Power distribution network large industrial load regulation and control method considering rebound effect

Abstract

The invention discloses a power distribution network large industrial load regulation and control method considering a bounce effect, and the method comprises the steps: recognizing and quantifying the internal relation of the bounce effect in an industrial load response through building a bounce effect model, building a feasible operation domain description model to obtain a constraint range of each physical quantity of the large industrial load bounce of a power distribution network, building a response characteristic model, and carrying out the recognition of the bounce effect. The response characteristic model comprises a day-ahead scheduling model, an intra-day rolling scheduling model, a real-time scheduling model and a self-adaptive regulation and control model, a power generation and scheduling plan of the second day is made through industrial load requirements, rebound effects, industrial equipment requirements and power grid constraints predicted by the day-ahead scheduling model, and the intra-day rolling scheduling model is updated according to real-time data. A power generation plan is adjusted and optimized, a real-time scheduling model monitors and controls a power grid in real time at the moment of operation of a power system, ordered peak clipping and multi-time-scale self-evolution scheduling of an industrial load system are achieved through a self-adaptive regulation and control model, and self-evolution updating of the scheduling model is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial load regulation, and specifically relates to a method for regulating large industrial loads in a distribution network considering the rebound effect. Background Technique

[0002] The participation of large industrial loads in the demand response of a new distribution network may trigger a rebound effect, resulting in load imbalance in the distribution network. With the continuous change of the power system and the dynamic fluctuation of loads, ensuring the self-adaptability of the scheduling model and adjusting the efficient participation of industrial loads in real time can not only relieve the supply-demand pressure of the power grid but also have high economic value.

[0003] The complexity of the rebound effect lies in that it is usually driven by multiple factors, such as load type, user behavior, and economic environment, etc., which makes it difficult to accurately model. At the same time, the existence of the rebound effect often leads to the actual energy efficiency improvement being lower than expected. Therefore, a model that can cover multiple uncertainties and dynamic factors is needed to describe it.

[0004] Existing scheduling strategies lack accurate analysis of rebound effect factors and the characterization of the constraint ranges of various physical quantities of load rebound, and the scheduling strategies for large industrial load systems lack timeliness and flexibility. This application introduces an incremental learning algorithm aiming to achieve self-evolution and update of the scheduling model, so as to better adapt to the changing needs of the power system. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for regulating large industrial loads in a distribution network considering the rebound effect to solve the problems in the above background technique that existing scheduling strategies lack accurate analysis of rebound effect factors, the characterization of the constraint ranges of various physical quantities of load rebound, and the scheduling strategies for large industrial load systems lack timeliness and flexibility.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A method for regulating large industrial loads in a distribution network considering the rebound effect, characterized by including the following steps:

[0007] Step S1: Collect historical data of the industrial load system. By analyzing the historical data, establish a rebound effect model of the industrial load system. Use the rebound effect model to identify and quantify the internal relationships of rebound effect factors in the response of the industrial load system. The rebound effect model decomposes the rebound effect factors into direct effect, indirect effect, environmental impact, interaction effect, and random perturbation, and quantifies and analyzes the weights of the four factors through data.

[0008] Step S2: Establish a feasible operating region characterization model for the industrial load system based on the impact of the rebound effect. The feasible operating region characterization model characterizes the feasible operating region of the large industrial load system load operation based on a deterministic optimization method, and obtains the constraint ranges of various physical quantities of the load rebound of the industrial load system, including ramp rate constraints, load tracking constraints, load change constraints, minimum load constraints by time period, daily load total constraints, and peak shaving and valley filling constraints.

[0009] Step S3: Establish a response characteristic model for the industrial load system based on the rebound effect. The response characteristic model includes a day-ahead scheduling model, an intraday rolling scheduling model, a real-time scheduling model, and an adaptive regulation model. Based on the constraint ranges of various physical quantities of the load rebound of the industrial load system obtained in Step S2, establish an industrial load system constraint system for the response characteristic model, including system dynamic constraints, load balance constraints, rebound characteristic constraints, and process safety constraints, and clarify the boundary of the load rebound effect of the industrial load system.

[0010] Step S4: Formulate the power generation and scheduling plan for the next day based on the industrial load demand, rebound effect, industrial equipment demand, and grid constraints predicted by the day-ahead scheduling model, and optimize the power generation plan with the highest cost-benefit through an optimization algorithm while meeting the physical and operating constraints of the grid.

