Residential user equipment regulation and control method and system considering optimal segmentation of time interval

Abstract

The invention discloses a residential user equipment regulation and control method and system considering optimal segmentation of a time interval, and belongs to the field of residential user equipment regulation and control. The residential user equipment regulation and control method considering the optimal segmentation of the time interval comprises the following steps: establishing a renewable energy parameter and electricity price prediction model and evaluation index; constructing a fan and photovoltaic physical model, inputting the renewable energy source parameters into the physical model, and outputting a calculation result; building a residential user equipment model by taking the minimum regulation and control cost as a target, and building a time interval optimal segmentation regulation and control model by considering an evaluation index; and importing the prediction data into the time interval optimal segmentation regulation and control model, and solving the model by using an improved MOPSO algorithm to obtain a residential user equipment regulation and control result considering time interval optimal segmentation.

Description

Technical Field

[0001] The present invention belongs to the field of residential user equipment control, and in particular relates to a residential user equipment control method and system considering optimal time interval segmentation. Background Art

[0002] With the increasing prominence of optimization techniques and probabilistic forecasting techniques, new research results and challenges have emerged in the study of optimal control of microgrids (MGs) or residential user equipment. For example, forecasting models such as LSTM are used to obtain deterministic (point estimate) weather forecasts, which are then used to solve the cost and pollution minimization problems of MGs. However, it is well known that deterministic forecasts do not contain the uncertainty of different scenarios that may affect the design, size and control of MGs. Although probabilistic forecasts are used for uncertainty modeling, these works do not address the problem of forecast accuracy decaying over time, but rely on bias parameters to achieve robustness. Some of the same works use rolling-horizon forecasting and optimization; however, the fixed time resolution used in the prior art ignores the impact of accuracy decay within the optimization range and the associated cost-accuracy trade-off. Therefore, multi-time block selection with variable time scales provides a beneficial way to better utilize probabilistic forecasting and robust optimization in microgrid management. Summary of the invention

[0003] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a residential user equipment control method and system considering the optimal division of time intervals, thereby solving the problems in the prior art.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] A residential user equipment control method considering optimal time interval segmentation includes the following steps:

[0006] Establish prediction models and evaluation indicators for renewable energy parameters and electricity prices;

[0007] Construct wind turbine and photovoltaic physical models, input renewable energy parameters into the physical models, and output calculation results;

[0008] Based on the calculation results, a residential user equipment model is constructed with the goal of minimizing the control cost, and an optimal time interval segmentation control model is constructed by considering the evaluation index;

[0009] The predicted data is imported into the time interval optimal segmentation control model, and the improved MOPSO algorithm is used to solve the model to obtain the residential user equipment control results considering the optimal segmentation of time intervals.

[0010] Furthermore, the renewable energy prediction model is an LSTM network model; and the electricity price prediction model is a gradient boosting regression model.

[0011] Furthermore, the evaluation indicators of the prediction model of renewable energy and electricity price include: mean absolute error, root mean square error, determination coefficient and continuous sorting probability score.

[0012] Furthermore, the physical model of the wind turbine is:

[0013]

[0014] in, To predict wind power output, v in , v rat , v out They represent the cut-in speed, rated speed and cut-in speed of the turbine respectively; Indicates the rated power of the turbine; is the predicted wind speed at the turbine hub height, h wt The turbine hub is high. is the predicted near-surface wind speed, h ns is the height at which near-surface measurements are made and z0 represents the surface roughness.

[0015] Furthermore, the photovoltaic physical model is:

[0016]

[0017] In the formula, Output power for solar energy; is the predicted incident solar irradiance; is the rated power, G inc,STC is the incident solar irradiance, f dr is the conversion factor.

