Aggregation regulation and control system for demand-side adjustable resources

Abstract

The invention discloses an aggregation regulation and control system for demand side adjustable resources. The aggregation regulation and control system comprises an adjustable resource pool, a resource regulation and control management module and a resource dynamic interaction decision module, the adjustable resource pool is used for acquiring load node data of adjustable resource load nodes, clustering to obtain different types of adjustable resource loads, and constructing load models corresponding to the various types of adjustable resource loads; the resource regulation and control management module is used for constructing a load prediction model corresponding to each type of adjustable resource load, and then predicting the predicted load output of each type of adjustable resource load according to the load prediction model; and the resource dynamic interaction decision module is used for analyzing the adjustable power of each type of adjustable resource load in different scheduling scenes according to the predicted load output of each type of adjustable resource load, and scheduling the demand response of each type of adjustable resource load. According to the invention, aggregation regulation and control of demand-side adjustable resources can be realized.

Description

Technical Field

[0001] The present invention relates to the field of computer application technology, and in particular to an aggregation control system for demand-side adjustable resources. Background Art

[0002] In the current power market environment, with the large-scale access of renewable energy and the diversification of user electricity demand, the stable operation of the integrated energy system faces huge challenges. As an important means of regulation in the integrated energy system, the potential of demand-side resources has not yet been fully developed and utilized. The multi-source data resources collected by the existing demand-side resource platform have not been fully utilized, and there is a lack of complete technical support for the hierarchical aggregation and regulation of demand-side resources, making it difficult to achieve accurate aggregation and regulation of demand-side adjustable resources. Summary of the invention

[0003] The present invention provides an aggregation and control system for demand-side adjustable resources to solve the technical problems that the multi-source data resources collected by the existing demand-side resource platform are not fully utilized, there is a lack of complete technical support in the hierarchical aggregation and control of demand-side resources, and it is difficult to achieve accurate aggregation and control of demand-side adjustable resources.

[0004] In order to solve the above technical problems, an embodiment of the present invention provides an aggregation control system of demand-side adjustable resources, including: an adjustable resource pool, a resource control management module and a resource dynamic interactive decision module;

[0005] The adjustable resource pool is used to obtain load node data of adjustable resource load nodes, cluster the load node data, obtain different types of adjustable resource loads by clustering, and construct load models corresponding to each type of adjustable resource load; wherein the types of adjustable resource loads include: industrial load, commercial load, residential load and emerging load;

[0006] The resource regulation and management module is used to take the historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, train the load model corresponding to each type of adjustable resource load, obtain the load prediction model corresponding to each type of adjustable resource load, and then predict the predicted load output of each type of adjustable resource load according to the load prediction model;

[0007] The resource dynamic interactive decision module is used to analyze the adjustable power of each type of adjustable resource load in different scheduling scenarios according to the predicted load output of each type of adjustable resource load, and then schedule the demand response of each type of adjustable resource load according to the adjustable power.

[0008] As a preferred solution, the adjustable resource pool includes: a data acquisition unit, a load multi-level clustering unit and an adjustable resource model library;

[0009] The data acquisition unit is used to connect with the measuring device on the bus of the integrated energy system, and obtain the load node data of the adjustable resource load node through the measuring device; wherein the load node data includes: voltage, current, frequency and power;

[0010] The load multi-level clustering unit is used to extract power consumption indicators representing user power consumption characteristics according to the load node data, and then cluster according to the power consumption indicators to obtain different types of adjustable resource loads; wherein the power consumption indicators include: peak load, load factor and load change rate;

[0011] The adjustable resource model library is used to construct load models corresponding to various types of adjustable resource loads.

[0012] As a preferred solution, the extracting of the power consumption index representing the power consumption characteristics of the user according to the load node data includes:

[0013] The load node data is input into a preset electricity consumption index extraction model so that the electricity consumption index extraction model performs data preprocessing on the load node data, and extracts high-level features in the load node data after data preprocessing to obtain the corresponding implicit state vector, and then analyzes the implicit state vector to obtain and output the electricity consumption index used to represent the user's electricity consumption characteristics; wherein the data preprocessing includes: data cleaning, data normalization and data missing value filling.

[0014] As a preferred solution, clustering is performed according to the power consumption indicators to obtain different types of adjustable resource loads, including:

[0015] According to the power consumption index and the preset silhouette coefficient method, the corresponding average silhouette coefficient is calculated, and the number of clusters when the average silhouette coefficient is the largest is found;

[0016] According to the power consumption characteristics, the number of clusters and a preset K-means algorithm, the adjustable resource load is divided into clusters of the same number as the number of clusters;

[0017] The central feature vector of each cluster is calculated, and the type of the adjustable resource load of each cluster is determined according to the central feature vector.