[0011] Based on the day-ahead scheduling model, the intraday rolling scheduling model updates according to real-time data, adjusts and optimizes the power generation plan, adopts a model predictive control (MPC) model, and through a rolling update mechanism, continuously adjusts the scheduling strategy according to the latest grid state and market information, and analyzes the different contributions of different equipment to the rebound through the rebound effect model to help optimize the equipment regulation sequence and timing.

[0012] At the moment of the operation of the power system, the real-time scheduling model monitors and controls the grid in real time according to the weights of the rebound effect factors of the industrial load system obtained in S1 to cope with grid emergencies. Grid emergencies include load changes and equipment failures. By finely adjusting the load distribution, ensure the robustness of the power system in the face of uncertainties and dynamic changes, and reduce the impact of the rebound on the grid through staged restoration or peak shifting scheduling.

[0013] Step S7: Achieve orderly peak shaving and multi-time scale self-evolving scheduling of the industrial load system through the adaptive regulation model, and realize the self-evolving update of the scheduling model. The adaptive regulation model is established based on real-time data analysis and feedback mechanism. The adaptive regulation model includes an incremental learning model, a scheduling time interval adaptive algorithm module, a rolling window length adaptive algorithm module, and a parameter dynamic update module.

[0014] Compared with the prior art, the beneficial effects of the present invention are:

[0015] This application proposes an orderly peak shaving and multi-time scale self-evolving scheduling strategy considering various industrial loads, introduces an incremental learning algorithm, aims to achieve self-evolving update of the scheduling model, so as to better adapt to the changing demands of the power system. And the established rebound effect model uses the method of Bayesian mediation analysis to identify and quantify the internal connection of the rebound effect in industrial load response. The focus is to decompose the rebound effect into direct effect, indirect effect, environmental impact, interaction effect and random perturbation. The final result is to quantitatively analyze through data which of these four factors has the greatest influence weight on the rebound effect of the research object. The feasible operating region characterization model is a method for characterizing the feasible operating region of large industrial loads based on deterministic optimization methods, considering the influence of the rebound effect, and describes how to optimize the load curve under uncertain conditions. In the process of establishing the feasible operating region characterization model, it mainly considers the load instant increase caused by the sharp increase in current when all equipment restarts after the load regulation ends in large-load industrial production and the load increase caused by each enterprise's eagerness to resume production, and jointly causes the load rebound effect, and establishes a functional relationship model between the load rebound value and the time after load regulation. Based on the load rebound effect model, a method for characterizing the feasible operating region is proposed, and the constraint ranges of various physical quantities of the load rebound are obtained. Brief Description of the Drawings

[0016] Figure 1 It is a flowchart of a method for regulating large industrial loads in a distribution network considering the rebound effect according to the present invention. Detailed Embodiments

[0017] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments.

[0018] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed present invention, but merely represents some embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0019] It should be noted that, without conflict, the embodiments in the present invention and the features and technical solutions in the embodiments can be combined with each other. It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0020] Such as Figure 1As shown in the figure, the present invention provides a method for regulating large industrial loads in a distribution network considering the rebound effect. The method includes the following steps:

[0021] Step S1: Collect historical data of the industrial load system. By analyzing the historical data, establish a rebound effect model for the industrial load system. Use the rebound effect model to identify and quantify the internal relationships of the rebound effect factors in the industrial load system response. The rebound effect model decomposes the rebound effect factors into direct effect, indirect effect, environmental impact, interaction effect, and random perturbation, and quantifies the weights of the four factors through data analysis.

[0022] By deeply studying the technological process, actual participation situation, electricity consumption characteristics of large industrial loads, and their impacts on the power grid, analyze the characteristics and mechanisms of the rebound effect of large industrial loads participating in the new distribution network demand response. Establish a method for quantifying and comprehensively evaluating the rebound effect, accurately evaluate the impact of the rebound effect brought by large industrial loads participating in the new distribution network demand response, provide a scientific basis for coping with the rebound effect, and help the safe and stable operation of the power grid.

[0023] Analyze various types of historical data, extract key parameters to quantitatively measure indicators such as the intensity and duration of the rebound effect caused by large industrial loads participating in the demand response. At the same time, comprehensively consider factors such as market mechanisms, policy environments, and power grid operating states, construct a comprehensive evaluation method for the overall impact of the rebound effect of large industrial loads participating in the new distribution network demand response. Then, use technologies such as risk assessment and probability analysis to study and construct its uncertainty and risk measurement methods.

