Day-ahead interactive adjustment method and system for residential user load considering uncertain risk

Abstract

The application discloses a kind of resident user load day-to-day interaction adjustment method and system considering uncertain risk, belong to smart grid technical field.The method includes: the uncertain risk caused by energy use behavior and power grid regulation demand is established probability distribution model;Utilize Latin hypercube sampling to sample and generate uncertain scene to probability distribution model;Two-stage stochastic optimization model considering the interaction of resident user load and power grid is constructed;Two-stage stochastic optimization model is solved using optimization solver, and resident user load regulation strategy under different scenes is obtained.The application improves the friendliness of power grid and user interaction, enhances the economic efficiency, stability and robustness of regulation resident user load strategy, reduces uncertain risk loss.

Description

Technical Field

[0001] This invention belongs to the field of smart grid technology, specifically relating to a day-ahead interactive regulation method and system for residential user load that takes into account uncertain risks. Background Technology

[0002] As user electricity consumption continues to increase, the use of flexible resources such as air conditioners, electric water heaters, and electric vehicles is becoming increasingly complex. Renewable energy power generation is characterized by randomness and volatility, which introduces a series of uncertain risks to the safe and stable operation of the power system. However, traditional residential load regulation methods are typically based on deterministic day-ahead forecasts, relying on single load forecasts or user behavior to model and regulate loads. This makes it difficult to effectively address the complex uncertainties arising from user energy consumption behavior and grid regulation needs. These uncertainties not only increase the difficulty of grid dispatch but may also lead to higher electricity purchase costs for users, reduced comfort, and even impact on grid operating efficiency. Summary of the Invention

[0003] The purpose of this invention is to provide a day-ahead interactive adjustment method for residential user load that takes into account uncertain risks, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a day-ahead interactive adjustment method for residential user load considering uncertain risks, comprising:

[0005] Step 1: Establish a probability distribution model for the uncertain risks arising from energy consumption behavior and grid regulation demand;

[0006] Step 2: Use Latin hypercube sampling to sample the probability distribution model and generate uncertain scenarios;

[0007] Step 3: Construct a two-stage stochastic optimization model that considers the interaction between residential user load and the power grid, including:

[0008] Phase 1: Before the occurrence of current uncertainties, determine adjustment decisions that are applicable to all scenarios;

[0009] The second stage: After the actual scenario is determined the next day, based on the adjustment decision in the first stage, adaptive adjustments are made according to the optimization objective function that takes into account user economy, comfort and power grid adjustment needs, in order to adapt to the uncertainty of the current scenario.

[0010] The two-stage stochastic optimization model introduces downside risk constraints to quantify and control downside risk caused by uncertainties. By quantifying downside risk and limiting the overall risk level through downside risk constraints, it achieves risk management of uncertainty.

[0011] Step 4: Use the optimization solver to solve the two-stage stochastic optimization model to obtain the load adjustment strategies for residential users under different scenarios.

[0012] Further optimization,

[0013] In step three, based on the overall interaction between users and the power grid, and considering user economy, comfort, and power grid regulation needs, a comprehensive optimization objective function is constructed:

[0014] ;

[0015] In the formula, To minimize the overall cost; , These represent the weighting coefficients for the user and the power grid, respectively. The user's expected electricity purchase cost; Costs related to user discomfort; Cost of demand-side adjustment deviation in the power grid;

[0016] ;

[0017] In the formula, For the scene The probability of occurrence; For time period C m Electricity purchase price in the scenario; For time period C m Power grid power in the scenario; Where m is the time step, m is the scene index, and M is the number of scenes;

[0018] ;

[0019] In the formula, For the user's expected indoor temperature, For time period C m The actual indoor temperature in the scenario; Preset the minimum water temperature for users; For time period C m The actual water temperature in the scenario; and These are the penalty coefficients for indoor temperature difference and water temperature difference, respectively. For indoor temperature difference, This refers to the water temperature difference.

[0020] ;

[0021] In the formula, For time period C m Total load power of all users in the scenario; For time period C mThe expected regulation power demand of the power grid in the scenario.