[0018] As a preferred solution, the load model includes: a ZIP static model;

[0019] The construction of load models corresponding to various types of adjustable resource loads includes:

[0020] For each type of adjustable resource load, constructing a constant impedance model when the equivalent impedance of the adjustable resource load is constant, a constant current model when the equivalent current of the adjustable resource load is constant, and a constant power model when the power of the adjustable resource load is constant according to the load node data;

[0021] According to the constant impedance model, the constant current model and the constant power model, a ZIP static model corresponding to the adjustable resource load is constructed.

[0022] As a preferred solution, the ZIP static model is:

[0023] P=P0·[k Z (V t / V0) 2 +k I (V t / V0)+k P ](1+L P Δf / f0);

[0024] Q=Q0·[k′ Z (V t / V0) 2 +k′ I (V t / V0)+k′ P ](1+L Q Δf / f0);

[0025] k Z +k I +k P =1;

[0026] k′ Z +k′ I +k′ P =1;

[0027] Among them, P and Q are respectively the active power and reactive power actually absorbed by the total load; P0 and Q0 are respectively the active power and reactive power absorbed by all loads in rated operation state; k Z , k I and k P are the proportions of the active constant impedance part, the constant current part and the constant power part in the ZIP static model respectively; k′ Z , k′ I and k′ P are the proportions of the reactive constant impedance part, the constant current part and the constant power part in the ZIP static model respectively; V t and V0 are the voltage and initial voltage at time t respectively; f and f0 are the actual frequency and rated frequency of the power grid respectively; (1+LP Δf / f0) and (1+L Q Δf / f0) are the active frequency characteristics and reactive frequency characteristics of the load respectively.

[0028] As a preferred solution, the load model further includes: a dynamic model;

[0029] The construction of load models corresponding to various types of adjustable resource loads includes:

[0030] For each type of adjustable resource load, fitting the load node data to obtain a corresponding load node data curve;

[0031] According to the voltage variation range in the load node data curve, the load node data curve is divided into a plurality of intervals, and the load node data curve in each interval is analyzed to obtain a mathematical expression corresponding to each interval;

[0032] The mathematical expressions corresponding to each interval are combined to obtain a dynamic model corresponding to the adjustable resource load.

[0033] As a preferred solution, the resource regulation and management module includes: a load model parameter identification unit and a multivariate load prediction unit;

[0034] The load model parameter identification unit is used to take the historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, train the load model corresponding to each type of adjustable resource load, and perform parameter identification on each load model during the training process to obtain the load prediction model corresponding to each type of adjustable resource load;

[0035] The multivariate load prediction unit is used to predict the predicted load output of each type of adjustable resource load according to the load prediction model.

[0036] As a preferred solution, the method uses historical load characteristic data as input, uses the predicted load output corresponding to the historical load characteristic data as output, trains the load models corresponding to each type of adjustable resource load, and performs parameter identification on each load model during the training process to obtain the load prediction models corresponding to each type of adjustable resource load, including:

[0037] Taking historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, the load model corresponding to each type of adjustable resource load is trained, and the optimal model action is selected according to the current state of the load model and the preset DDQN algorithm, and the model parameters of the load model are adjusted to minimize the error between the predicted load output and the actual load output of the load model, so as to obtain the load prediction model corresponding to each type of adjustable resource load.

[0038] As a preferred solution, the multivariate load forecasting unit is also used to evaluate the model performance of the load forecasting model according to preset mean square error index and mean absolute error index after predicting the predicted load output of each type of adjustable resource load.

[0039] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0040] The present invention provides an aggregated control system for demand-side adjustable resources, including an adjustable resource pool, a resource control management module and a resource dynamic interactive decision module; the adjustable resource pool is used to obtain load node data of adjustable resource load nodes, cluster according to the load node data, cluster to obtain different types of adjustable resource loads, and construct load models corresponding to each type of adjustable resource load; wherein the types of the adjustable resource loads include: industrial loads, commercial loads, residential loads and emerging loads; the resource control management module is used to take historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, train the load model corresponding to each type of adjustable resource load, obtain the load prediction model corresponding to each type of adjustable resource load, and then predict the predicted load output of each type of adjustable resource load according to the load prediction model; the resource dynamic interactive decision module is used to analyze the adjustable power of each type of adjustable resource load in different scheduling scenarios according to the predicted load output of each type of adjustable resource load, and then schedule the demand response of each type of adjustable resource load according to the adjustable power.