[0024] The established rebound effect model is as follows:

[0025] E(t) = A(t) + B(t) + C(t) + D(t) + ζ(t) (1);

[0026] In formula (1), E(t) is the load rebound amplitude at time point t, A(t) represents the direct effect, B(t) represents the equipment state effect; C(t) represents the environmental impact; D(t) represents the interaction effect; ζ(t) represents the random perturbation term. The physical meanings and mathematical expressions of each component are as follows:

[0027] (1) The direct effect describes the direct impact of demand response intervention on the load:

[0028] A(t) = α·AE(t) (2);

[0029] In formula (2), AE(t) represents the intensity of the demand response, and α is the direct effect coefficient.

[0030] (2) The equipment state effect reflects the indirect impact caused by changes in equipment state:

[0031]

[0032] In Equation (3), B(t) represents the state of the i-th device, β is the corresponding influence coefficient, and n is the total number of devices.

[0033] (3) The environmental impact considers the regulatory effects of external factors such as temperature:

[0034] C(t) = γ·T(t)·AE(t) (4);

[0035] In Equation (4), T(t) is the environmental temperature and γ is the environmental regulation coefficient.

[0036] (4) The interaction effect describes the interaction between devices:

[0037]

[0038] In Equation (5), δ i,j represents the interaction coefficient between devices i and j

[0039] (5) The random disturbance term characterizes the random fluctuations in the system, and it follows a normal distribution:

[0040] ζ(t) ~ N(0,σ 2 ) (5);

[0041] The Bayesian inference framework plays an important role in the analysis of the rebound effect, especially in the face of uncertainty. The Bayesian inference framework relies on the setting of the prior distribution of parameters. The prior distribution reflects the expectation or uncertainty of the parameter values in the absence of observed data. In the framework of the present invention, the setting of the prior distribution takes into account the physical characteristics of the various components of the rebound effect and the relevant background knowledge. For example, for the direct effect coefficient and the device state influence coefficient, the prior distribution is set as a normal distribution to reflect the initial assumptions about the possible values of these coefficients. For the environmental regulation coefficient and the interaction effect coefficient, a distribution form that conforms to the actual physical phenomena is also adopted to ensure that the model is more in line with the actual situation.

[0042] By constructing the likelihood function, the parameters in the model are related to the observed data. The role of the likelihood function is to represent the likelihood of the observed data occurring given the parameters. In the analysis of the rebound effect, the likelihood function not only reflects the evolution process of the device state but also describes how factors such as the demand response intensity and the environmental temperature affect the observed value of the rebound effect. By comparing these variables with the actual observed data, the estimation of each parameter can be gradually corrected. The constructed likelihood function is as follows:

[0043] For each device i, the likelihood function of its state observation is:

[0044]

[0045] In formula (6), f i () is the state transition equation for each device i, and θ represents all relevant parameters. is the observation noise variance of device i.

[0046] For the observed bounce effect E(t), its likelihood function is:

[0047]

[0048] In formula (7), μ R (t) = A(t) + B(t) + C(t) + D(t) represents the expected value of the bounce effect

[0049] Through the above steps, the Bayesian inference framework can combine prior knowledge and observed data to provide an optimal estimate of the parameters through the posterior distribution. This process can not only effectively quantify the influence degree of each effect, but also capture the uncertainties existing in the system, thus providing strong support for the further analysis and decision-making of the load bounce effect.

[0050] Step S2: Based on the influence of the bounce effect, establish a feasible operating region characterization model for the industrial load system. The feasible operating region characterization model characterizes the feasible operating region of the large industrial load system based on a deterministic optimization method to obtain the constraint ranges of various physical quantities of the load bounce in the industrial load system.

[0051] To establish the feasible operating region characterization model, the following steps are included:

[0052] First, based on the description of the large industrial load characteristics formed in S1, considering the complex multi-coupling of industrial electricity loads typical of the cement industry, each industrial electricity device needs to follow strict order and processes. Use the stochastic process theory to analyze the stochastic characteristics of industrial loads by time-sharing and in stages, and establish a load model for industrial users. Then, take industrial users in the cement industry as an example, deeply analyze the adjustable load resources of industrial users, understand the regulation categories, regulation time limits, and the proportion of adjustable loads of various resources; analyze the load characteristics of different adjustable resources from aspects such as energy consumption characteristics, start-stop time, and regulation potential to provide a basis for establishing a response characteristic model. Finally, for the analysis of the industrial user load curve, study the reasons and internal mechanisms of the load rebound phenomenon; comprehensively consider the industrial user load model and the characteristics of adjustable resources, and establish a numerical model based on factors such as the adjustable potential of the load to describe the large industrial load rebound characteristics, providing important decision-making support for system operation and scheduling.