[0022] Further optimization, the downside risk constraint is expressed as follows:

[0023] ;

[0024] ;

[0025] In the formula: C0 represents the expected overall downside risk for each scenario, and C0 represents the expected cost. For the scene Downside risk; γ is a risk control parameter, with a value range of [0,1]. When it is 0, it is a risk aversion strategy; when it is 1, it is a risk neutral strategy; when it is between 0 and 1, it is a risk trade-off strategy; C0 is the expected cost; For the scene Operating costs without considering risks.

[0026] Further optimization reveals that the constraints of the two-stage stochastic optimization model include physical law constraints, power balance constraints, power grid security constraints, electric vehicle charging constraints, air conditioner operation constraints, and electric water heater operation constraints; among which, the physical law constraints are expressed as follows:

[0027] ;

[0028] In the formula, Indicates time period C m State of charge in the scenario; Indicates the charging efficiency of electric vehicles; This represents the charging power of the electric vehicle during time period t; Indicates time period C m Air conditioning operating power in the scenario; Indicates time period C m Outdoor temperature in the scene; Indicates time period C m Operating power of electric water heater in the following scenario; For electric vehicle battery capacity; and These are the building's thermal resistance and heat capacity, respectively. To improve air conditioning operating efficiency; This is the heat loss coefficient; To improve the operating efficiency of electric water heaters; For time period C m The capacity of the electric water heater used in the given scenario; This refers to the heat capacity of the water tank.

[0029] Specifically, in order to analyze energy consumption behavior and grid regulation demand, the preset usage status and data of each load are collected through user-side and grid-side systems within a certain time window before the day, and the collected data is preprocessed.

[0030] Specifically, step one includes:

[0031] To address the uncertainty of user behavior, a probability distribution model is established using a time-varying Markov chain.

[0032] For the uncertainty of environmental factors, a probability distribution model is established using the normal distribution;

[0033] Uncertain events that may occur during electric vehicle charging:

[0034] For the insertion time at the start of charging, a probability distribution model is constructed using a normal distribution;

[0035] For the initial state of charge and the target state of charge, a probability distribution model is established using the Beta distribution;

[0036] The load regulation demand on the demand side of the power grid is modeled using a normal distribution model to obtain a probability distribution model;

[0037] The prediction error of renewable energy output is modeled using a Gaussian mixture model to comprehensively model the probabilities of different scenarios, resulting in a probability distribution model. The Gaussian mixture model constructs multiple sub-Gaussian models to fit the distribution of renewable energy output prediction errors under different operating conditions. The overall probability is then calculated based on the weights of each sub-Gaussian model to reflect the uncertainty of renewable energy output prediction errors.

[0038] Specifically, in step two, based on the probability distribution model of uncertain risks, Latin hypercube sampling (LHS) is used to generate multiple energy consumption behavior and grid regulation demand scenarios. Sampling is performed for each time period within a day, and after scenario reduction technology, the scenarios are spliced ​​together into a whole uncertain scenario for a single day.

[0039] Specifically, the process of step two is as follows:

[0040] Calculate the cumulative probability distribution function F(P) for all variables, and then divide the scenario into N scenarios;

[0041] Latin hypercube sampling is performed within each interval. For the i-th interval, i=1,2,...,N, a random number r in the range [0,1] is randomly generated. i Then the cumulative probability function value q corresponding to the i-th interval i for:

[0042] ;

[0043] The inverse function of F(P) is obtained as F -1 (q), q i Substituting the values ​​into the following formula yields the sample value P for an uncertain scenario over a single time period. i :

[0044] ;

[0045] By reducing the number of individual time periods, the 24 individual time periods are spliced ​​together to obtain the uncertain scenario of the entire day-ahead residential user load.

[0046] Further optimization involves reducing the number of scenarios within a single time period, as follows;

[0047] The K-means clustering algorithm was used for reduction. A scenario set includes the electric vehicle plug-in time, the state of charge at the start and expected end of charging, the air conditioner set temperature and time, the electric water heater temperature and water volume, the grid regulation demand, and the renewable energy output prediction error. A total of N scenarios were divided, and the number of target scenarios after reduction is M, which are used as the initial cluster centers.

[0048] For each scenario S i Its probability of occurrence is P i Calculate the Euclidean distance between it and the M cluster centers, and assign it to the class C to which the nearest cluster center belongs. k ;

[0049] A new cluster center scenario is obtained by weighted averaging across all scenarios. :

[0050] ;

[0051] Repeatedly calculate the Euclidean distance between each scene and the current cluster center and assign it to the nearest category. Update the cluster center for each scene in each category using a probability-weighted average until the cluster center no longer changes, thus obtaining the final set of M scenes {C1, C2, ..., C} for time period t. M}, C M This is the Mth scene.