[0041] The present invention can construct a load model corresponding to each type of adjustable resource load, and then obtain a corresponding load prediction model based on the load model training, and obtain the predicted load output of each type of adjustable resource load through the load prediction model to schedule the demand response of each type of adjustable resource load. The present invention can use the load data of the demand-side adjustable resources to regulate the demand-side adjustable resources, and solve the problem of realizing the aggregate regulation and optimal utilization of the demand-side adjustable resources in the context of a large number of demand-side adjustable resources participating in the grid interaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a structural diagram of a system for aggregating and regulating demand-side adjustable resources provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0045] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined.

[0046] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0047] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0048] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0049] In the description of the embodiments of the present application, unless otherwise clearly specified and limited, technical terms such as "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0050] Embodiment 1

[0051] Please refer to Figure 1 , is a structural diagram of an aggregation control system of demand-side adjustable resources provided by an embodiment of the present invention, including: an adjustable resource pool, a resource control management module and a resource dynamic interactive decision module;

[0052] The adjustable resource pool is used to obtain load node data of adjustable resource load nodes, cluster the load node data, obtain different types of adjustable resource loads by clustering, and construct load models corresponding to each type of adjustable resource load; wherein the types of adjustable resource loads include: industrial load, commercial load, residential load and emerging load;

[0053] The resource regulation and management module is used to take the historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, train the load model corresponding to each type of adjustable resource load, obtain the load prediction model corresponding to each type of adjustable resource load, and then predict the predicted load output of each type of adjustable resource load according to the load prediction model;

[0054] The resource dynamic interactive decision module is used to analyze the adjustable power of each type of adjustable resource load in different scheduling scenarios according to the predicted load output of each type of adjustable resource load, and then schedule the demand response of each type of adjustable resource load according to the adjustable power.

[0055] Specifically, the purpose of the present invention is to provide a hierarchical aggregation and control system for adjustable resources on the demand side, so as to solve the problem of realizing multi-level aggregation and optimal utilization of resources in the context of a large number of adjustable resources on the demand side participating in the grid interaction. Figure 1 As shown, the demand-side adjustable resource hierarchical aggregation control system includes: a primary platform adjustable resource pool, a secondary platform resource control management module, and a tertiary platform resource dynamic interactive decision-making module;

[0056] The adjustable resource pool is used to interact with the measuring device and the multi-load general model library to complete the construction of a general model of massive adjustable resources; the resource regulation and management module is used to interact with the adjustable resource load model, the multi-load forecasting system, and the auxiliary service system to complete multi-level aggregation of resources for comprehensive modeling and analysis to participate in interactive response; the resource dynamic interactive decision-making module is used to interact with the user demand system, the energy optimization strategy system, and the resource response potential assessment system to complete interactive response and optimization management of multiple types of resources.

[0057] Preferably, the adjustable resource pool includes: a data acquisition unit, a load multi-level clustering unit and an adjustable resource model library; the data acquisition unit is used to connect to the measuring device on the bus of the integrated energy system, and obtain the load node data of the adjustable resource load node through the measuring device; wherein the load node data includes: voltage, current, frequency and power; the load multi-level clustering unit is used to extract the power consumption index representing the user's power consumption characteristics according to the load node data, and then cluster according to the power consumption index to obtain different types of adjustable resource loads; wherein the power consumption index includes: peak load, load factor and load change rate; the adjustable resource model library is used to construct load models corresponding to each type of adjustable resource load.

[0058] Preferably, the extracting of electricity consumption indicators representing the user's electricity consumption characteristics based on the load node data includes: inputting the load node data into a preset electricity consumption indicator extraction model, so that the electricity consumption indicator extraction model performs data preprocessing on the load node data, and extracts high-level features in the load node data after data preprocessing to obtain a corresponding implicit state vector, and then analyzing the implicit state vector to obtain and output the electricity consumption indicator representing the user's electricity consumption characteristics; wherein the data preprocessing includes: data cleaning, data normalization and filling of missing data values.

[0059] Preferably, clustering is performed according to the power consumption index to obtain different types of adjustable resource loads, including: calculating the corresponding average silhouette coefficient according to the power consumption index and a preset silhouette coefficient method, and finding the number of clusters when the average silhouette coefficient is maximum; dividing the adjustable resource load into clusters of the same number as the number of clusters according to the power consumption characteristics, the number of clusters and a preset K-means algorithm; calculating the central feature vector of each cluster, and determining the type of adjustable resource load of each cluster based on the central feature vector.

[0060] Preferably, the load model includes: a ZIP static model; the construction of load models corresponding to various types of adjustable resource loads includes: for each type of adjustable resource load, based on the load node data, constructing a constant impedance model when the equivalent impedance of the adjustable resource load is constant, a constant current model when the equivalent current of the adjustable resource load is constant, and a constant power model when the power of the adjustable resource load is constant; based on the constant impedance model, the constant current model and the constant power model, constructing a ZIP static model corresponding to the adjustable resource load.

