Distributed resource dynamic aggregation method based on improved regression strategy

Abstract

The invention provides a distributed resource dynamic aggregation method based on an improved regression strategy, and the method specifically comprises the steps: constructing a quantile regression forest model, and training the quantile regression forest model according to the historical data of distributed resources; performing output prediction on the distributed resources according to the quantile regression forest model, and correspondingly constructing a probability prediction curve of the output of the distributed resources; and dynamically aggregating distributed resource output into a generalized energy storage model based on a probability prediction curve in combination with grid constraint and equipment constraint, and participating in power grid dispatching. According to the method, the quantile regression forest model of the improved regression strategy is utilized to be compatible with the spatio-temporal correlation and uncertainty of the distributed resources, accurate prediction of the output condition of the distributed resources is achieved, the output of the distributed resources is dynamically aggregated into the generalized energy storage model under the specific constraint condition to support power grid dispatching, and the power grid dispatching efficiency is improved. The utilization rate of distributed resources is improved, and the accuracy of power grid dispatching is further improved.

Description

Technical Field

[0001] The present invention relates to the field of smart grids, and in particular to a distributed resource dynamic aggregation method based on an improved regression strategy. Background Art

[0002] With the large-scale access of distributed renewable energy sources such as solar energy and wind energy, the structure and operating characteristics of the power system have changed significantly. The output of such distributed resources has obvious spatio-temporal correlation and uncertainty. Taking wind power generation as an example, for distributed resources in different geographical locations, the wind energy resource conditions are greatly affected by factors such as terrain and climate, which will lead to different output situations. Taking photovoltaic power generation as another example, the output of the same distributed resource at different times is also greatly fluctuated due to the influence of light conditions.

[0003] Affected by the uncertainty of the output of the above-mentioned distributed resources, it is difficult to effectively integrate and utilize the output of distributed resources. This leads to difficulties in reasonably utilizing distributed resources when scheduling and planning the access of distributed resources to the power grid through traditional power system scheduling methods, and the accuracy of power grid scheduling is relatively low. Summary of the Invention

[0004] The object of the present invention is to overcome the disadvantages in the prior art that it is difficult to effectively integrate and utilize the output of distributed resources, and it is difficult to reasonably utilize distributed resources when scheduling and planning the access of distributed resources to the power grid through traditional power system scheduling methods, and the accuracy of power grid scheduling is relatively low. A distributed resource dynamic aggregation method based on an improved regression strategy is provided. The quantile regression forest model of the improved regression strategy is used to be compatible with the spatio-temporal correlation and uncertainty of distributed resources, so as to accurately predict the output of distributed resources. Then, under the specific grid constraints and equipment constraints, the output of distributed resources is dynamically aggregated into a generalized energy storage model to support power grid scheduling, improve the utilization rate of distributed resources, and further improve the accuracy of power grid scheduling.

[0005] The object of the present invention is achieved by the following technical solutions:

[0006] A distributed resource dynamic aggregation method based on an improved regression strategy includes:

[0007] Construct a quantile regression forest model and train the quantile regression forest model according to the historical data of distributed resources;

[0008] Predict the output of distributed resources according to the quantile regression forest model;

[0009] Construct a probability prediction curve of the output of distributed resources according to the output prediction result;

[0010] Based on the probability prediction curve, combined with grid constraints and device constraints, the output of distributed resources is dynamically aggregated into a generalized energy storage model to participate in power grid dispatching.

[0011] The quantile regression forest model using an improved regression strategy is used to learn and train the historical data of distributed resources to better accommodate its spatio-temporal correlation and uncertainty, so as to accurately predict the output of distributed resources. Furthermore, under the conditions of considering specific grid constraints and device constraints, the output of distributed resources is dynamically aggregated into a generalized energy storage model, which can not only effectively integrate dispersed and intermittent distributed resources, smooth their output fluctuations, avoid unreasonable operation of devices, but also provide more flexibility and controllability for power grid dispatching. And aggregating the output of distributed resources into a generalized energy storage model to participate in power grid dispatching can also flexibly adjust the output of distributed resources according to the real-time needs of the power grid, achieve fast response and balance of power, further fully tap the potential of distributed resources, improve their utilization rate, reduce energy waste, and at the same time provide accurate information and reliable means for power grid dispatching to avoid dispatching errors.