[0052] The present invention also provides a day-ahead interactive adjustment system for residential user load considering uncertain risks, including a memory and a processor. The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor implements the day-ahead interactive adjustment method for residential user load considering uncertain risks.

[0053] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0054] By establishing a differentiated probability distribution model based on user energy consumption behavior (such as temperature-controlled load use and electric vehicle charging), environmental factors, grid regulation demand, and renewable energy output prediction errors, and combining Latin hypercube sampling and K-means clustering to generate scenarios, this model effectively covers various uncertain risks, solves the problem that traditional methods rely on deterministic predictions and are difficult to deal with complex uncertainties, and reduces the difficulty of grid dispatching.

[0055] The two-stage stochastic optimization model takes into account both day-ahead general decision-making and next-day adaptive adjustment. The objective function integrates user electricity purchase cost, comfort cost, and grid regulation deviation cost. Furthermore, it flexibly adapts to demand through weight coefficients, ensuring user energy comfort and reducing electricity purchase cost while meeting grid regulation needs and improving interactivity.

[0056] By introducing downside risk constraints, risk avoidance, neutrality, or trade-off strategies can be implemented through the risk control parameter γ to prevent excessive deviation between actual and expected costs. The resulting adjustment strategies are adaptable to different scenarios, enhancing the stability and robustness of power system operation and reducing economic losses and efficiency degradation caused by uncertain risks. Attached Figure Description

[0057] Figure 1 Flowchart of a day-ahead interactive adjustment method for residential user load considering uncertain risks;

[0058] Figure 2 A flowchart for randomly generating uncertain scenarios using the Latin hypercube sampling method. Detailed Implementation

[0059] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. These embodiments are exemplary and intended to explain the present invention, and should not be construed as limiting the invention.

[0060] like Figure 1 As shown, a day-ahead interactive adjustment method for residential user load considering uncertain risks includes:

[0061] Step 1: Establish a probability distribution model for the uncertain risks arising from energy consumption behavior and grid regulation demand;

[0062] Step 2: Use Latin hypercube sampling to sample the probability distribution model and generate uncertain scenarios;

[0063] Step 3: Construct a two-stage stochastic optimization model that considers the interaction between residential user load and power grid;

[0064] Step 4: Use the optimization solver to solve the two-stage stochastic optimization model to obtain the load adjustment strategies for residential users under different scenarios.

[0065] To support day-ahead interactive regulation of residential user loads, it is necessary to predict user-side energy consumption behavior and grid-side regulation needs in advance. Within a certain time window before the day, the preset usage and specific data of each load should be collected through user-side and grid-side systems, and the collected data should be preprocessed.

[0066] The specific load data that needs to be collected on the user side is as follows:

[0067] 1) The air conditioner's operating time and set temperature, as well as the actual indoor temperature and air conditioner's operating power;

[0068] 2) The water usage time and volume of the electric water heater, the temperature change of the water tank, and the heating power;

[0069] 3) The daily start and end times of charging for electric vehicles, the initial state of charge (SOC) and the state of charge (SOC) at the end of charging, and the actual charging power.

[0070] The specific load data that needs to be collected on the grid side is as follows:

[0071] 1) Renewable energy output: Due to the deviation between the actual output of photovoltaic power generation and wind power generation and the forecast, the grid's response to the demand side changes.

[0072] 2) Fluctuations in real-time electricity prices in the electricity market may affect the incentive mechanisms for load regulation by the power grid;

[0073] 3) Sudden equipment failures, severe weather, and other emergencies can also become uncertain risks that affect the adjustment needs of residential users' load.

[0074] The collected data was preprocessed as follows:

[0075] 1) Handling missing data values. Linear interpolation can be used to fill in missing values.

[0076] 2) Handling outliers. Outliers in the data can be identified using the Z-score method, and then deleted or replaced.

[0077] 3) Normalization of multi-source interactive data. The Min-Max normalization method is used to ensure that data of different dimensions are represented by values ​​in the range [0,1].