[0061] Preferably, the ZIP static model is:

[0062] P=P0·[k Z (V t / V0) 2 +k I (V t / V0)+k P ](1+L P Δf / f0);

[0063] Q=Q0·[k′ Z (V t / V0) 2 +k′ I (V t / V0)+k′ P ](1+L Q Δf / f0);

[0064] k Z +k I +k′ P =1;

[0065] k′ Z +k′ I +k′ P =1;

[0066] Among them, P and Q are respectively the active power and reactive power actually absorbed by the total load; P0 and Q0 are respectively the active power and reactive power absorbed by all loads in rated operation state; k Z , k I and k P are the proportions of the active constant impedance part, the constant current part and the constant power part in the ZIP static model respectively; k′ Z , k′ I and k′ P are the proportions of the reactive constant impedance part, the constant current part and the constant power part in the ZIP static model respectively; V tand V0 are the voltage and initial voltage at time t respectively; f and f0 are the actual frequency and rated frequency of the power grid respectively; (1+L P Δf / f0) and (1+L Q Δf / f0) are the active frequency characteristics and reactive frequency characteristics of the load respectively.

[0067] Preferably, the load model also includes: a dynamic model; the construction of the load model corresponding to each type of adjustable resource load includes: for each type of adjustable resource load, fitting the load node data to obtain the corresponding load node data curve; according to the voltage variation range in the load node data curve, dividing the load node data curve into several intervals, and analyzing the load node data curve in each interval to obtain the mathematical expression corresponding to each interval; combining the mathematical expressions corresponding to each interval to obtain the dynamic model corresponding to the adjustable resource load.

[0068] Specifically, the adjustable resource pool includes a data acquisition unit, a load multi-level clustering unit, and an adjustable resource model library. It is used to interact with the measurement device and the multi-load general model library to complete the construction of a massive adjustable resource general model; considering the characteristics of different loads, according to the data obtained by the measurement device, the power load data clustering analysis method is used to classify the adjustable resource loads and build a resource load model.

[0069] The data acquisition unit measures the tested load data through a PMU device configured on the bus of the integrated energy system to obtain voltage, current, frequency and power information of the load node connected to the adjustable resource, including historical load curves, power consumption, and power consumption time.

[0070] The load multi-level clustering unit pre-processes the collected data and uses the LSTM network (i.e., the user index extraction model) to extract key indicators that can represent the user's power consumption characteristics, such as peak load, load factor, and load change rate, from the original data. User power consumption characteristics often have strong time correlation (for example, daily cycle, weekly cycle, seasonality), and LSTM can effectively capture the time-related characteristics of user power consumption through its unique gating mechanism (input gate, forget gate, and output gate). First, the LSTM network inputs data and preprocesses data, including data cleaning, normalization, and missing value filling; second, through its cyclic structure, it gradually receives each time point of the time series data (such as hourly electricity consumption data), and selectively memorizes or forgets historical information through a gating mechanism; finally, through a multi-layer LSTM structure, the network will gradually extract high-level features and eventually output a latent state vector (such as the periodicity, peak characteristics, and volatility of electricity consumption). This vector can be regarded as a high-dimensional representation of the user's electricity consumption behavior and can reflect the user's electricity consumption habits; finally, the output of LSTM can be directly used to generate key indicators: for example, by analyzing the time pattern represented in the latent state, the user's daily or weekly peak electricity consumption time can be identified, and by analyzing the amplitude of the change in the latent state, the degree of fluctuation of the user's electricity load can be calculated. By observing the LSTM's memory ability for long-term historical data, the user's periodic electricity consumption pattern rules can be extracted, and finally the key indicators that can represent the user's electricity consumption characteristics are extracted.

[0071] Different load patterns are identified using the K-means clustering method. First, the time series data of user electricity consumption is collected and preprocessed, including data cleaning, feature extraction and standardization. Subsequently, the silhouette coefficient method is used to find the value of the number of clusters K that maximizes the average silhouette coefficient S, and the user's characteristics are input into the K-means algorithm. K-means divides users into K clusters through iterative optimization, and each cluster represents a load pattern. Finally, by analyzing the central feature vector of each cluster, different load patterns can be identified and explained, providing support for power dispatching, demand response and user behavior analysis. According to the response characteristics of different types of loads in the virtual power plant, the adjustable resource loads are divided into four categories: industrial loads, commercial loads, residential loads, and emerging loads.