[0053] Characterization of the feasible operating region based on the deterministic optimization method: The deterministic optimization method assumes that all parameters are known and directly solves the feasible region by establishing an optimization model.

[0054] First, an optimization model is established, and the objective function is to minimize the total cost of load fluctuation:

[0055]

[0056] In Equation (8), P t is the load value at time period t, and P r is the reference load curve. The constraint conditions for each part of the index are as follows:

[0057] 1. Ramp rate constraint

[0058] -△P max ≤P t -P t-1 ≤△P max (9);

[0059] In Equation (9), P t-1 is the load value at time period t - 1, and ΔP max is the maximum change rate of the load.

[0060] 2. Load tracking constraint

[0061] |P t -P r |≤ε (10);

[0062] In Equation (10), ε is the allowable load deviation range.

[0063] 3. Load change constraint

[0064] P min ≤P t ≤P max (11);

[0065] In Equation (11), P min is the minimum allowable value of the load, and P max is the maximum allowable value of the load.

[0066] 4. Minimum load constraint for each time period

[0067]

[0068] In Equation (12), P t min is the minimum load requirement for time period t.

[0069] 5. Daily total load constraint

[0070]

[0071] In Equation (13), P total is the fixed value of the daily total load.

[0072] 6. Peak shaving and valley filling constraints

[0073] P valley ≤ P t ≤ P peak (14);

[0074] In formula (14), P valley is the lower limit of the valley value of the load, and P peak is the upper limit of the peak value of the load.

[0075] Then, a quadratic programming solution method is used to solve the feasible solution set that satisfies all constraint conditions, and the minimization of the total load fluctuation cost is written in the standard quadratic programming form:

[0076]

[0077] In formula (15), P is the load value vector, Q = 2I, I is the identity matrix, and c = -2[P r,1 , P r,2 ,...] T .

[0078] Next, the interior point method is used. By introducing a barrier function, the constrained optimization problem is transformed into an unconstrained optimization problem, and the barrier parameter is continuously reduced until convergence to the optimal solution, completing the iterative solution.

[0079] Finally, by solving the above optimization model, a set of load values that satisfy all constraint conditions can be obtained. Plotting these solution sets in the load-time coordinate system can obtain the boundary of the operation feasible region.

[0080] Through the above steps, the solution process of the operation feasible region of large industrial loads can be obtained, and the ramp rate constraint, load tracking constraint, load change constraint, minimum load constraint by time period, daily load total constraint, and peak shaving and valley filling constraint of the operation feasible region of large industrial loads can be solved respectively;

[0081] In the large industrial system, the load rebound characteristics, the uncertainty of load changes, multi-objective optimization and trade-off, as well as the requirements of real-time and responsiveness are comprehensively processed. The quadratic programming problem is solved using a deterministic optimization method to accurately describe the operation feasible region of the system, and decisions are made under the worst distribution to provide an optimized control strategy to improve the stability, reliability, and economy of load regulation in the large industrial system under the background of load rebound.

[0082] Step S3: Establish a response characteristic model for the industrial load system based on the rebound effect. The response characteristic model includes a day-ahead scheduling model, an intraday rolling scheduling model, a real-time scheduling model, and an adaptive regulation model. Based on the constraint ranges of the physical quantities of the load rebound of the industrial load system obtained in Step S2, establish an industrial load system constraint system for the response characteristic model, including system dynamic constraints, load balance constraints, rebound characteristic constraints, and process safety constraints, and clarify the boundary of the load rebound effect of the industrial load system, providing a solid theoretical basis and practical guidance for subsequent optimization and regulation.