[0018] Furthermore, the residential user equipment model constructed with the goal of minimizing the regulation cost is:

[0019]

[0020] Where W VPP For the benefit of residential users, They represent the market revenue of residential users, internal electricity sales revenue, micro gas turbine operating cost, electric energy storage operating cost, demand response cost and internal electricity purchase cost respectively; represents the market electricity price during period t, represents the amount of electricity bid by residential users in the market during period t, When purchasing electricity on the market on behalf of residential users, The opposite is true when N Jrepresents the number of micro gas turbines managed by residential users, a j , b j is the cost coefficient of the jth micro gas turbine, is the power generation of the jth micro gas turbine in period t; N e The amount of electric energy storage managed for residential users, c i is the depreciation cost coefficient of the i-th electric energy storage, and are the charging power and discharging power of the i-th energy storage in period t, both of which are positive values; and The internal electricity purchase and sales prices set for residential users in period t, is the initial residential user load in period t, is the demand response compensation unit price during period t, is the demand response amount during period t; It is the contractual transaction electricity between residential users and internal power generation resources during period t.

[0021] Furthermore, the improved MOPSO algorithm uses non-dominated sorting MOPSO as a framework, incorporates the variation of NSGAII, and uses the improved MOPSO algorithm to solve the time interval optimal segmentation control model. The steps include:

[0022] Step 1: Input the time block t, PAD estimation, data prediction value and system parameters; initialize the position and velocity of particles, which represent individuals in the population; set the parameters of the algorithm, including the maximum number of iterations maxgen and the initial value of the inertia weight;

[0023] Step 2: Generate the initial population W using the input parameters. This population is the potential solution set of the problem.

[0024] Step 3: Calculate the objective function value for each particle in the population Q, perform non-dominated sorting on the particles according to their fitness values ​​to determine their superiority and inferiority, and stratify the non-inferior solution set to maintain population diversity;

[0025] Step 4: Select the optimal individual and global solutions of the particles, and use the improved MOPSO algorithm to update the velocity and position of the particles;

[0026] Step 5: Iterate and repeat steps 3 and 4 in each generation. At the end of each generation, update the set of non-inferior solutions to form a new generation of population P;

[0027] Step 6: Check whether the current number of iterations gen has reached the maximum number of iterations maxgen. If not, continue iterating; if reached, end the iteration;

[0028] Step 7: Calculate the objective function value of each particle, perform non-dominated sorting according to the fitness value, stratify the non-inferior solution set, and finally output the optimal Pareto solution set.

[0029] A residential user equipment control system considering optimal time interval segmentation includes:

[0030] Prediction model building module: Establish prediction models and evaluation indicators for renewable energy parameters and electricity prices;

[0031] Physical model building module: builds wind turbine and photovoltaic physical models, inputs renewable energy parameters into the physical models, and outputs calculation results;

[0032] A control model building module: based on the calculation results, a residential user equipment model is built with the goal of minimizing the control cost, and a time interval optimal segmentation control model is built considering the evaluation index;

[0033] And, the result output module: imports the predicted data into the time interval optimal segmentation control model, and uses the improved MOPSO algorithm to solve the model to obtain the residential user equipment control result considering the optimal segmentation of the time interval.

[0034] A computer storage medium stores a readable program, which can execute the residential user equipment control method considering the optimal division of time intervals when the program is running.

[0035] An electronic device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus;

[0036] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned residential user equipment control method considering optimal division of time intervals.

[0037] Beneficial effects of the present invention:

[0038] 1. A set of equipment control methods that take prediction accuracy into consideration is provided, which further improves the high error problem caused by prediction uncertainty faced by traditional scheduling methods.

[0039] 2. Through probabilistic prediction and robust optimization, the present invention can better handle the problem of prediction accuracy decaying over time, rather than relying on bias parameters to achieve robustness.