[0012] Further, constructing the probability prediction curve of the output of distributed resources according to the output prediction result includes:

[0013] Obtain the output prediction value of the distributed resource at each quantile according to the output prediction result;

[0014] Taking the quantile level and the corresponding output prediction value as the corresponding horizontal and vertical coordinate values, draw the corresponding probability prediction curve for each time point.

[0015] Further, constructing the quantile regression forest model and training the quantile regression forest model according to the historical data of distributed resources includes:

[0016] Taking date, time, season and geographical location as feature variables and the historical output data of distributed resources as the target variable, construct a quantile regression forest model;

[0017] Based on the power grid dispatching requirements and the historical output data of distributed resources, set the quantiles of the quantile regression forest model;

[0018] Construct a training set according to the output data of distributed resources and train the quantile regression forest model according to the training set.

[0019] Further, the setting of the quantiles of the quantile regression forest model based on the power grid dispatching requirements and the historical output data of distributed resources includes:

[0020] Determine the dispatching scenario of the distributed resource according to the power grid dispatching requirements of the power grid in which the distributed resource participates;

[0021] Generate corresponding data distribution characteristics based on the historical output data of distributed resources;

[0022] Set the quantile interval according to the scheduling scenario of the distributed resources and the data distribution characteristics of the historical output data, and correspondingly set the quantiles of the quantile regression forest model.

[0023] Further, based on the probability prediction curve, combining grid constraints and device constraints, dynamically aggregating the distributed resource output into a generalized energy storage model includes:

[0024] Set the evaluation index and corresponding weight of the output prediction value;

[0025] According to the corresponding probability prediction curve, calculate the evaluation index value of the output prediction value at each quantile;

[0026] According to the corresponding evaluation index value, calculate the comprehensive evaluation value of the output prediction at each quantile by combining the corresponding weight;

[0027] Obtain the probability prediction curve at the corresponding quantile with the highest comprehensive evaluation value, and combine grid constraints and device constraints to dynamically aggregate the distributed resource output into a generalized energy storage model.

[0028] Further, based on the probability prediction curve, combining grid constraints and device constraints, dynamically aggregating the distributed resource output into a generalized energy storage model further includes:

[0029] Based on the obtained probability prediction curve, determine the output prediction value of the distributed resources in each time period;

[0030] According to the corresponding output prediction value, obtain the power regulation ability of the distributed resources in each time period;

[0031] Summarize the power regulation ability of the distributed resources in each time period to the corresponding connection point;

[0032] Based on the power regulation range of the connection point, combining grid constraints and device constraints, establish a generalized energy storage model.

[0033] Further, the grid constraints at least include the network node voltage and branch power flow constraints of the distributed resources.

[0034] Further, the device constraints at least include the power upper and lower limit constraints, ramp constraints and energy constraints of the devices corresponding to the distributed resources.

[0035] The beneficial effects of the present invention are:

[0036] The quantile regression forest model using an improved regression strategy is used to learn and train the historical data of distributed resources to better accommodate their spatio-temporal correlation and uncertainty, so as to accurately predict the output of distributed resources. Furthermore, under the conditions of considering specific grid constraints and equipment constraints, the output of distributed resources is dynamically aggregated into a generalized energy storage model, which can not only effectively integrate dispersed and intermittent distributed resources, smooth their output fluctuations, and avoid unreasonable operation of equipment, but also provide more flexibility and controllability for power grid dispatching. And aggregating the output of distributed resources into a generalized energy storage model to participate in power grid dispatching can also flexibly adjust the output of distributed resources according to the real-time demand of the power grid, achieve rapid response and balance of power, and further fully tap the potential of distributed resources, improve their utilization rate, reduce energy waste, and at the same time provide accurate information and reliable means for power grid dispatching to avoid dispatching errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a schematic flow chart of the present invention;

[0038] Figure 2 is a schematic diagram of the 24-hour-ahead power curve of a generalized energy storage corresponding to distributed resources in a certain place according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The present invention will be further described below with reference to the drawings and embodiments.