[0072] Finally, it is necessary to design corresponding models according to the characteristics and regulation capabilities of different load types, so as to form a comprehensive adjustable resource model library to support the demand response plan of the power grid. Analyzing the load types of adjustable resources provides a basis for building an adjustable resource model library. By clarifying the characteristics and adjustable equipment of industrial, commercial, residential and emerging loads, the regulation characteristic parameters of each type of load can be extracted and the corresponding mathematical model can be constructed. According to whether the dynamic characteristics of the load are reflected, the mathematical model of multivariate load can be divided into static model and dynamic model. At the same time, according to the regulation strategies of different load types, these models are integrated into the adjustable resource model library to provide accurate resource scheduling basis and decision support for demand response, peak shaving and valley filling and power grid optimization scheduling. The load accounts for a large proportion of industrial loads and the power consumption equipment that is easy to regulate is rolling mills and electric arc furnaces. During the demand response period, according to the different needs of adjusting the load upward or downward, the power impact of the steel rolling load can be used to arrange the production / stop periods of the steel rolling mill into specific plans to complete the demand response instructions, or the demand response instructions can be completed by adjusting the production rhythm before and after the arc furnace; commercial loads usually reach their peak during the day on weekdays and may decrease at night and on weekends. By adjusting the power consumption patterns of adjustable loads including air-conditioning load, lighting load, and refrigeration load, in response to electricity price signals or grid demand, economic incentives can be obtained; the characteristics of residential loads are greatly affected by their living and working habits. The adjustable loads are mainly air-conditioning, water heaters, and lighting electricity, which encourage residential users to participate in demand response; emerging loads mainly refer to communication base stations. During the peak load period, the macro station load can participate in peak response regulation. By disconnecting the external AC power supply, the energy storage battery provides power to the macro station.

[0073] By establishing a mathematical model in which multiple loads are independent of each other, the uncertainty of load resources and the diversity of load modeling are clarified. It should be noted that the adjustable resource model library includes static models and dynamic models. The static model adopts a comprehensive ZIP load model. The model uses polynomials to describe the functional relationship between node voltage amplitude, system frequency and power. It connects the constant impedance model (Z), constant current model (I) and constant power model (P) in parallel. The constant impedance model (Z), constant current model (I) and constant power model (P) are the basic forms of static load models, which are obtained through experimental measurement, theoretical derivation and engineering experience. By actually testing the response of the load at different voltage levels, recording the relationship between the active power and reactive power of the load and the voltage, and fitting the mathematical relationship between power and voltage based on the measured data, the load model is determined. At the same time, according to the physical characteristics of the load, the relationship between load power and voltage can also be derived from the circuit principle. In the constant impedance model, the load power is proportional to the square of the voltage, assuming that the equivalent impedance of the load is constant; in the constant current model, the load power is proportional to the voltage, assuming that the equivalent current of the load is constant; in the constant power model, the load power is independent of the voltage, assuming that the power of the load is constant. The construction of the static model is based on the voltage-power characteristics of the load, and is mainly used for steady-state analysis of the power system, voltage stability research, and demand response assessment scenarios, and has important theoretical and engineering application value.

[0074] Wherein, the ZIP static load model is:

[0075]

[0076] Among them, P and Q are respectively the active power and reactive power actually absorbed by the total load; P0 and Q0 are the active power and reactive power absorbed by all loads under rated operation; k Z , k I , k P are the proportions of active constant impedance, constant current, and constant power in the ZIP model; k' Z , k I ', k' P are the proportions of reactive constant impedance, constant current, and constant power in the ZIP model; V t and V0 are the voltage and initial voltage at time t respectively; f and f0 are the actual frequency and rated frequency of the power grid; (1+L P Δf / f0)、(1+L Q Δf / f0) describes the active power-frequency characteristic and reactive power-frequency characteristic of the load respectively.

[0077] It should be noted that the dynamic model takes into account the time factor and dynamic response characteristics, and can more realistically reflect the time response characteristics of the load to disturbances. The dynamic model fits the load data curve based on the historically collected load data, and uses the segmented curve analysis method to construct the mathematical model of the load. Under different operating conditions, the actual response data of the load is measured, including the changing relationship between active power, reactive power and voltage or frequency. The data is processed in segments, and the operating characteristic curve of the load is divided into several intervals according to the range of voltage or frequency. In each interval, the characteristics of the load can be approximated as a certain linear or nonlinear relationship. The measured data is curve-fitted using an appropriate function form (such as linear function, polynomial, exponential function) to obtain a segmented mathematical expression. Finally, the segmented relationships are combined to form a complete load mathematical model. The equivalent models of multivariate loads constitute a model library, which provides various types of load composition analysis for the load aggregation of virtual power plants, and is helpful for the energy resource integration and operation optimization of virtual power plants.