[0083] In step S3, a multi-time scale optimization model of large industrial loads considering an orderly peak shaving strategy is studied. The objective function is constructed with the minimum value of the total system operation cost, and the inequality constraint conditions that must be satisfied during the optimization process are constructed to achieve the economy of system operation. For large industrial loads at multiple time scales, an optimization control strategy with self-evolution ability is studied. Using a hierarchical and zonal scheduling mode, a multi-time scale self-evolution hierarchical control optimization strategy for large industrial loads under wide-area distribution is realized. Research on the multi-time scale optimization model of large industrial loads considering an orderly peak shaving strategy: First, based on the response characteristic model constructed in step S3, combined with the rebound effect, provincial-local coordination, and orderly peak shaving strategy, a day-ahead time scale optimization scheduling model is studied and constructed. The objective function is constructed with the minimum value of the total system operation cost, and inequality constraint conditions are established based on the feasible operation region of large industrial loads. An intelligent optimization algorithm, such as a genetic algorithm, fish swarm algorithm, or particle swarm algorithm, is used to optimize and solve the model to obtain the day-ahead scheduling instruction, which is used as the reference value for the intra-day scheduling of subsequent research content. Then, the actual output of industrial load scheduling resources is determined, and a continuous rolling strategy is adopted. Based on the objective function constructed during the day-ahead scheduling modeling process and considering the constraint conditions of the operation feasible region, an intra-day rolling model combined with the rebound effect, provincial-local coordination, and orderly peak shaving strategy is established. Finally, a prediction model is constructed. Based on the state variables and control variables of each load at the current moment, the output of each load at the next moment is predicted, and the minimum error between the predicted output variable and the intra-day scheduling reference trajectory is used as the objective function. Constraint conditions are constructed based on the operation feasible region, and a real-time feedback model combined with the rebound effect, provincial-local coordination, and orderly peak shaving strategy is established. Research on the multi-time scale self-evolution hierarchical optimization scheduling strategy for large industrial loads considering provincial-local coordination: First, combined with the real-time feedback and rolling optimization ideas of model predictive control, using the day-ahead scheduling instruction plan obtained in the previous part and the intra-day and real-time scheduling models, the intra-day scheduling prediction plan and real-time scheduling instruction are obtained to regulate the power of the load. Then, the regulation result is fed back in real time to roll-optimize the intra-day scheduling prediction plan, and an optimization control strategy with self-evolution ability is proposed to realize the self-evolution optimization control strategy for large industrial loads at multiple time scales and suppress the impact of the rebound characteristic on the power grid. Second, for the provincial-local coordinated scheduling problem of large-scale industrial controllable loads, relying on the current top-down centralized scheduling of the power grid, a hierarchical and zonal scheduling mode of "centralized overall coordination, decentralized local autonomy" is adopted, and a hierarchical coordinated scheduling strategy based on the provincial-local joint scheduling mode is proposed to realize the "source-load" interaction under wide-area distribution. Finally, based on the proposed multi-time scale self-evolution optimization control and hierarchical coordinated scheduling strategies, a dynamic regulation intelligent terminal device and system are developed and applied in the demonstration area to meet the operation requirements of the multi-time scale large industrial load optimization scheduling under provincial-local coordinated regulation.

[0084] In step S3, the response characteristic model modeling method adopts statistical analysis, time series modeling and nonlinear dynamic system theory, and is verified and optimized to ensure the accuracy and reliability of the model.

[0085] The objective function of the day-ahead scheduling model is to minimize the sum of the total operating costs of the system, and calculate the two cases of whether to enable peak shaving and valley filling, and count peak shaving and valley filling as rewards.

[0086] The day-ahead scheduling model formulates the power scheduling plan for large industrial sectors such as cement, steel, and electrolytic aluminum in advance through an optimization algorithm. On the premise of meeting the specific constraints of each sector, it minimizes the total operating cost to ensure the reliability and economy of the power system's power supply the next day.

[0087] In step S3, the intraday rolling scheduling model adopts model predictive control (MPC) technology to calculate future control strategies, aiming to minimize the deviation between the actual load and the day-ahead plan and reduce the operating cost at the same time.

[0088] In step S3, the AGC controller of the real-time scheduling model adopts a PI control algorithm to automatically adjust the output of the generator to respond to the real-time frequency deviation of the power grid.