[0040] 3. By using multiple time block selections with variable time scales, the present invention can better adapt to different operating conditions and demand changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 is a flow chart of the residential user equipment control method of the present invention;

[0043] Figure 2 is a general schematic diagram of the time rolling window of the present invention;

[0044] Figure 3 It is the flow chart of the improved MOPSO algorithm of the present invention;

[0045] Figure 4 is a wind speed prediction diagram of the present invention;

[0046] Figure 5 is a solar illumination intensity prediction diagram of the present invention;

[0047] Figure 6 It is the time block division diagram under the Pareto optimal 10 scenarios of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Example 1

[0050] like Figure 1 As shown, the residential user equipment control method considering the optimal division of time intervals includes the following steps:

[0051] S1, establish prediction models and indicators for renewable energy parameters such as wind speed and solar radiation intensity and electricity prices;

[0052] ML (machine learning) models are used to predict renewable energy and electricity prices with an emphasis on multi-step ahead forecasts in the electricity market, in this particular case with a forecast horizon of 24 hours (one day). These models are designed for multivariate time series data with multiple outputs, ensuring that spatial relationships and time-based patterns of change in the data are effectively captured. In terms of forecasting, the model uses data from several previous time periods to predict future trends. The observed past time intervals are denoted as (hm, h-m+1,..., h). Using this data, subsequent intervals can be predicted, i.e. (h+1, h+2,..., h+n). This forecasting framework can be expressed as:

[0053]

[0054] D output =[d(h+1),d(h+2),...,d(h+n)](2)

[0055] Among them, d varp The input time series data representing multiple parameters; m and n specify the range of previous readings and the length of the desired forecast period, respectively; D input , D output They are the model input and output data respectively.

[0056] Before adding data to the AI ​​model, preprocessing operations are required: detecting and removing outliers, filling missing data, and applying the MinMaxScaler (maximum-minimum) transformation to obtain normalized data to optimize the model's prediction accuracy.

[0057] (1) Renewable energy parameter prediction

[0058] Accurate prediction of wind speed and solar radiation in renewable energy forecasting is crucial for effective energy management of microgrid systems. LSTM is an advanced recurrent neural network architecture that effectively solves the gradient vanishing or gradient exploding problems encountered by traditional RNNs when processing long sequence data through a unique gating mechanism. This network structure consists of a forget gate, an input gate, and an output gate, which work together to control the flow and update of information in the cell state, similar to an information conveyor belt that can transmit important long-term dependency information across time steps. At the same time, the hidden state of LSTM, as the output of the network, captures and retains the key information of the previous time step, providing a basis for predicting the next time step. In addition, LSTM shares the same weight parameters in all time steps, which not only simplifies the model structure, but also enhances the model's ability to learn the same pattern at different positions in the sequence. Through the gradient truncation technique, LSTM is able to maintain a stable backpropagation process, allowing the network to learn long-term dependencies, which makes LSTM perform well in a variety of sequence prediction tasks such as language models, text generation, speech recognition, and time series prediction. In the present invention, a Long Short-Term Memory (LSTM) network is integrated into a sequence-to-sequence architecture, and an ensemble approach is used to generate prediction intervals to address the inherent uncertainty in prediction.

[0059] In summary, the renewable energy prediction model adopts the LSTM network model.

[0060] (2) Electricity price forecast

[0061] Electricity price forecasting is an important component of microgrid operation management and optimization, enabling efficient energy trading and resource allocation. Since electricity prices are affected by multiple factors such as supply, demand, and local market dynamics, it is critical to capture these fluctuations.

