focus on and adjust different aspects of the time series data. Then, feature extraction is performed through a convolutional layer (CNN), and the size of the convolution kernel is set to (k, F), where k is the size of the convolution kernel and F is the number of features. The convolution operation can capture the local relationship of features in the time dimension. Finally, the cold, heat and electricity data are processed through the LSTM layer. The LSTM unit can memorize and update the information in the sequence, thereby realizing the long-term dependency modeling of the sequence data. Finally, the features extracted by the LSTM layer are integrated through the fully connected layer, and the final prediction results are output. In the fully connected layer, a Dropout layer can be added to prevent overfitting. In summary, the TAM-CNN-LSTM model can dynamically and selectively focus on important information at different time points through the temporal attention mechanism, extract local features through CNN and improve the representation ability of time series, and capture long-term dependencies and process memory through LSTM, thereby effectively extracting important features from time series data. The model can capture the time correlation, periodicity, trend and other features in load data, thereby achieving accurate load forecasting. The TAM-CNN-LSTM model can effectively process time series data, extract important features, and achieve accurate forecasting.

[0077] 1. Power load forecasting at instantaneous and short-term response levels.

[0078] Since transient response requires prediction of instantaneous load and more accurate prediction is required, the Seasonal Autoregressive Integrated Moving Average (SARIMA) model is added to the aggregation control method of the transient response level.

[0079] First, the SARIMA model is used to predict the electric load data. The prediction results of the SARIMA model are input into the TAM-CNN-LSTM model as additional features. The input data of the TAM-CNN-LSTM model includes: the original electric load data, that is, the historical load data sequence. The prediction results of the SARIMA model, that is, the predicted value of the load data at a future time point. These two parts of data will be used as input features of the TAM-CNN-LSTM model for the model to perform load forecasting.

[0080] The output of the TAM-CNN-LSTM model is the predicted value of the electric load data at a future time point. By learning the historical load data and its association with the prediction results of the SARIMA model, the TAM-CNN-LSTM model can generate more accurate load forecast results. This can fully utilize the prediction ability of the time series model and provide richer and more accurate input features.

[0081] The prediction results of the SARIMA model are combined with the features extracted by the TAM-CNN-LSTM model. A feature concatenation method, such as the weighted average method, is used.

[0082] The SARIMA model and the TAM-CNN-LSTM model are serially integrated into an overall model. First, the SARIMA model is used to predict the data to obtain preliminary prediction results. Then, these prediction results are used as the input of the TAM-CNN-LSTM model to predict again and obtain the final prediction results of the overall model. This serial integration method can fully utilize the advantages of the two models and improve the accuracy and stability of the prediction.

[0083] When training the integrated model, the model parameters and hyperparameters are optimized through cross-validation and other techniques to improve the model's generalization ability and predictive performance.

[0084] After the model training is completed, the integrated model is evaluated and tuned. The model is evaluated on the validation set, and the model parameters and feature fusion strategy are adjusted according to the evaluation results to further improve the prediction accuracy and stability of the model.

[0085] Through the above steps, the SARIMA model and the TAM-CNN-LSTM model are organically combined to make full use of their respective advantages to improve the accuracy and reliability of load forecasting. The SARIMA model is a classic time series model that is suitable for modeling and forecasting data with seasonal, trend and cyclical changes. Instantaneous forecasting usually involves forecasting data within a short time range in the future, while the SARIMA model can better capture the seasonal and cyclical changes of the data and has better forecasting capabilities for data with obvious seasonal characteristics. In addition, the modeling and forecasting process of the SARIMA model is relatively simple, and the calculation speed is fast, which is suitable for instantaneous forecasting tasks that require fast response and real-time forecasting.

[0086] 2. Power load forecasting at medium and long-term response levels.

[0087] Variational autoencoder (VAE) is added to the medium and long-term load forecasting model. First, historical load data is prepared as input, and the corresponding labels (i.e., load data at future moments) are used as the target of supervised learning. At the same time, the historical load data is normalized to ensure the stability and convergence of model training. Then, a variational autoencoder (VAE) model is constructed, including encoder and decoder parts. The encoder is responsible for mapping the input historical load data to the distribution parameters in the latent space, and a multi-layer neural network structure is used for modeling. The decoder is responsible for decoding the representation in the latent space into the reconstructed input data, and a multi-layer neural network structure is also used for modeling. Finally, the VAE model is used at the input end of the TAM-CNN-LSTM model to encode the historical load data to obtain the latent representation, which is input into the TAM-CNN-LSTM model together with the original historical load data. The overall model is trained end-to-end, and the model parameters are adjusted by minimizing the loss function between the predicted value and the true label. At the same time, the reconstruction error and KL divergence in the VAE model are minimized to ensure that the VAE model learns a suitable latent representation. The integrated model is evaluated, and the performance of the model is evaluated by methods such as validation set or cross-validation. According to the evaluation results, the model is tuned, and hyperparameters and network structures are adjusted to improve the prediction accuracy and generalization ability of the model. By integrating VAE and TAM-CNN-LSTM models, the potential representation of data learned by the VAE model can be fully utilized to improve the feature extraction and modeling capabilities of the TAM-CNN-LSTM model for load data, thereby improving the accuracy and stability of load forecasting.

[0088] S5 adjusts the power load of the electrical equipment according to the prediction result.