[0089] In step S3: Through the rebound effect modeling method and the design of a constraint system based on the rebound effect, the key factors affecting the development of pumped storage are comprehensively analyzed and controlled. First, a feasible operating domain characterization model is constructed using statistical analysis, time series modeling and nonlinear dynamic system theory to capture the response characteristics of the system under different operating conditions, and is verified and optimized to ensure the accuracy and reliability of the model. Then, a comprehensive constraint system is designed, including system dynamic constraints, load balance constraints, rebound characteristic constraints and process safety constraints. Among them, the system dynamic constraints ensure that the operating parameters are kept within a safe range, the load balance constraints evaluate the power supply and demand relationship, the rebound characteristic constraints control the reaction amplitude and speed of the system, and the process safety constraints ensure the safety and stability of the entire system. Through this integration, the anti-interference ability of the system can be effectively improved, supporting the sustainable development of pumped storage projects. It should be noted that since this patent mainly focuses on multi-time scale optimization algorithms and adaptive control algorithms, the constraint formulas in step S3 will not be elaborated in detail.

[0090] Step S4: Formulate the power generation and scheduling plan for the next day based on the industrial load demand, rebound effect, industrial equipment demand and grid constraints predicted by the day-ahead scheduling model. Through the optimization algorithm, the power generation plan with the highest cost-benefit is obtained, while meeting the physical and operating constraints of the power grid. The key to day-ahead scheduling lies in accurate prediction and optimization to ensure the reliability and economy of power supply.

[0091] In day-ahead scheduling, considering the impact of the rebound effect, it is necessary to first determine the time period when the rebound effect may occur. For this purpose, the following formula (1) is introduced to judge whether the load reduction will cause a rebound

[0092]

[0093] In formula (16), T is the threshold, F is the predicted load, R is the threshold ratio, A is the optimized load, and D is the reduction amount.

[0094] The objective function of the day-ahead scheduling model is to minimize the sum of the total system operating costs, and calculate whether to enable peak shaving and valley filling in two cases, and regard peak shaving and valley filling as rewards. The objective function expression is:

[0095]

[0096] In formula (17), f C , f D , f T are the electricity cost, load tracking cost, and load change cost respectively. R p is the peak shaving and valley filling reward, L t is the optimized load at time period t, P t is the electricity price, W t is the tracking weight, D t + and D t - are the positive and negative deviations; C t is the load change; R p and R v are the reward weights, C p,t and F v,t are the peak shaving and valley filling amounts, I p,t and I v,t are the peak and valley indicators, and E and D are the electricity cost and change cost weights.

[0097] The objective function without enabling peak shaving and valley filling is:

[0098] F2 = e × f C + f T + d × f D (18);

[0099] In the daily operation of the power system, a series of technical, economic, and safety constraints must be adhered to ensure the stable and reliable power supply of the system. These constraints not only involve the matching of power generation and load but also include compliance with equipment operation limits and consideration of the operating characteristics of the power grid. In the steps of this patent, we will define and describe these constraints in detail so that in subsequent modeling and optimization analysis, the operating requirements and limitations of the power system can be accurately reflected. These constraints will serve as the basis for the optimization model to ensure that the proposed solutions are both practical and meet the actual operating requirements.

[0100] The ramp rate constraint can ensure that the change in power generation capacity is within a safe range, preventing equipment damage or grid instability caused by too rapid a change in power generation capacity.

[0101]

[0102] The load tracking constraint can ensure that the actual power generation closely follows the predicted load demand to maintain the power supply-demand balance of the power grid.

[0103]

[0104] The load change constraint limits the amount of load change in adjacent time periods to avoid affecting the power grid stability due to sudden load changes.

[0105] C t ≥L t -L t-1 C t ≥L t-1 -L t (21);

[0106] The time-period minimum load constraint ensures that in each time period, the power generation is not lower than the minimum load demand of that period to meet the basic power supply demand.

[0107] L t ≥M t (22);

[0108] The daily load total constraint ensures that the total power generation in a day meets the daily load demand to maintain the daily power supply-demand balance of the power grid.

[0109]

[0110] The peak shaving and valley filling constraint is used to adjust the load during peak and valley periods to reduce the load pressure on the power grid during peak periods and the excess power generation during valley periods.

[0111] For each time period t, if it is a peak period:

[0112] P t ≥Ft -L t (24);

[0113] Otherwise P t = 0;

[0114] If it is the valley period:

[0115] V t = L t -F t (25);

[0116] Otherwise V t = 0.

[0117] For the above constraint formula, R represents the ramping rate, T t represents the target load, C t represents the load change, M t represents the minimum load, D represents the total minimum daily load, P t , V t represents the peak load and the valley adjustment amount, F t represents the load forecast value.