[0062] Gradient Boosting Regression (GBR) is an ensemble learning technique that improves the prediction performance of a model by gradually adding weak predictive models (usually decision trees) to minimize the loss function. The core idea of ​​GBR is to build a strong predictor, which is a linear combination of multiple weak predictors, each of which is adjusted on the basis of the previous one to correct the residuals of the previous model. In each iteration of GBR, the prediction error of the current model is calculated, and then a new weak model is trained to fit these residuals. Finally, the prediction results of the new model are weighted by a certain learning rate and integrated into the final prediction results. The advantage of this method is that it can handle various types of data and problems, and it can usually produce high-accuracy prediction results, but it is also necessary to prevent overfitting and model complexity. In order to help residential users make informed decisions, gradient boosting regression (GBR) is used to provide reasonable prediction estimates. This method can not only predict electricity prices, but also evaluate the associated uncertainties through PI. GBR is an ensemble learning method that has proven its effectiveness in regression tasks. The technique improves predictions by processing the residuals of previous models. Although GBR can adapt to different loss functions, it performs well when paired with a quantization loss, transitioning from standard mean regression to quantization-based predictions. For a specified target quantization value τ (0 < τ < 1), the quantization loss function is defined as follows:

[0063]

[0064] In the formula, represents the loss function, τ represents the target score and 0<τ<1; I(·) represents the indicator function, which takes the value 1 when the condition in the parentheses is true, otherwise it takes the value 0.

[0065] Through this approach centered on quantifying losses, GBR forecasts can be customized to different scenarios of electricity price distribution. For example, by modeling the 0.1, 0.5, and 0.9 quantiles, it is possible to obtain forecasts that align with the 10th percentile (indicating the lower bound), median, and 90th percentile (indicating the upper bound) of the expected price distribution.

[0066] In summary, the electricity price prediction model is a gradient boosting regression model.

[0067] (3) Indicators

[0068] In deterministic forecasting, the accuracy and efficiency of a single-point forecast are evaluated using a set of established metrics. These include mean absolute error (MAE), root mean square error, and coefficient of determination (R 2). MAE provides a straightforward interpretation by expressing the average deviation from the true value. On the other hand, RMSE places more emphasis on significant errors and penalizes them more heavily. Finally, R 2 It helps to show how well the model's predictions agree with the actual values, and its value fluctuates between 0 and 1; values ​​closer to 1 indicate higher model accuracy. These indicators are mathematically defined as follows:

[0069]

[0070] In the formula, y i , and N are the true value, predicted value, mean value, and number of samples, respectively.

[0071] On the other hand, in order to evaluate the probability prediction, the continuous ranked probability score (CRPS) has been developed as an important metric. It is constructed by including the The predicted distribution in CRP is compared with the ground truth to capture the accuracy and sensitivity of the ensemble forecast. and the unit step function Where y is the actual value. Its mathematical representation is given by the following formula:

[0072]

[0073] S2, constructs the physical model of wind turbines and photovoltaics, inputs the forecast results of renewable energy and electricity prices into the physical model, and outputs the calculation results;

[0074] (1) Physical model of fan

[0075] When modeling wind power generation, two main aspects are often considered: height adjustment and wind turbine power curve. For the former, the difference between the wind speed measurement near the wind turbine surface and the expected hub height can be solved using the logarithmic wind profile equation. The adjustment result is expressed as:

[0076]

[0077] in, is the predicted wind speed at the turbine hub height, h wt The turbine hub is high. is the predicted near-surface wind speed, h ns is the height at which near-surface measurements are made and z0 represents the surface roughness.

[0078] After adjustment for height, the power output of the wind turbine is determined by its characteristic power curve. The power generation predicted from the wind speed at the turbine hub height is given by:

[0079]

[0080] In the formula, To predict wind power output, v in , v rat , v out They represent the cut-in speed, rated speed and cut-in speed of the turbine respectively; Indicates the rated power of the turbine.

[0081] (2) Photovoltaic physical model

[0082] Modeling the output power of a photovoltaic system is essential to understanding its performance under various lighting conditions. The solar output power can be estimated as:

[0083]

[0084] In the formula, is the solar output power; the predicted incident solar irradiance It is measured in watts per square meter (W / m 2 ) is measured in units; standardized performance indicators of photovoltaic systems, such as rated power Usually specified under standard test conditions (STC). The incident solar irradiance under these standard test conditions is denoted as G inc,STC Another important factor in this model is the reduction factor f dr , which takes into account system losses due to inefficiency, fouling, temperature and other factors.