[0078] Preferably, the resource regulation and management module includes: a load model parameter identification unit and a multivariate load prediction unit; the load model parameter identification unit is used to take historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, to train the load model corresponding to each type of adjustable resource load, and to perform parameter identification on each load model during the training process to obtain the load prediction model corresponding to each type of adjustable resource load; the multivariate load prediction unit is used to predict the predicted load output of each type of adjustable resource load based on the load prediction model.

[0079] Preferably, the method uses historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, trains the load models corresponding to each type of adjustable resource load, and performs parameter identification on each load model during the training process to obtain the load prediction models corresponding to each type of adjustable resource load, including: using historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, trains the load models corresponding to each type of adjustable resource load, selects the optimal action of the model according to the current state of the load model and a preset DDQN algorithm, adjusts the model parameters of the load model to minimize the error between the predicted load output of the load model and the actual load output, and obtains the load prediction models corresponding to each type of adjustable resource load.

[0080] Preferably, the multivariate load forecasting unit is further used to evaluate the model performance of the load forecasting model according to a preset mean square error index and a mean absolute error index after predicting the predicted load output of each type of adjustable resource load.

[0081] Specifically, the resource regulation and management module includes a load model parameter identification unit and a multi-element load forecasting unit. It is used to interact with the adjustable resource load model, the multi-element load forecasting system, and the auxiliary service system, complete the multi-level aggregation of resources for comprehensive modeling analysis and interactive response. It identifies the parameters of the general model based on the historical load data, establishes the equivalent model of the multi-element adjustable resource load, and uses deep reinforcement learning to predict the output of the load model to participate in the virtual power plant demand response.

[0082] The load model parameter identification unit uses the deep reinforcement learning DDQN algorithm to identify the parameters of the mathematical model of the adjustable resource model library. The goal of parameter identification is to find the optimal parameters through input and output data so that the model output is as close to the output of the real system as possible. In the reinforcement learning framework, the parameter identification problem is regarded as a sequential decision problem, the model parameters are regarded as the state s, the parameter adjustment direction and amplitude are regarded as the action a, and the negative value of the error between the model output and the true value is defined as the reward r. The DDQN algorithm is used to optimize the parameter adjustment strategy in exploration and utilization, and the target Q value is calculated through the main network and the target network. The model parameters are gradually updated iteratively so that its output is close to the behavior of the real system, and accurate parameter identification is achieved. The intelligent agent of the DDQN algorithm considers the influence of environmental factors by observing the changes in environmental status and reward signals, learns and formulates strategies through interaction with the environment, maximizes its long-term rewards, and achieves the goal of minimizing the power error of the equivalent model, ensuring that the equivalent model can accurately reflect the characteristics and behavior of the actual load, and avoid decision deviations or system operation risks caused by excessive model errors. When the agent trains a large number of samples during the learning process, it can fully understand the dynamic characteristics of the environment and make reliable decisions to find a strategy that can maximize the cumulative reward, thereby achieving effective control and optimization of the complex environment and making the load equivalent model have high accuracy.

[0083] The training of the DDQN model is achieved through interaction with the environment. The agent selects action a according to the current state at each time step, obtains rewards through interaction with the environment and transfers to the next state s, and then stores the experience in the experience replay pool. Using randomly selected experience samples, the target Q value is calculated and the parameters of the main network are updated. At the same time, the parameters of the target network are updated regularly to gradually optimize the strategy. Load forecasting is achieved by building a prediction model based on historical data and optimizing the model parameters in combination with the DDQN algorithm. Using the historical data of the load as input, DDQN selects the optimal action according to the current state and adjusts the model parameters to minimize the error between the predicted value and the actual load value. In the prediction stage, the relevant feature data of the future is input, and the model generates the corresponding load prediction value according to the optimized parameters, thereby realizing accurate prediction of future loads, dynamically adapting to different environmental characteristics, and finally achieving high-precision load forecasting.

[0084] The multivariate load forecasting module uses the trained DDQN model for load forecasting, and uses the mean square error (MSE) indicator and the mean absolute error (MAE) indicator to evaluate the model prediction performance. The multivariate adjustable resource load forecast output participates in the virtual power plant demand response to achieve effective scheduling and optimal utilization of resources. When using the mean square error (MSE) and the mean absolute error (MAE) to evaluate the model prediction performance, it is first necessary to calculate the error between the model prediction value and the true value. MSE reflects the overall volatility of the error by calculating the square average of all prediction errors, and is more sensitive to larger errors; its formula is: in is the predicted value, y i is the true value, and n is the number of samples. MAE reflects the average level of error by calculating the average absolute value of all prediction errors; its formula is Through these two indicators, the prediction performance of the model can be comprehensively evaluated. MSE focuses more on large errors, while MAE more intuitively reflects the average error size.