[0078] Changes in user energy consumption behavior and grid demand may lead to some uncertain risks. Therefore, a probability distribution model is established for the uncertain risks brought about by energy consumption behavior and grid regulation demand.

[0079] (1) For temperature-controlled loads such as air conditioners and electric water heaters, the main uncertainties in load changes come from user behavior and environmental changes.

[0080] 1) To address the uncertainty of user behavior, i.e., their in-home state, a time-varying Markov chain can be used to model it and establish a probability distribution model. Using one hour as the time step and a total of 24 hours, two states are defined: user at home (S1) and user not at home (S2). The state transition probability matrix is ​​as follows:

[0081] ;

[0082] In the formula, Let be the state transition matrix at time t. Let S be the probability that the state remains unchanged from S1. Let S be the probability that the state remains unchanged from S2. Let S be the probability of the state changing from S1 to S2. Let S be the probability of the state changing from S2 to S1.

[0083] Define the initial state distribution P(0). The transition probabilities and the initial state distribution P(0) for each time period can be calculated from historical data. Then, by random sampling, the state probability for each subsequent time period is calculated, and finally, a 24-hour state sequence X(t) is generated.

[0084] 2) The uncertainty of environmental factors can be described using a normal distribution, and a normal distribution model can be established. As a probability distribution model:

[0085] ;

[0086] In the formula, This represents the probability density at temperature T, where T represents the actual ambient temperature or the water tank temperature. To predict the temperature to be reached, This represents the temperature standard deviation.

[0087] (2) Uncertain events that may occur during electric vehicle charging.

[0088] 1) The insertion time at the start of charging The normal distribution can be used to construct a probability distribution model. . Average charging start time for users, This represents the standard deviation of the charging start time.

[0089] 2) Regarding the initial state of charge during charging and target state of charge A probability distribution model can be built using the Beta distribution:

[0090] ;

[0091] ;

[0092] In the formula, Represents the initial state of charge or target state of charge , For the collected state of charge, For the Gamma function, It follows a Beta distribution. and These are two parameters of the Beta distribution. By using the Beta distribution and combining them with actual observation data, the parameters can be adjusted to obtain a more accurate predicted probability.

[0093] (3) Load regulation demand on the demand side of the power grid This refers to the expected load change of the power grid over a certain period of time, which can also be obtained by modeling it using a normal distribution model. , This represents the average load adjustment demand. The standard deviation of load regulation demand is used as a probability distribution model.

[0094] (4) Renewable energy output forecasting error Gaussian mixture models (GMMs) can be used to comprehensively model the probabilities of different scenarios and obtain a probability distribution model. In this model, multiple sub-Gaussian models are constructed to fit the distribution of renewable energy output prediction errors under different operating conditions. The overall probability is then calculated based on the weights of each sub-Gaussian model to reflect the uncertainty of renewable energy output prediction errors.

[0095] ;

[0096] In the formula, Indicates in the parameter Prediction error of renewable energy output The probability density function, The probability that the observed data belongs to the k-th sub-Gaussian model; Let be the Gaussian distribution function of the k-th sub-Gaussian model. , Let be the mean of the k-th sub-Gaussian model. Let be the standard deviation of the k-th sub-Gaussian model; a total of K sub-Gaussian models are fitted.

[0097] Based on a probability distribution model of uncertain risks, Latin hypercube sampling (LHS) is used to generate multiple energy consumption behavior and grid regulation demand scenarios. Sampling is performed for each time period within a day, and after scenario reduction techniques, the scenarios are stitched together to form a complete single-day uncertain scenario, such as... Figure 2As shown, it includes:

[0098] (1) Calculate the cumulative probability distribution function F(P) for all variables, and then divide each time period of the day into N scenarios to accurately improve the calculation efficiency and distribution coverage.

[0099] (2) Perform Latin hypercube sampling within each scene. For the i-th scene (i=1,2,...,N), randomly generate a random number r in the range [0,1]. i Then the cumulative probability function value q corresponding to the i-th scenario i for:

[0100] ;

[0101] The inverse function of F(P) is obtained as F -1 (q), q i Substituting the values ​​into the following formula yields the sample value P for an uncertain scenario over a single time period. i :

[0102] ;

[0103] (3) Reduce the number of scenes in a single time period.