[0085] S3, comprehensively considering the economic and evaluation indicators, constructs the time interval optimal segmentation control model based on the minimum control cost and maximum prediction accuracy of residential user equipment, and proposes an improved MOPSO algorithm to solve the multi-objective problem;

[0086] The steps of constructing a time interval optimal segmentation control model considering the minimum control cost and maximum prediction accuracy of residential user equipment include:

[0087] S31, based on the calculation result output in S2, constructing a residential user equipment model with the goal of minimizing the control cost;

[0088] The distributed resources managed by residential users set in the present invention include: distributed photovoltaic units, micro gas units, electric energy storage equipment and controllable loads of residential users. Therefore, the income of residential users is W VPP Maximization (i.e., minimizing regulation cost) can be expressed as:

[0089]

[0090] Where: They represent the market revenue of residential users, internal electricity sales revenue, micro gas turbine operating costs, electric energy storage operating costs, demand response costs and internal electricity purchase costs. The calculation formulas for each variable are as follows:

[0091] 1) Market returns

[0092]

[0093] Where: represents the market electricity price during period t, represents the amount of electricity bid by residential users in the market during period t, When purchasing electricity on the market on behalf of residential users, The opposite is true.

[0094] 2) Micro gas turbine operating costs

[0095] Since the micro gas turbines managed by residential users are generally small in size, it is assumed that they are online all the time and their start-up and shutdown costs are not considered. Their operating costs can be measured by a linear function:

[0096]

[0097] Where: N J represents the number of micro gas turbines managed by residential users, a j , b j is the cost coefficient of the jth micro gas turbine, is the power generation of the jth micro gas turbine in period t.

[0098] 5) Electric energy storage operating costs

[0099] The operating cost of electric energy storage is the loss cost of the battery due to charging and discharging, which is directly related to the charging and discharging amount and can be approximately measured by a linear function:

[0100]

[0101] Where: N e The amount of electric energy storage managed for residential users, c i is the depreciation cost coefficient of the i-th electric energy storage, and are the charging power and discharging power of the i-th energy storage in period t, both of which are positive values. In order to minimize the total operating cost, the final optimization result is and One of the items must be 0.

[0102] 6) Internal management benefits

[0103] Internal management benefits include internal electricity sales revenue, demand response compensation costs and electricity purchase costs.

[0104]

[0105] Where: and The internal electricity purchase and sales prices set for residential users in period t, is the initial residential user load in period t, is the demand response compensation unit price during period t, is the demand response amount during period t; It is the contractual transaction electricity between residential users and internal power generation resources during period t.

[0106] The constraints are as follows:

[0107] 1) Power balance constraints

[0108]

[0109] 2) Output constraints of renewable energy units

[0110]

[0111] Where: P pv,max It is the maximum installed capacity of distributed photovoltaic units.

[0112] 3) Gas unit output constraints

[0113] The power response speed of the micro gas unit managed by residential users is relatively fast, which is different from the ramp constraints required for hourly scheduling. Therefore, the present invention only considers the upper and lower limit constraints of the micro gas unit.

[0114]

[0115] Where: and are the lower and upper limits of the output of the hth gas generator set respectively.

[0116] 4) Energy storage state of charge constraints

[0117] The state of charge of the energy storage device at each time period during operation must be maintained within a certain range. In addition, in order to facilitate scheduling the next day, the state of charge of the energy storage device at the end of the day should be equal to the state at the beginning of the day.

[0118]

[0119] Where: and are the upper limits of the charging and discharging power of the i-th energy storage device; E n,i,t is the surplus of the i-th energy storage device in period t; S i,t is the state of charge of the i-th energy storage device during period t; and

[0120] are the lower and upper limits of the state of charge of the i-th energy storage device respectively; is the charging and discharging efficiency factor of the i-th energy storage device; and Especially the starting and ending capacity of i energy storage devices.