[0040] Embodiment:

[0041] A method for dynamically aggregating distributed resources based on an improved regression strategy, as Figure 1 shown, includes:

[0042] Construct a quantile regression forest model and train the quantile regression forest model according to the historical data of distributed resources;

[0043] Predict the output of distributed resources according to the quantile regression forest model;

[0044] Construct a probability prediction curve of the output of distributed resources according to the output prediction result;

[0045] Based on the probability prediction curve, combine grid constraints and equipment constraints, and dynamically aggregate the output of distributed resources into a generalized energy storage model to participate in power grid dispatching.

[0046] Distributed resources such as solar photovoltaics and wind power often exhibit complex correlations in time and space. For example, spatially, adjacent distributed photovoltaic power plants may have a certain correlation in their output power due to similar geographical locations and similar lighting conditions. Temporally, the load demands at different times of the day also have obvious periodicity and correlation. The quantile regression forest can construct multiple decision trees and consider multiple spatio-temporal feature variables during the node splitting process of the trees to extract this hidden spatio-temporal correlation information, thereby more accurately describing the variation law of distributed resources.

[0047] The uncertainty of distributed resources mainly comes from natural factors such as the random variation of light and wind speed, and human factors such as the uncertainty of user electricity consumption behavior. The quantile regression forest model can provide a specific prediction interval by estimating different quantiles to reflect the possible value range of distributed resources at different probability levels, thereby effectively handling uncertainty.

[0048] Therefore, in this embodiment, the quantile regression forest model is specifically used for the output power prediction of distributed resources to be compatible with the spatio-temporal correlation and uncertainty of distributed resources and ensure the accuracy of the distributed output power prediction results.

[0049] Among them, constructing the quantile regression forest model and training the quantile regression forest model according to the historical data of distributed resources includes:

[0050] Taking date, time, season, and geographical location as feature variables and the historical output data of distributed resources as the target variable, construct the quantile regression forest model;

[0051] Based on the power grid scheduling requirements and the historical output data of distributed resources, set the quantiles of the quantile regression forest model;

[0052] Construct a training set according to the output data of distributed resources and train the quantile regression forest model according to the training set.

[0053] The expression of the constructed quantile regression forest model is:

[0054] QRF(d,t,s,l,P pv (t),P ess (t),P ev (t),input)=(P pv (t+1),P ess (t+1),P ev (t+1),P h (t+1));

[0055] Among them, d is the date, t is the time, s is the season, l is the geographical location, Ppv P(t) is the photovoltaic output at time t. ess P(t) is the charge and discharge power of the energy storage at time t. ev P(t) is the charge and discharge power of the electric vehicle at time t. pv P(t + 1) is the photovoltaic output at time t + 1. ess P(t + 1) is the charge and discharge power of the energy storage at time t + 1. ev P(t + 1) is the charge and discharge power of the electric vehicle at time t + 1. h P(t + 1) is the small hydropower output at time t + 1, and input is the input of other factors affecting the output of distributed resources.

[0056] And considering that in actual applications, different business scenarios have different focuses on the selection of quantiles. For example, for the scheduling and planning of the power system, in order to ensure the stability of power supply, it is necessary to pay attention to lower quantiles, such as the output situation at 5% or 1%, to cope with the possible minimum power generation scenarios and make preparations for power reserve and allocation in advance. For the energy trading market, traders may be more concerned about higher quantiles, such as the output prediction at 90% or 95%, because this represents the power generation capacity of distributed resources under better conditions and helps to evaluate the potential maximum benefits.

[0057] At the same time, considering that the historical data distribution may have a skewed distribution phenomenon, for example, most output values are concentrated in the lower interval, and when high values occasionally appear, it is necessary to select more detailed quantiles, such as 1%, 5%, 10%, etc., to more comprehensively capture the change range of the data and the output situation at different probability levels.

[0058] Therefore, when constructing the quantile regression forest model, combined with the grid scheduling requirements and the historical output data distribution of distributed resources, set the quantiles of the quantile regression forest model to ensure that the output prediction of the quantile regression forest model can be closer to the actual scenario and further improve the prediction accuracy.

[0059] Specifically, based on the grid scheduling requirements and the historical output data of distributed resources, set the quantiles of the quantile regression forest model, including:

[0060] Determine the scheduling scenario of distributed resources according to the grid scheduling requirements of the grid in which distributed resources participate;

[0061] Generate corresponding data distribution characteristics according to the historical output data of distributed resources;

[0062] Set the quantile interval according to the scheduling scenario of distributed resources and the data distribution characteristics of historical output data, and correspondingly set the quantiles of the quantile regression forest model.