[0104] 1) K-means clustering algorithm can be used for scenario reduction. A scenario set includes electric vehicle plug-in time, state of charge at the start and expected end of charging, air conditioner set temperature and time, electric water heater temperature and water volume, grid regulation demand, and renewable energy output prediction error. A total of N scenarios are divided. The number of target scenarios after reduction is M, which serve as the initial cluster centers.

[0105] 2) For each scenario S i Its probability of occurrence is P i Calculate the Euclidean distance between it and the M cluster centers, and assign it to the class C to which the nearest cluster center belongs. k .

[0106] 3) A new cluster center scenario is obtained by weighted averaging of all scenarios. :

[0107] ;

[0108] 4) Repeatedly calculate the Euclidean distance between each scene and the current cluster center and assign it to the nearest category. Update the cluster center for each scene in each category using a probability-weighted average until the cluster center no longer changes, thus obtaining the final set of M scenes {C1, C2, ..., C} for time period t. M}, C M This is the Mth scene.

[0109] (4) Finally, the 24 time periods are spliced ​​together to obtain the uncertain scenario of the entire daytime residential user load.

[0110] A two-stage stochastic optimization model considering the interaction between residential user load and power grid is constructed as follows:

[0111] Phase 1: Before the occurrence of day-ahead uncertainties, determine adjustment decisions that are applicable to all scenarios. These adjustment decisions are independent of specific scenarios, such as electric vehicle charging power and basic load scheduling benchmarks, and provide a basic framework for the next day's adjustments. The variables in this phase are independent of the scenario and are consistent across all scenarios.

[0112] The second stage: After the actual scenario (such as real-time plug-in time and outdoor temperature) is determined the next day, based on the adjustment decision of the first stage, adaptive adjustment is carried out according to the optimization objective function that takes into account the user's economy, comfort and power grid adjustment needs. This includes dynamic optimization of variables such as air conditioner operating power, electric water heater operating power and real-time power supply from the power grid to adapt to the uncertainty of the current scenario.

[0113] The variables at this stage will be adjusted accordingly based on the impact of uncertain risks, including:

[0114] The operating power of the air conditioner, using Indicates time period C m Air conditioning operating power in the scenario;

[0115] Electric water heater operating power, using Indicates time period C m Operating power of electric water heater in the following scenario;

[0116] Grid power, using Indicates time period C m Real-time power provided by the power grid in the scenario;

[0117] Based on the overall interaction between users and the power grid, and considering user economy, comfort, and power grid regulation needs, a comprehensive optimization objective function is constructed:

[0118] ;

[0119] In the formula, To minimize the overall cost; , These represent the weighting coefficients for the user and the power grid, respectively. The user's expected electricity purchase cost; Costs related to user discomfort; Cost of demand-side adjustment deviation in the power grid;

[0120] ;

[0121] In the formula, For the scene The probability of occurrence; t time period C m Electricity purchase price in the scenario; For time period C m Power grid power in the scenario; Where m is the time step, m is the scene index, and M is the number of scenes;

[0122] ;

[0123] In the formula, For the user's expected indoor temperature, For time period C m The actual indoor temperature in the scenario; Preset the minimum water temperature for users; For time period C m The actual water temperature in the scenario; and These are the penalty coefficients for indoor temperature difference and water temperature difference, respectively. For indoor temperature difference, The water temperature difference can be adjusted according to user preferences.

[0124] ;

[0125] In the formula, For time period C m Total load power of all users in the scenario; For time period C m The expected regulation power demand of the power grid in the scenario.

[0126] The constraints for constructing the two-stage stochastic optimization model are as follows:

[0127] (1) Constraints of physical laws

[0128] ;

[0129] ;

[0130] In the formula, Indicates time period C m State of charge in the scenario; Indicates the charging efficiency of electric vehicles; Indicates time period C m Charging power of electric vehicles in the specified scenarios; Indicates time period C m Air conditioning operating power in the scenario; Indicates time period C m Outdoor temperature in the scene; Indicates time period C m Operating power of electric water heater in the following scenario; For electric vehicle battery capacity; and These are the building's thermal resistance and heat capacity, respectively. To improve air conditioning operating efficiency; This is the heat loss coefficient; To improve the operating efficiency of electric water heaters; For time period C m The capacity of the electric water heater used in the given scenario; This refers to the heat capacity of the water tank.