[0121] 5) Demand response constraints

[0122]

[0123] Where: The electricity reduction cap is agreed upon between residential users in the market through contracts or other agreements.

[0124] 6) Market power purchase and sales constraints

[0125]

[0126] Where: P DA,max It is the upper limit of the market bidding electricity.

[0127] S32, based on the user equipment control cost model, additionally consider the prediction accuracy under different time block divisions, and build a time interval optimal segmentation control model;

[0128] like Figure 2 As shown in Figure 1, a schematic diagram of the segmented rolling window of the optimal time segmentation interval block is shown. The time range T is subdivided into planning windows of length M, Nm, and arbitrary start and end times. The dark gray and light gray blocks represent the start and end times for scheduling for this particular window, respectively. This process will continue until the complete time range is covered, and these planning windows will be called time blocks. The choice of time blocks will affect the prediction accuracy, and the corresponding indicators are shown in the aforementioned mean absolute error (MAE), root mean square error, and determination coefficient (R 2 )wait.

[0129] The optimal time interval segmentation control model is:

[0130]

[0131] The time interval optimal segmentation control model is a multi-objective model. During the solution process:

[0132] First, data is collected from meteorological measurements and historical records. This data is used to train ML models to predict relevant values, including near-surface wind speed, incident solar irradiance, and local marginal electricity prices, as well as sample curves for predicting load demand and electricity costs.

[0133] Secondly, the forecast data is processed according to the defined time span (T) and the design of the specific renewable energy generator. The prediction accuracy decay (PAD) value of each prediction variable is estimated to evaluate the accuracy of different prediction horizons, including the aforementioned mean absolute error (MAE), root mean square error and coefficient of determination (R 2 ), etc. Wind speed and solar irradiance are used to estimate the power output of renewable energy sources given the equipment characteristics of wind turbines and solar photovoltaic panels. Other specifications such as rated power and design factors are also simulated.

[0134] Third, the time horizon, PAD estimates, forecasts, and system data are loaded into a model that determines the optimal split points within the time horizon to minimize scheduling costs while maximizing accuracy.

[0135] Fourth, and finally, the Pareto optimal frontier (POF) of cost and accuracy is obtained; based on the calculation results, a non-uniform rolling horizontal window can be specified for the optimal scheduling.

[0136] In the process of solving multi-objective problems, MOPSO still has problems such as insufficient stability and insufficient solution accuracy in dealing with nonlinear multi-objective problems. At present, many literatures have improved the optimization process of MOPSO. Since the non-dominated sorting strategy of NSGA II can effectively help particles quickly find the dominant solution, and has a better inhibitory effect on the optimal value of the optimization problem being too premature and falling into the local optimum, this paper improves the MOPSO algorithm, uses the non-dominated sorting MOPSO as a framework, and incorporates the variation of NSGA II. The advantage is that in addition to inheriting the faster optimization speed of the original MOPSO, it further absorbs NSGA-II's ability to solve the problem of premature optimization process falling into the local optimum.

[0137] This paper uses the improved MOPSO algorithm to solve the optimal segmentation control model of the time interval. Figure 3 As shown, gen represents the iteration number and maxgen represents the maximum number of iterations. The model solution process includes the following steps:

[0138] Step 1: Initialization and parameter setting. Input the time block tt, PAD estimation, data prediction value and system parameters. Initialize the positions and velocities of particles, which represent individuals in the population. Set the algorithm parameters, such as the maximum number of iterations maxgen and the initial value of the inertia weight.

[0139] Step 2: Generate the initial population. Generate the initial population WW using the input parameters. This population is the potential solution set of the problem.

[0140] Step 3: Calculate fitness and non-dominated sorting. Calculate the objective function value for each particle in the population QQ. Perform non-dominated sorting on the particles according to their fitness values ​​to determine their superiority. Stratify the non-inferior solution set to maintain population diversity.