[0063] In this embodiment, the quantiles of the quantile regression forest model are specifically set to 10%, 25%, 50%, 75%, and 95%.

[0064] Then, collect the historical output data of distributed resources to construct a corresponding training set, and realize the learning and training of the quantile regression forest model.

[0065] Among them, distributed resources specifically include photovoltaic, small hydropower, flexible load, and energy storage. Collect the historical output data, load data, meteorological data, and other factors that may affect the output of distributed resources corresponding to each device of distributed resources. The meteorological data mainly includes information such as temperature, humidity, wind speed, and wind direction.

[0066] Then, extract the features that can reflect spatio-temporal correlation from the collected original data, such as time features such as date, time, and season, spatial features such as geographical location and adjacent relationship, and meteorological features.

[0067] Construct a training set with the extracted feature data. On this basis, construct and learn and train the quantile regression forest model.

[0068] After completing the learning and training of the quantile regression forest model, predict the output of distributed resources for a period of time in the future according to the trained quantile regression forest model.

[0069] Among them, the quantile regression forest model can generate corresponding prediction values for each time point and each quantile value.

[0070] On this basis, construct a probability prediction curve for the output of distributed resources according to the output prediction results, which specifically includes the following steps:

[0071] Obtain the output prediction values of distributed resources at each quantile according to the output prediction results;

[0072] Use the quantile level and the corresponding output prediction value as the corresponding horizontal and vertical coordinate values, and draw a corresponding probability prediction curve for each time point.

[0073] The focuses of different grid dispatching requirements are different, and the dispatching focuses reflected by the probability prediction curves of quantile levels are also different. Therefore, specific evaluation indicators need to be set in combination with the actual grid dispatching requirements to select the most suitable probability prediction curve of the quantile, so as to realize the subsequent dynamic aggregation of the output of distributed resources and ensure the accuracy of grid dispatching when distributed resources participate in grid dispatching subsequently.

[0074] Among them, the specific steps for selecting the probability prediction curve include:

[0075] Set the evaluation indicators and corresponding weights of the output prediction values;

[0076] Calculate the evaluation index values of the output prediction values at each quantile according to the corresponding probability prediction curve;

[0077] According to the corresponding evaluation index values, calculate the comprehensive evaluation values of the output prediction at each quantile by combining the corresponding weights;

[0078] Obtain the probability prediction curve at the corresponding quantile with the highest comprehensive evaluation value, and combine the grid constraint and device constraint to dynamically aggregate the distributed resource output into a generalized energy storage model.

[0079] For the evaluation index of the output prediction value, it can be set as the load tracking error, output volatility, root mean square error, etc. Starting from the power supply stability and prediction accuracy, the selection of the quantile is realized.

[0080] For the weights of each evaluation index, they can be allocated and set according to requirements.

[0081] In this embodiment, specifically select the probability prediction curve at the 95% quantile, and combine the grid constraint and device constraint to dynamically aggregate the distributed resource output into a generalized energy storage model.

[0082] Specifically, based on the probability prediction curve, combining the grid constraint and device constraint, dynamically aggregating the distributed resource output into a generalized energy storage model includes:

[0083] Based on the obtained probability prediction curve, determine the output prediction values of the distributed resources in each time period;

[0084] According to the corresponding output prediction values, obtain the power regulation capabilities of the distributed resources in each time period;

[0085] Summarize the power regulation capabilities of the distributed resources in each time period to the corresponding connection point;

[0086] Based on the power regulation range of the connection point, combining the grid constraint and device constraint, establish a generalized energy storage model.

[0087] For the constraint conditions of the distributed resources, based on the linear power flow model, the linear operation constraint conditions among the decision variables can be simply expressed as a compact matrix form constraint: Kx ≤ L. Where x represents the decision variable vector composed of the active and reactive powers of each flexible distributed resource device at each moment, and K and L respectively represent the constant matrix and vector constituting the linear inequality constraint parameters.

[0088] Based on the linear power flow model, a power balance model P0 = Bx + c can be constructed, where P0 is the active power at the connection point at each moment, and B and c are the constant matrix and vector constituting the linear equality constraint parameters.