[0131] (2) Power balance constraint

[0132] ;

[0133] In the formula, Indicates time period C m Real-time power provided by the power grid in the scenario; For time period C m The base load is not adjustable in this scenario; For time period C m Photovoltaic power output in the scenario.

[0134] (3) Power grid security constraints

[0135] ;

[0136] In the formula, This represents the maximum power that the power grid can operate at.

[0137] (4) Electric vehicle charging constraints

[0138] ;

[0139] In the formula, This indicates the charging on / off state; 1 means charging is on, and 0 means charging is off. For the t-time period scenario The charging power of electric vehicles. For the scene State of charge at the end of charging; For the scene The target state of charge for the next charge.

[0140] (5) Air conditioning operation constraints

[0141] ;

[0142] In the formula, and Scenes The minimum and maximum allowable temperatures for air conditioning operation. Represents the scenario during time period t The maximum power of the air conditioner when it is running. This formula represents the on / off state of the air conditioner and indicates the temperature and operating power constraints of the air conditioner.

[0143] (6) Operating constraints of electric water heaters

[0144] ;

[0145] In the formula, and Scenes The minimum and maximum allowable water temperatures for electric water heaters; Representing a scene The maximum allowable power of the electric water heater is as follows: This indicates the on / off state of the electric water heater.

[0146] Uncertainties and risks during system operation may cause deviations between the actual overall cost and the expected target of the two-stage stochastic optimization model. Downside risk constraints can be used to manage the risks of the optimization model, as follows:

[0147] The downside scenario is defined as the situation where the actual cost is higher than the expected cost. Downside risk is the difference between the actual cost and the expected cost in this downside scenario. When the actual cost is less than or equal to the expected cost, the downside risk is zero. The estimated total downside risk considering all scenarios is as follows:

[0148] ;

[0149] In the formula: The expected overall downside risk for each scenario; C0 represents the expected cost; For the scene Downside risks.

[0150] To balance cost and risk, a downside risk constraint is imposed in optimizing the two-stage stochastic optimization model:

[0151] ;

[0152] In the formula: γ is a risk control parameter, which takes values ​​in the range [0,1]. A value of 0 indicates that the existence of risk is not allowed, which is called a risk aversion strategy; a value of 1 indicates that no risk is controlled, which is called a risk neutral strategy; a value between 0 and 1 indicates that some risk is allowed, which is called a risk trade-off strategy. For the scene Operating costs without considering risks.

[0153] A two-stage stochastic optimization model was solved using a commercial optimization solver to obtain residential user load adjustment strategies under different scenarios, as follows:

[0154] (1) Configure the parameters of the solver

[0155] 1) Set the maximum iteration time for the solution, based on the adjustment time window of the current day, to ensure that the solution is completed before the current day (D-1 day);

[0156] 2) Setting a relative gap improves convergence accuracy;

[0157] 3) Selection of the number of parallel threads.

[0158] (2) Solve the two-stage stochastic optimization model and output the results.

[0159] The solver searches for the global optimum using a branch and bound algorithm. Under all constraints, it outputs: the objective function value, the day-ahead scheduling plan, and the real-time adjustment strategy for residential user load regulation.

[0160] 1) Output the first stage result, which is the daily scheduling plan that is irrelevant to the uncertain risks of the next day: obtain the charging power of electric vehicles and the charging plan for each time period of the next day, that is, the charging power curve of electric vehicles for 24 hours on the next day;

[0161] 2) Output the results of the second stage, namely, real-time adjustment of residential user load regulation strategy: for air conditioner power, electric water heater power, and electricity purchase power under many uncertain scenarios;

[0162] Parallel computing can improve solution efficiency by allocating the computation of the second-stage scenario to multi-core processors. The scenario decomposition formula is as follows:

[0163] ;

[0164] In the formula, Z is the objective function value, and x and y are the variables in the first stage and the second stage, respectively. For the scene The objective function value under the given scenario reflects the optimization effect based on the first and second stage variables in this uncertain scenario.

[0165] 3) Output the objective function value, and use the solver to obtain the expected total cost (user electricity cost, discomfort cost, grid deviation adjustment cost) and the specific cost under each scenario, and finally obtain the optimal function output value.

[0166] The convergence condition for this solver is: the maximum iteration time is reached or the accuracy of the objective function is lower than the set value ε.