[0141] Step 4: Update particle position and velocity. Select the optimal individual and global solutions for the particles. Use the improved MOPSO algorithm to update the velocity and position of the particles, which may include linearly decreasing inertia weights and mutation operations.

[0142] Step 5: Iteration process. Iterate and repeat steps 3 and 4 in each generation. At the end of each generation, update the set of non-inferior solutions to form a new generation of population PP.

[0143] Step 6: Check the termination condition. Check whether the current iteration number gen reaches the maximum iteration number maxgen. If not, continue iterating; if reached, end the iteration.

[0144] Step 7: Output the optimal solution set. Calculate the objective function value of each particle and perform non-dominated sorting based on the fitness value. Layer the non-inferior solution set and finally output the optimal Pareto solution set.

[0145] S4, importing the predicted data into the time interval optimal segmentation control model, using the improved MOPSO algorithm to solve the model, and obtaining the residential user equipment control result considering the optimal segmentation of the time interval;

[0146] Example 2

[0147] In this embodiment, a specific case is provided to illustrate the effect of the control method of the present invention. In this case, a residential user device cluster consisting of a wind turbine, a solar photovoltaic panel, energy storage and a grid connection is studied, and the relevant device parameters are shown in Table 1. Figure 4 and Figure 5 Displays probabilistic forecast data for renewable energy sources wind and solar.

[0148] Table 1 Equipment parameters

[0149]

[0150] like Figure 6As shown, during the first few hours of each day, there is a wind forecast but no solar; this period coincides with a very short time period. Then, once the solar forecast increases, fewer time blocks are used in favor of longer execution time horizons. It can be seen that there is a direct correlation between solar generation and longer forecast periods. Since wind speed has more randomness than solar radiation, the PAD associated with wind power increases faster over the time horizon than the PAD associated with solar power. Since the error in solar generation is small and varies less across the time horizon, the regulation strategy is able to exploit the loss in accuracy to obtain lower dispatch costs. This suggests that microgrids dominated by wind generation require more frequent forecasts to maintain reasonable accuracy, but at the expense of not dispatching for longer periods (which is cheaper). On the other hand, solar systems allow for longer and cheaper schedules, but may require additional installation costs.

[0151] The method of the present invention may be implemented in hardware, firmware, or as software or computer code that may be stored in a recording medium (such as a CDROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code that is originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded over a network and will be stored in a local recording medium, so that the method described herein may be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by a computer, processor, or hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method shown herein.

[0152] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A residential user equipment control method considering optimal time interval segmentation, characterized in that: The following steps are involved: Establish prediction models and evaluation indicators for renewable energy parameters and electricity prices; Construct wind turbine and photovoltaic physical models, input renewable energy parameters into the physical models, and output calculation results; Based on the calculation results, a residential user equipment model is constructed with the goal of minimizing the control cost, and an optimal time interval segmentation control model is constructed by considering the evaluation index; The predicted data is imported into the time interval optimal segmentation control model, and the improved MOPSO algorithm is used to solve the model to obtain the residential user equipment control results considering the optimal segmentation of time intervals.

2. The residential user equipment control method considering optimal time interval segmentation according to claim 1, characterized in that: The renewable energy prediction model is an LSTM network model; the electricity price prediction model is a gradient boosting regression model.

3. The residential user equipment control method considering optimal time interval segmentation according to claim 1 or 2, characterized in that: The evaluation indicators of the prediction model of renewable energy and electricity price include: mean absolute error, root mean square error, determination coefficient and continuous sorting probability score.

4. The residential user equipment control method considering optimal time interval segmentation according to claim 1, characterized in that: The physical model of the fan is: in, To predict wind power output, v in , v rat , v out They represent the cut-in speed, rated speed and cut-in speed of the turbine respectively; Indicates the rated power of the turbine; is the predicted wind speed at the turbine hub height, h wt The turbine hub is high. is the predicted near-surface wind speed, h ns is the height at which near-surface measurements are made and z0 represents the surface roughness.