[0089] Specifically:

[0090] For the instantaneous and short-term response level power load forecast, the trend and change rules of the forecast load data are analyzed. When the actual load is higher or lower than the forecast load by more than 10%, the controllable equipment will be adjusted to reduce or increase the load. According to the formulated control strategy, the corresponding control measures are implemented to adjust the actual load level to make it close to the forecast load.

[0091] The cost of purchasing energy during peak hours is high. Instantaneous and short-term control measures can reduce the cost of purchasing energy, improve energy utilization efficiency and system operation efficiency, and avoid energy waste and unnecessary cost expenditures. Timely response to load fluctuations and emergencies can reduce the operating risk and instability of the system and ensure the safety and stable operation of the system.

[0092] For the medium and long-term response level electricity load forecast, since the medium and long-term forecast takes a longer time than the instantaneous short-term forecast, before the load change is predicted, control measures should be executed in advance to adjust the load of controllable equipment to adapt to the expected changes in future loads. Through the operation of the intelligent control system or operation and maintenance personnel, control measures should be implemented in a timely manner and the changes in actual load should be monitored.

[0093] By predicting and planning the load trends and changes over a long period of time in the future, we can optimize the investment and operation strategies of equipment, avoid overinvestment and waste of resources, and thus reduce the overall cost of the system. We can rationally allocate resources and schedule the operation of equipment to maximize the use of existing resources and improve the overall efficiency and production capacity of the system. We can identify and respond to potential risks and problems in advance, reduce the operational risks and uncertainties of the system, and improve the risk resistance and reliability of the system.

[0094] Embodiment 2

[0095] Embodiment 2 of the present disclosure provides a multi-level hierarchical aggregation control system for adjustable resources, including:

[0096] The clustering module is configured to: obtain historical electricity consumption data and cluster the buildings according to the electricity consumption;

[0097] The load response level determination module is configured to: perform controllability analysis on the electrical equipment of each building complex and determine the load response level of the electrical equipment;

[0098] The energy consumption mode classification module is configured to: classify the building energy consumption into different energy consumption modes;

[0099] The model prediction module is configured to: input historical electricity consumption data into the prediction model to obtain electricity load prediction results of different response levels under different energy consumption modes;

[0100] The control module is configured to: control the power load of the electrical equipment according to the prediction result.

[0101] The more detailed steps are the same as those in the first embodiment and will not be repeated here.

[0102] Embodiment 3

[0103] Embodiment 3 of the present disclosure provides a medium on which a program is stored. When the program is executed by a processor, the steps of a multi-level hierarchical aggregation control method for adjustable resources as described in Embodiment 1 of the present disclosure are implemented.

[0104] The more detailed steps are the same as those in the first embodiment and will not be repeated here.

[0105] Embodiment 4

[0106] Embodiment 4 of the present disclosure provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in a multi-level hierarchical aggregation control method for adjustable resources as described in Embodiment 1 of the present disclosure are implemented.

[0107] The more detailed steps are the same as those in the first embodiment and will not be repeated here.

[0108] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A multi-level hierarchical aggregation control method for adjustable resources, characterized in that: The following steps are involved: Obtain historical electricity consumption data and cluster buildings based on electricity consumption; Conduct controllability analysis on the electrical equipment in each building complex and determine the load response level of the electrical equipment; Classify building energy consumption into different energy consumption modes; Input historical electricity consumption data into the prediction model to obtain electricity load prediction results of different response levels under different energy consumption modes; According to the prediction results, the power load of the electrical equipment is regulated.

2. A multi-level hierarchical aggregation control method for adjustable resources as claimed in claim 1, characterized in that: Controllability analysis refers to the response and controllability of equipment to the power system.

3. A multi-level hierarchical aggregation control method for adjustable resources as described in claim 2, characterized in that: The load response levels include instantaneous response, short-time response and medium- and long-time response.

4. A multi-level hierarchical aggregation control method for adjustable resources as claimed in claim 1, characterized in that: Energy consumption modes include cooling, heating and electricity.

5. A multi-level hierarchical aggregation control method for adjustable resources as claimed in claim 4, characterized in that: The three energy consumption modes have different peak power consumption periods and power consumption characteristics.

6. A multi-level hierarchical aggregation control method for adjustable resources as claimed in claim 1, characterized in that: According to the prediction results, the power load of the electrical equipment is regulated, specifically: Analyze the trend and change pattern of the predicted load. When the actual load is higher or lower than the predicted load and the range exceeds the set threshold, the controllable equipment will be adjusted to reduce or increase the load.

7. A multi-level hierarchical aggregation control method for adjustable resources as claimed in claim 1, characterized in that: The historical electricity consumption data includes the electricity consumption, energy consumption time distribution and load curve of each building.

8. A multi-level hierarchical aggregation control system for adjustable resources, characterized in that: include: The clustering module is configured to: obtain historical electricity consumption data and cluster the buildings according to the electricity consumption; The load response level determination module is configured to: perform controllability analysis on the electrical equipment of each building complex and determine the load response level of the electrical equipment; The energy consumption mode classification module is configured to: classify the building energy consumption into different energy consumption modes; The model prediction module is configured to: input historical electricity consumption data into the prediction model to obtain electricity load prediction results of different response levels under different energy consumption modes; The control module is configured to: control the power load of the electrical equipment according to the prediction result.

9. A medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of a multi-level hierarchical aggregation control method for adjustable resources as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the multi-level hierarchical aggregation control method for adjustable resources as described in any one of claims 1-7 are implemented.

Images