[0089] Based on this, the technical constraint scope of the overall distributed resource cluster is as follows:

[0090] P1 = {[x P0] T |Kx ≤ L; P0 = Bx + c};

[0091] Among them, P1 is the adjustable range of active power of the distributed resource cluster, which is jointly defined by linear equality and inequality constraint conditions.

[0092] After realizing dynamic aggregation, the corresponding constraint conditions also need to be adjusted accordingly, which can be specifically represented by the power constraint conditions at each moment at the grid connection point: P2 = {P0|AP0 ≤ d}, where A and d are the parameter matrix and vector of P2 respectively.

[0093] In this embodiment, the grid constraints at least include the network node voltage and branch power flow constraints of the distributed resources. The equipment constraints at least include the upper and lower power limits, ramp constraints, and energy constraints of the equipment corresponding to the distributed resources. According to the setting basis of the above constraint conditions and combined with the actual situation of the given constraint conditions, the construction of the generalized energy storage model is realized.

[0094] Moreover, the generalized energy storage model aggregates various types of distributed resources. For different distributed resources, different processing methods are set to realize the identification of power regulation capabilities and constraint conditions, and then the subsequent construction of the generalized energy storage model is realized.

[0095] Among them, for the photovoltaic power generation system, the schedulable photovoltaic power generation in the region is aggregated according to the regional geographical location to form N photovoltaic groups. The output of the i-th photovoltaic group at time t predicted by the quantile regression forest is denoted as P iPV (t).

[0096] Then, for the small hydropower stations in the region, the following expressions of output and constraint conditions are constructed:

[0097]

[0098]

[0099] Among them, represents the output of the i-th small hydropower station at time t; represents the power generation efficiency of the i-th small hydropower station; represents the reservoir water volume of the i-th small hydropower station; and respectively represent the upper and lower limits of the output of the i-th small hydropower station; and respectively represent the power generation water consumption, water inflow and water abandonment of the i-th small hydropower station at time t; Indicates the change in reservoir water volume of the \(i\)-th small hydropower station at a certain moment; and respectively represent the upper and lower limits of the reservoir storage capacity of the \(i\)-th small hydropower station.

[0100] For flexible loads, taking electric vehicles as an example, the corresponding constraint conditions can be expressed as:

[0101]

[0102] Among them, and are respectively the charging and discharging powers of the \(v\)-th electric vehicle at time \(t\), are respectively the charging and discharging efficiencies of the \(v\)-th electric vehicle at time \(t\), and That is, the charging and discharging power under the agency mode of the electric vehicle aggregator EVA at time \(t\). and are the maximum charging power under the agency mode of the electric vehicle aggregator EVA at time \(t\), and is the maximum discharging power. and are the upper limits of the maximum charging power and the maximum discharging power of the \(v\)-th vehicle.

[0103] For the energy storage system, its constraint conditions are specifically:

[0104]

[0105]

[0106] Among them, and are respectively the charging and discharging powers of the \(d\)-th distributed energy storage at time \(t\), are respectively the charging and discharging efficiencies of the \(d\)-th distributed energy storage at time \(t\). and That is, the charging and discharging power under the management mode of the distributed energy storage aggregator DESA at time \(t\). and are the maximum charging power and the maximum discharging power at time \(t\) under the management mode of the distributed energy storage aggregator DESA.

[0107] Based on the output and constraint conditions of the above-mentioned distributed resources, the distributed resources are aggregated into an equivalent energy storage, thereby establishing the generalized energy storage model \(P\) GES .

[0108] Among them, the constraint conditions of the generalized energy storage model \(P\) GES are:

[0109]

[0110] Among them, and are the maximum charging power and maximum discharging power of the generalized energy storage at time t, and are the charging and discharging power of the generalized energy storage at time t. and respectively represent the minimum and maximum values of the state of charge of the generalized energy storage, represents the charging efficiency of the generalized energy storage, represents the discharging efficiency of the generalized energy storage, is the baseline energy consumption, and are the upper and lower limits of the ramping constraint.

[0111] Taking the actual power grid grid of a certain place as the computing power, using the collected 24-hour historical data of this area throughout the year as the training set, the quantile regression forest model is trained. Based on the trained model, the distributed resource output of this grid is predicted, and the 24-hour power curve of the day before of this grid equivalent to the generalized energy storage is output, specifically as Figure 2 shown.