5. The residential user equipment control method considering optimal time interval segmentation according to claim 4, characterized in that: The photovoltaic physical model is: In the formula, Output power for solar energy; is the predicted incident solar irradiance; is the rated power, G inc,STC is the incident solar irradiance, f dr is the conversion factor.

6. The residential user equipment control method considering optimal time interval segmentation according to claim 5, characterized in that: The residential user equipment model constructed with the goal of minimizing the regulation cost is: Where W VPP For the benefit of residential users, They represent the market revenue of residential users, internal electricity sales revenue, micro gas turbine operating cost, electric energy storage operating cost, demand response cost and internal electricity purchase cost respectively; represents the market electricity price during period t, P t DA represents the amount of electricity bid by residential users in the market during period t, P t DA >0 means that residential users purchase electricity in the market, P t DA When <0, the opposite is true; N J represents the number of micro gas turbines managed by residential users, a j , b j is the cost coefficient of the hth micro gas turbine, is the power generation of the jth micro gas turbine in period t; N e The amount of electric energy storage managed for residential users, c i is the depreciation cost coefficient of the i-th electric energy storage, and are the charging power and discharging power of the i-th energy storage in period t, both of which are positive values; and The internal electricity purchase and sales prices set for residential users in period t, is the initial residential user load during period t, is the demand response compensation unit price during period t, P t dr is the demand response amount during period t; P t buyins It is the contractual transaction electricity between residential users and internal power generation resources during period t.

7. The residential user equipment control method considering optimal time interval segmentation according to claim 6, characterized in that: The improved MOPSO algorithm uses the non-dominated sorting MOPSO as a framework and incorporates the variation of NSGAII. The steps of using the improved MOPSO algorithm to solve the time interval optimal segmentation control model include: Step 1: Input the time block t, PAD estimation, data prediction value and system parameters; initialize the position and velocity of particles, which represent individuals in the population; set the parameters of the algorithm, including the maximum number of iterations maxgen and the initial value of the inertia weight; Step 2: Generate the initial population W using the input parameters. This population is the potential solution set of the problem. Step 3: Calculate the objective function value for each particle in the population Q, perform non-dominated sorting on the particles according to their fitness values ​​to determine their superiority and inferiority, and stratify the non-inferior solution set to maintain population diversity; Step 4: Select the optimal individual and global solutions of the particles, and use the improved MOPSO algorithm to update the velocity and position of the particles; Step 5: Iterate and repeat steps 3 and 4 in each generation. At the end of each generation, update the set of non-inferior solutions to form a new generation of population P; Step 6: Check whether the current number of iterations gen has reached the maximum number of iterations maxgen. If not, continue iterating; if reached, end the iteration; Step 7: Calculate the objective function value of each particle, perform non-dominated sorting according to the fitness value, stratify the non-inferior solution set, and finally output the optimal Pareto solution set.

8. A residential user equipment control system considering optimal time interval segmentation, characterized in that: include: Prediction model building module: Establish prediction models and evaluation indicators for renewable energy parameters and electricity prices; Physical model building module: builds wind turbine and photovoltaic physical models, inputs renewable energy parameters into the physical models, and outputs calculation results; A control model building module: based on the calculation results, a residential user equipment model is built with the goal of minimizing the control cost, and a time interval optimal segmentation control model is built considering the evaluation index; And, the result output module: imports the predicted data into the time interval optimal segmentation control model, and uses the improved MOPSO algorithm to solve the model to obtain the residential user equipment control result considering the optimal segmentation of the time interval.

9. A computer storage medium storing a readable program, characterized in that: When the program is running, it can execute the residential user equipment control method considering the optimal division of time intervals as described in any one of claims 1-7.

10. An electronic device, characterized in that: include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the residential user equipment control method considering optimal time interval segmentation as described in any one of claims 1-7.

Images