[0167] The optimization results are transformed into user-friendly scheduling plans and pushed to the app, including: electric vehicle charging plans: charging time and power; air conditioning operation plans: allowed operating times and temperature ranges; and electric water heater plans: heating times and water temperature in the electric water heater tank.

[0168] Through IoT protocols, scheduling plans that take into account uncertainties are distributed to smart devices to ensure execution and support real-time adjustments to cope with uncertainties. Residential user load regulation strategies. The format of the document is as follows:

[0169] ;

[0170] Based on the uncertain risks brought about by uncertain scenarios, users can adjust load usage plans, change the original parameters of the model, and then re-establish constraints and solve the two-stage optimization model after feeding back to the power grid system. This allows for further adjustments to the optimization objective, enabling the power grid and users to interact and minimize overall costs.

[0171] Another embodiment of the present invention provides a day-ahead interactive adjustment system for residential user load considering uncertain risks, including a memory and a processor. The memory stores computer-readable instructions, which, when executed by the processor, cause the processor to implement the above-described day-ahead interactive adjustment method for residential user load considering uncertain risks.

[0172] The above embodiments are only used to illustrate the effects of the present invention, and the described embodiments are only some embodiments of this application, not all embodiments. Finally, it should be noted that all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

Claims

1. A day-ahead interactive regulation method for residential user load considering uncertain risks, characterized in that, include: Step 1: Establish a probability distribution model for the uncertain risks arising from energy consumption behavior and grid regulation demand; Step 2: Use Latin hypercube sampling to sample the probability distribution model and generate uncertain scenarios; Step 3: Construct a two-stage stochastic optimization model that considers the interaction between residential user load and the power grid, including: Phase 1: Before the occurrence of current uncertainties, determine adjustment decisions that are applicable to all scenarios; The second stage: After the actual scenario is determined the next day, based on the adjustment decision in the first stage, adaptive adjustments are made according to the optimization objective function that takes into account user economy, comfort and power grid adjustment needs, in order to adapt to the uncertainty of the current scenario. The two-stage stochastic optimization model introduces downside risk constraints to quantify and control downside risk caused by uncertainties. By quantifying downside risk and limiting the overall risk level through downside risk constraints, it achieves risk management of uncertainty. Based on the overall interaction between users and the power grid, and considering user economy, comfort, and power grid regulation needs, a comprehensive optimization objective function is constructed: ； In the formula, To minimize the overall cost; , These represent the weighting coefficients for the user and the power grid, respectively. The user's expected electricity purchase cost; Costs related to user discomfort; Cost of demand-side adjustment deviation in the power grid; ； In the formula, For the scene The probability of occurrence; For time period C m Electricity purchase price in the scenario; For time period C m Power grid power in the scenario; Where m is the time step, m is the scene index, and M is the number of scenes; ； In the formula, For the user's expected indoor temperature, For time period C m The actual indoor temperature in the scenario; Preset the minimum water temperature for users; For time period C m The actual water temperature in the scenario; and These are the penalty coefficients for indoor temperature difference and water temperature difference, respectively. For indoor temperature difference, This refers to the water temperature difference. ； In the formula, For time period C m Total load power of all users in the scenario; For time period C m The expected regulation power demand of the power grid in the scenario; Step 4: Use the optimization solver to solve the two-stage stochastic optimization model to obtain the load adjustment strategies for residential users under different scenarios.

2. The day-ahead interactive adjustment method for residential user load according to claim 1, characterized in that, The downside risk constraint is expressed as follows: ； ； In the formula: C0 represents the expected overall downside risk for each scenario, and C0 represents the expected cost. For the scene Downside risk; γ is a risk control parameter, with a value range of [0,1]. When it is 0, it is a risk aversion strategy; when it is 1, it is a risk neutral strategy; when it is between 0 and 1, it is a risk trade-off strategy; C0 is the expected cost; For the scene Operating costs without considering risks.