[0118] Through the comprehensive application of these constraint conditions, we can build a comprehensive power system optimization model, which can not only meet the daily power dispatching requirements, but also maintain robustness in the face of uncertainties and changes. The integration and optimization of these constraint conditions are the key steps to achieve intelligent and automated dispatching of the power system, and also one of the core innovation points of this patent.

[0119] Step S5. Based on the day-ahead scheduling model, the intra-day rolling scheduling model updates according to real-time data, adjusts and optimizes the generation plan, adopts the model predictive control (MPC) model, and through the rolling update mechanism, continuously adjusts the scheduling strategy according to the latest grid state and market information. By analyzing the different contributions of different devices to the rebound through the feasible operating region characterization model, it helps to optimize the order and timing of device regulation, improve the flexibility and response speed of scheduling, so as to adapt to intra-day load changes and the fluctuations of renewable energy;

[0120] Step S6: The real-time scheduling model depicts the feasible operating region of the industrial load system based on the result of S1. At the moment of the power system operation, it monitors and controls the power grid in real time to respond to power grid emergencies, which include load changes and equipment failures. By finely adjusting the load distribution, it ensures the robustness of the power system in the face of uncertainties and dynamic changes. Through staged restoration or peak shaving scheduling, it mitigates the impact of the rebound on the power grid. The core of this step is the automatic generation control (AGC) model, which ensures the frequency and power balance of the power grid through a fast response mechanism and real-time calculation methods. The goal of real-time optimization is to maintain the stability and security of the power grid while optimizing the utilization of power generation resources.

[0121] Step S7: Implement the orderly peak shaving and multi-time scale self-evolving scheduling of the industrial load system through the adaptive regulation model to achieve the self-evolving update of the scheduling model. The adaptive regulation model is established based on real-time data analysis and feedback mechanism. The adaptive regulation model includes an incremental learning model, a scheduling time interval adaptive algorithm module, a rolling window length adaptive algorithm module, and a parameter dynamic update module to improve the overall system performance and stability.

[0122] The incremental learning model is responsible for constructing a powerful learning framework that can learn and optimize based on continuously increasing new data. The incremental learning model dynamically updates the model by continuously receiving and analyzing new operation data, using incremental learning technology to improve learning efficiency and reduce computational resource consumption. The scheduling time interval adaptive algorithm module automatically adjusts the scheduling time interval according to the current load changes and system status to achieve fast response and save scheduling resources. This module first collects historical data and real-time operation data, extracts features through machine learning algorithms, and trains the model. The key to incremental learning lies in its ability to quickly update through existing knowledge when new data arrives, without retraining the entire model. This method not only improves learning efficiency but also effectively reduces computational resource consumption. Finally, the incremental learning model will provide accurate predictions and suggestions for the regulation strategy to enhance the flexibility and adaptability of the system.

[0123] The above embodiments are only used to illustrate the present invention, rather than limiting the technical solutions described therein. Although this specification details the above embodiments, the present invention is not limited to the above specific implementation manners. Therefore, any technical solutions and their improvements that modify or equivalently replace the present invention without departing from the scope of the present invention are covered by the scope of the claims of the present invention.

[0124] It is easy to understand that those skilled in the art can combine, split, reorganize, etc. the embodiments of the present application based on one or several embodiments provided by the present application to obtain other embodiments, and these embodiments do not exceed the protection scope of the present application.