[0085] Specifically, the resource dynamic interactive decision-making module includes an optimization and control strategy unit block and a resource response potential assessment unit. It is used to interact with the user demand system, energy optimization strategy system, and resource response potential assessment system to complete interactive response and optimization management of multiple types of resources. Since the integrated energy system has different requirements for regulation at different time scales, according to the different technical characteristics of each type of demand, the adjustable resource load is involved in short-term scale regulation, intraday load regulation, and long-term scale regulation, and the demand response potential of the adjustable resources is analyzed.

[0086] The optimization and control strategy unit guides the load resource subject to actively participate, predicts and develops quantitative modeling for user electricity demand and adjustable resources, comprehensively considers response capacity, response rate, and response time, and shows the adjustable resource response characteristics under various market incentive levels. First, it is necessary to collect historical electricity consumption data, environmental factors, and user behavior pattern related data, extract key features, build a prediction model based on historical data, and optimize the model parameters in combination with the DDQN algorithm to achieve this. Then, based on these features, a load model is built, and through DDQN agent training, the model can accurately capture the dynamic change characteristics of user electricity demand and adjustable resources, so as to minimize the error between the predicted value and the actual load value. In the prediction stage, the relevant feature data of the future is input, and the model generates the corresponding load prediction value according to the optimized parameters, thereby realizing accurate prediction of future loads.

[0087] The resource response potential evaluation unit is constructed based on the resource response potential under a given excitation signal, and comprehensively considers the influence of multiple response characteristics such as response capacity, response rate, and response time. The acquisition of resource response potential is based on resource characteristic modeling and historical data analysis. First, by collecting the operation data of the resource (load curve, equipment operation status) and external factors (environmental conditions, user behavior patterns), the maximum adjustable range, response speed and duration key characteristics of the resource are evaluated in combination with the load model. Then, given an excitation signal (such as local electricity price incentives and carbon emission restrictions), the dynamic response behavior of the resource is analyzed to quantify the adjustable power or capacity of the adjustable resource, that is, the resource response potential. The power consumption curves under various incentives are subtracted from the power consumption curves under no incentives, and the adjustable resource response potential under different incentives is evaluated from multiple perspectives to adapt to the scenarios of short-term scale regulation, intraday load regulation, and long-term scale regulation. To evaluate the adjustable resource response potential under different incentives from multiple perspectives, it is necessary to comprehensively consider the technical characteristics, economics and multi-dimensional factors of user behavior of the resource. From a technical perspective, the response speed, adjustment range and duration of the load model under different incentive signals (such as price incentives and carbon emission restrictions) are solved to quantify its dynamic adjustment ability; from an economic perspective, the impact of incentive signals on user benefits is evaluated, and the user's willingness to participate in regulation is analyzed and determined; from a behavioral perspective, historical data is combined to mine the impact of user behavior patterns on resource response and evaluate the resource response potential.

[0088] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A demand-side adjustable resource aggregation control system, characterized in that: include: Adjustable resource pool, resource regulation and management module, and resource dynamic interactive decision-making module; The adjustable resource pool is used to obtain load node data of adjustable resource load nodes, cluster the load node data, obtain different types of adjustable resource loads by clustering, and construct load models corresponding to each type of adjustable resource load; wherein the types of adjustable resource loads include: industrial load, commercial load, residential load and emerging load; The resource regulation and management module is used to take the historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, train the load model corresponding to each type of adjustable resource load, obtain the load prediction model corresponding to each type of adjustable resource load, and then predict the predicted load output of each type of adjustable resource load according to the load prediction model; The resource dynamic interactive decision module is used to analyze the adjustable power of each type of adjustable resource load in different scheduling scenarios according to the predicted load output of each type of adjustable resource load, and then schedule the demand response of each type of adjustable resource load according to the adjustable power.

2. The aggregate control system for demand-side adjustable resources as claimed in claim 1, characterized in that: The adjustable resource pool includes: a data acquisition unit, a load multi-level clustering unit and an adjustable resource model library; The data acquisition unit is used to connect with the measuring device on the bus of the integrated energy system, and obtain the load node data of the adjustable resource load node through the measuring device; wherein the load node data includes: voltage, current, frequency and power; The load multi-level clustering unit is used to extract power consumption indicators representing user power consumption characteristics according to the load node data, and then cluster according to the power consumption indicators to obtain different types of adjustable resource loads; wherein the power consumption indicators include: peak load, load factor and load change rate; The adjustable resource model library is used to construct load models corresponding to various types of adjustable resource loads.