3. The day-ahead interactive adjustment method for residential user load according to claim 1, characterized in that, The constraints of the two-stage stochastic optimization model include physical law constraints, power balance constraints, power grid security constraints, electric vehicle charging constraints, air conditioner operation constraints, and electric water heater operation constraints; among which, the physical law constraints are expressed as follows: ； In the formula, Indicates time period C m State of charge in the scenario; Indicates the charging efficiency of electric vehicles; This represents the charging power of the electric vehicle during time period t; Indicates time period C m Air conditioning operating power in the scenario; Indicates time period C m Outdoor temperature in the scene; Indicates time period C m The operating power of the electric water heater in the given scenario; For electric vehicle battery capacity; and These are the building's thermal resistance and heat capacity, respectively. To improve air conditioning operating efficiency; This is the heat loss coefficient; To improve the operating efficiency of electric water heaters; For time period C m The capacity of the electric water heater used in the given scenario; This refers to the heat capacity of the water tank.

4. The day-ahead interactive adjustment method for residential user load according to claim 1, characterized in that, In order to analyze energy consumption behavior and grid regulation demand, the preset usage status and data of each load are collected through user-side and grid-side systems within a certain time window before the day, and the collected data is preprocessed.

5. The day-ahead interactive regulation method for residential user load according to claim 1, characterized in that, Step one includes: To address the uncertainty of user behavior, a probability distribution model is established using a time-varying Markov chain. For the uncertainty of environmental factors, a probability distribution model is established using the normal distribution; Uncertain events that may occur during electric vehicle charging: For the insertion time at the start of charging, a probability distribution model is constructed using a normal distribution; For the initial state of charge and the target state of charge, a probability distribution model is established using the Beta distribution; The load regulation demand on the demand side of the power grid is modeled using a normal distribution model to obtain a probability distribution model; The prediction error of renewable energy output is modeled using a Gaussian mixture model to comprehensively model the probabilities of different scenarios, resulting in a probability distribution model. The Gaussian mixture model constructs multiple sub-Gaussian models to fit the distribution of renewable energy output prediction errors under different operating conditions. The overall probability is then calculated based on the weights of each sub-Gaussian model to reflect the uncertainty of renewable energy output prediction errors.

6. The day-ahead interactive regulation method for residential user load according to claim 1, characterized in that, In step two, based on the probability distribution model of uncertain risks, Latin hypercube sampling is used to generate multiple energy consumption behavior and grid regulation demand scenarios. Sampling is performed for each time period within a day, and after scenario reduction technology, the scenarios are spliced ​​together to form an overall uncertain scenario for a single day.

7. The day-ahead interactive regulation method for residential user load according to claim 6, characterized in that, Step two is as follows: Calculate the cumulative probability distribution function F(P) for all variables, and then divide the scenario into N scenarios; Latin hypercube sampling is performed within each interval. For the i-th interval, i=1,2,...,N, a random number r in the range [0,1] is randomly generated. i Then the cumulative probability function value q corresponding to the i-th interval i for: ； The inverse function of F(P) is obtained as F -1 (q), q i Substituting the values ​​into the following formula yields the sample value P for an uncertain scenario over a single time period. i : ； By reducing the number of individual time periods, the 24 individual time periods are spliced ​​together to obtain the uncertain scenario of the entire day-ahead residential user load.

8. The day-ahead interactive adjustment method for residential user load according to claim 7, characterized in that, The process of reducing scenes within a single time period is as follows; The K-means clustering algorithm was used for reduction. A scenario set includes the electric vehicle plug-in time, the state of charge at the start and expected end of charging, the air conditioner set temperature and time, the electric water heater temperature and water volume, the grid regulation demand, and the renewable energy output prediction error. A total of N scenarios were divided, and the number of target scenarios after reduction is M, which are used as the initial cluster centers. For each scenario S i Its probability of occurrence is P i Calculate the Euclidean distance between it and the M cluster centers, and assign it to the class C to which the nearest cluster center belongs. k ; A new cluster center scenario is obtained by weighted averaging across all scenarios. : ； Repeatedly calculate the Euclidean distance between each scene and the current cluster center and assign it to the nearest category. Update the cluster center for each scene in each category using a probability-weighted average until the cluster center no longer changes, thus obtaining the final set of M scenes {C1, C2, ..., C} for time period t. M }, C M This is the Mth scene.

9. A residential user load day-ahead interactive regulation system considering uncertain risks, comprising a memory and a processor, wherein the memory stores computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, the processor implements the day-ahead interactive adjustment method for residential user load considering uncertain risks as described in any one of claims 1-8.

Images