[0125] The above has schematically described the present invention and its embodiments. This description is not restrictive, and what is shown in the embodiments is only part of the embodiments of the present invention. The actual structure is not limited thereto. Therefore, if those of ordinary skill in the art are inspired by this and, without departing from the gist of the present invention, design similar structural forms and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. A method for controlling large industrial loads in a distribution network considering rebound effect, characterized in that: The following steps are involved: Step S1, collect historical data of industrial load system, establish a rebound effect model of industrial load system by analyzing the historical data, use the rebound effect model to identify and quantify the internal connection of rebound effect factors in the response of industrial load system, the rebound effect model decomposes the rebound effect factors into direct effect, indirect effect, environmental impact, interactive effect and random disturbance, and quantifies the weights of the four factors through data; Step S2, based on the influence of the rebound effect, a feasible operation domain characterization model of the industrial load system is established. The feasible operation domain characterization model characterizes the feasible operation domain of the load operation of the large industrial load system based on a deterministic optimization method, and obtains the constraint range of various physical quantities of the load rebound of the industrial load system including the ramp rate constraint, load tracking constraint, load change constraint, time period minimum load constraint, daily load total constraint, and peak shaving and valley filling constraint; Step S3, based on the rebound effect, a response characteristic model of the industrial load system is established, the response characteristic model includes a day-ahead scheduling model, an intraday rolling scheduling model, a real-time scheduling model and an adaptive control model, based on the constraint range of each physical quantity of the load rebound of the industrial load system obtained in step S2, an industrial load system constraint system including system dynamic constraints, load balance constraints, rebound characteristic constraints and process safety constraints is established about the response characteristic model, and the limit of the load rebound effect of the industrial load system is clarified; Step S4, formulate the power generation and dispatching plan for the next day through the industrial load demand, rebound effect, industrial equipment demand and grid constraints predicted by the day-ahead dispatching model, and optimize the most cost-effective power generation plan through the optimization algorithm while satisfying the physical and operational constraints of the grid; Step S5: Based on the day-ahead dispatch model, the intraday rolling dispatch model is updated according to real-time data to adjust and optimize the power generation plan. The model predictive control (MPC) model is adopted to continuously adjust the dispatch strategy according to the latest power grid status and market information through a rolling update mechanism. The rebound effect model is used to analyze the different contributions of different devices to the rebound, helping to optimize the order and timing of device regulation. Step S6, the real-time dispatching model monitors and controls the power grid in real time at the moment of power system operation according to the weight of the rebound effect factor of the industrial load system obtained in S1 to respond to power grid emergencies, which include load changes and equipment failures. By fine-tuning the load distribution, the robustness of the power system in the face of uncertainty and dynamic changes is ensured, and the impact of rebound on the power grid is reduced through phased recovery or peak-shifting dispatching; Step S7: Implement orderly peak shaving and multi-time scale self-evolving scheduling of the industrial load system through an adaptive control model, and realize self-evolving updating of the scheduling model; The adaptive control model is established based on real-time data analysis and feedback mechanism. The adaptive control model includes an incremental learning model, a scheduling time interval adaptive algorithm module, a rolling window length adaptive algorithm module and a parameter dynamic update module.

2. A method for controlling large industrial loads in a distribution network considering rebound effect according to claim 1, characterized in that: In step S3, the response characteristic model modeling method adopts statistical analysis, time series modeling and nonlinear dynamic system theory, and performs verification and optimization to ensure the accuracy and reliability of the model.

3. A method for controlling large industrial loads in a distribution network considering rebound effect according to claim 2, characterized in that: In step S3, the objective function of the day-ahead scheduling model is to minimize the sum of the total operating costs of the system, and calculate whether to enable peak shaving and valley filling, and calculate peak shaving and valley filling as a reward.

4. A method for controlling large industrial loads in a distribution network considering rebound effect according to claim 2, characterized in that: In step S3, the intraday rolling scheduling model uses model predictive control (MPC) technology to calculate future control strategies, aiming to minimize the deviation between the actual load and the day-ahead plan while reducing operating costs.

5. A method for controlling large industrial loads in a distribution network considering rebound effect according to claim 2, characterized in that: In step S3, the AGC controller of the real-time dispatch model adopts a PI control algorithm to automatically adjust the output of the generator to respond to the real-time frequency deviation of the power grid.

6. A method for controlling large industrial loads in a distribution network considering rebound effect according to claim 1, characterized in that: In step S7, the incremental learning model continuously receives and analyzes new operating data, dynamically updates the model using incremental learning technology, improves learning efficiency, and reduces computing resource consumption; the scheduling time interval adaptive algorithm module automatically adjusts the scheduling time interval according to the current load changes and system status to achieve rapid response and save scheduling resources; the rolling window length adaptive algorithm module dynamically adjusts the length of the observation window according to the real-time status and historical data of the system to ensure timely capture and analysis of the system status; the parameter dynamic update module is used to correct the key parameters in the algorithm in real time to ensure that it accurately reflects the current system status and environmental changes.

7. A method for controlling large industrial loads in a distribution network considering rebound effect according to claim 1, characterized in that: The response characteristic model takes into account the load rebound effect caused by the instantaneous increase in load due to the sharp increase in current when all equipment restarts after load regulation in large-load industrial production, as well as the load increase caused by the enterprises' eagerness to resume production, and establishes a functional relationship model between the load rebound value and the time after regulation.

Images