3. The aggregate control system for demand-side adjustable resources as claimed in claim 2, characterized in that: The step of extracting a power consumption index representing a user's power consumption characteristics based on the load node data includes: The load node data is input into a preset electricity consumption index extraction model so that the electricity consumption index extraction model performs data preprocessing on the load node data, and extracts high-level features in the load node data after data preprocessing to obtain the corresponding implicit state vector, and then analyzes the implicit state vector to obtain and output the electricity consumption index used to represent the user's electricity consumption characteristics; wherein the data preprocessing includes: data cleaning, data normalization and data missing value filling.

4. The aggregate control system for demand-side adjustable resources as claimed in claim 3, characterized in that: The clustering is performed according to the power consumption indicators to obtain different types of adjustable resource loads, including: According to the power consumption index and the preset silhouette coefficient method, the corresponding average silhouette coefficient is calculated, and the number of clusters when the average silhouette coefficient is the largest is found; According to the power consumption characteristics, the number of clusters and a preset K-means algorithm, the adjustable resource load is divided into clusters of the same number as the number of clusters; The central feature vector of each cluster is calculated, and the type of the adjustable resource load of each cluster is determined according to the central feature vector.

5. The aggregate control system for demand-side adjustable resources as claimed in claim 4, characterized in that: The load model includes: a ZIP static model; The construction of load models corresponding to various types of adjustable resource loads includes: For each type of adjustable resource load, constructing a constant impedance model when the equivalent impedance of the adjustable resource load is constant, a constant current model when the equivalent current of the adjustable resource load is constant, and a constant power model when the power of the adjustable resource load is constant according to the load node data; According to the constant impedance model, the constant current model and the constant power model, a ZIP static model corresponding to the adjustable resource load is constructed.

6. The aggregate control system for demand-side adjustable resources as claimed in claim 5, characterized in that: The ZIP static model is: P=P0·[k Z (V t / V0) 2 +k I (V t / V0)+k P ](1+L P Δf / f0); Q=Q0·[k' Z (V t / V0)) 2 +k′ I (V t / V0)+k' P ](1+L Q Δf / f0); k Z +k I +k P =1; k' Z +k′ I +k′ P =1; Among them, P and Q are respectively the active power and reactive power actually absorbed by the total load; P0 and Q0 are respectively the active power and reactive power absorbed by all loads in rated operation state; k Z , k I and k P are the proportions of the active constant impedance part, the constant current part and the constant power part in the ZIP static model respectively; k' Z , k′ I and k′ P are the proportions of the reactive constant impedance part, the constant current part and the constant power part in the ZIP static model respectively; V t and V0 are the voltage and initial voltage at time t respectively; f and f0 are the actual frequency and rated frequency of the power grid respectively; (1+L P Δf / f0) and (1+L Q Δf / f0) are the active frequency characteristics and reactive frequency characteristics of the load respectively.

7. The aggregate control system for demand-side adjustable resources as claimed in claim 6, characterized in that: The load model further includes: a dynamic model; The construction of load models corresponding to various types of adjustable resource loads includes: For each type of adjustable resource load, fitting the load node data to obtain a corresponding load node data curve; According to the voltage variation range in the load node data curve, the load node data curve is divided into a plurality of intervals, and the load node data curve in each interval is analyzed to obtain a mathematical expression corresponding to each interval; The mathematical expressions corresponding to each interval are combined to obtain a dynamic model corresponding to the adjustable resource load.

8. The aggregate control system for demand-side adjustable resources according to claim 1, characterized in that: The resource regulation and management module includes: a load model parameter identification unit and a multivariate load prediction unit; The load model parameter identification unit is used to take the historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, train the load model corresponding to each type of adjustable resource load, and perform parameter identification on each load model during the training process to obtain the load prediction model corresponding to each type of adjustable resource load; The multivariate load prediction unit is used to predict the predicted load output of each type of adjustable resource load according to the load prediction model.

9. The aggregate control system for demand-side adjustable resources as claimed in claim 8, characterized in that: The method uses historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, trains the load models corresponding to various types of adjustable resource loads, and performs parameter identification on each load model during the training process to obtain the load prediction models corresponding to various types of adjustable resource loads, including: Taking historical load characteristic data as input and the predicted load output corresponding to the historical load characteristic data as output, the load model corresponding to each type of adjustable resource load is trained, and the optimal model action is selected according to the current state of the load model and the preset DDQN algorithm, and the model parameters of the load model are adjusted to minimize the error between the predicted load output and the actual load output of the load model, so as to obtain the load prediction model corresponding to each type of adjustable resource load.

10. The aggregate control system for demand-side adjustable resources according to claim 9, characterized in that: The multivariate load forecasting unit is also used to evaluate the model performance of the load forecasting model according to a preset mean square error index and a mean absolute error index after predicting the predicted load output of each type of adjustable resource load.

Images