[0112] It can be seen from Figure 2 that the positive power value in the figure represents that this grid receives power from the superior power grid, equivalent to the charging of the generalized energy storage; the negative power value represents that this grid supports the power consumption of the superior power grid, equivalent to the discharging of the generalized energy storage. In the subsequent power grid dispatching process, the dispatching plan for the distributed resource output can be realized according to the power situation of the generalized energy storage.

[0113] The above-described embodiments are only a preferred solution of the present invention, and do not impose any form of limitation on the present invention. There are other variations and modifications without exceeding the technical solutions recorded in the claims.

Claims

1. A distributed resource dynamic aggregation method based on an improved regression strategy, characterized in that Including: Construct a quantile regression forest model and train the quantile regression forest model according to the historical data of distributed resources; Perform output prediction on distributed resources according to the quantile regression forest model; Construct a probability prediction curve for the output of distributed resources based on the output prediction results; Based on the probability prediction curve, combined with grid constraints and device constraints, dynamically aggregate the output of distributed resources into a generalized energy storage model to participate in power grid dispatching.

2. The distributed resource dynamic aggregation method based on an improved regression strategy according to claim 1, wherein The constructing a probability prediction curve for the output of distributed resources based on the output prediction results includes: Obtain the output prediction values of distributed resources at each quantile according to the output prediction results; Taking the quantile level and the corresponding output prediction value as the corresponding horizontal and vertical coordinate values, draw the corresponding probability prediction curve for each time point.

3. The distributed resource dynamic aggregation method based on an improved regression strategy according to claim 1, wherein The constructing a quantile regression forest model and training the quantile regression forest model according to the historical data of distributed resources includes: Construct a quantile regression forest model with date, time, season, and geographical location as feature variables and the historical output data of distributed resources as target variables; Based on the power grid dispatching requirements and the historical output data of distributed resources, set the quantiles of the quantile regression forest model; Construct a training set according to the output data of distributed resources and train the quantile regression forest model according to the training set.

4. The distributed resource dynamic aggregation method based on an improved regression strategy according to claim 3, wherein The setting the quantiles of the quantile regression forest model based on the power grid dispatching requirements and the historical output data of distributed resources includes: Determine the dispatching scenarios of distributed resources according to the power grid dispatching requirements of the power grid in which distributed resources participate; Generate corresponding data distribution characteristics according to the historical output data of distributed resources; Set the quantile intervals according to the dispatching scenarios of distributed resources and the data distribution characteristics of historical output data, and correspondingly set the quantiles of the quantile regression forest model.

5. The distributed resource dynamic aggregation method based on an improved regression strategy according to claim 1, wherein The dynamically aggregating the output of distributed resources into a generalized energy storage model based on the probability prediction curve, combined with grid constraints and device constraints, includes: Set the evaluation indexes and corresponding weights of the output prediction values; Calculate the evaluation index values of the output prediction values at each quantile according to the corresponding probability prediction curve; Calculate the comprehensive evaluation values of the output prediction at each quantile according to the corresponding evaluation index values, combined with the corresponding weights; Obtain the probability prediction curve at the corresponding quantile with the highest comprehensive evaluation value, and dynamically aggregate the output of distributed resources into a generalized energy storage model in combination with grid constraints and device constraints.

6. The distributed resource dynamic aggregation method based on an improved regression strategy according to claim 5, wherein The dynamically aggregating the output of distributed resources into a generalized energy storage model based on the probability prediction curve, combined with grid constraints and device constraints, further includes: Based on the obtained probability prediction curve, determine the output prediction values of distributed resources in each time period; According to the corresponding output prediction values, obtain the power regulation capabilities of distributed resources in each time period; Summarize the power regulation capabilities of distributed resources in each time period to the corresponding grid connection points; Based on the power regulation range of the grid connection points, combined with grid constraints and device constraints, establish a generalized energy storage model.

7. The distributed resource dynamic aggregation method based on an improved regression strategy according to claim 6, wherein The grid constraints at least include the network node voltage and branch power flow constraints of distributed resources.

8. The distributed resource dynamic aggregation method based on an improved regression strategy according to claim 6, characterized in that The device constraints at least include the upper and lower power limits, ramping constraints, and energy constraints of the devices corresponding to distributed resources.

Images