Adjustment capability rolling prediction method for multi-element regulation and control demand of power grid

Abstract

The invention relates to a power grid multi-element regulation and control demand-oriented regulation capability rolling prediction method, which comprises the following steps of 1, analyzing a power grid multi-element regulation and control demand, analyzing dynamic characteristics of a park integrated energy system, and extracting key indexes of the park integrated energy system participating in dynamic regulation and control of a power grid; 2, monitoring park load fluctuation and power grid regulation and control demand changes in real time; 3, predicting the adjustment capability of the park integrated energy system based on a data-driven robust optimization technology; 4, verifying the model precision and analyzing a prediction error according to a park load dynamic change and adjustment capability prediction result; and step 5, according to error analysis of the prediction model in the step 4, adopting a rolling prediction strategy, reducing prediction errors, optimizing the prediction model, and completing prediction of the adjustment capability of the park integrated energy system. According to the invention, the adjustment capability of the park integrated energy system can be predicted.

Description

Technical Field

[0001] The present invention belongs to the technical field of park regulation capacity prediction, and relates to a rolling prediction method for park regulation capacity, and in particular to a rolling prediction method for regulation capacity facing the diversified regulation needs of power grids. Background Art

[0002] As power users continue to increase their demand for flexible regulation of park energy consumption, predicting energy consumption regulation capabilities is one of the effective means to achieve deep participation of parks in grid peak regulation. Currently, parks can achieve flexible energy use through technologies such as demand response, energy management systems, energy storage systems, intelligent load control, and distributed energy. On this basis, a reasonable combination of energy consumption regulation prediction and other methods can better achieve deep peak regulation in parks.

[0003] In the context of the diversified regulation needs of the power grid, regulation capacity prediction can meet the peak regulation and dynamic balance needs of the power system by predicting the regulation capacity of the park at different time scales. Existing technologies are usually based on methods such as big data analysis, machine learning and numerical simulation, and comprehensively consider factors such as the historical operation data of the power grid, load demand, uncertainty in renewable energy output and scheduling strategies.

[0004] Traditional regulation capacity prediction methods mainly rely on deterministic models, which often have difficulty coping with the uncertainty problems caused by complex grid operation environments and high proportions of renewable energy access. Existing regulation capacity prediction models have computational accuracy problems in complex models, while simple models may not be able to accurately capture the diversity and dynamics of grid regulation. Grid regulation needs to be highly real-time, requiring prediction technology to respond quickly to changes in grid status. However, existing prediction technologies often have shortcomings in real-time performance, making it difficult to achieve instant perception and rapid response to grid status.

[0005] Therefore, how to effectively integrate and utilize these diverse data sources to improve prediction accuracy is a major challenge facing current technology.

[0006] After searching, no prior art documents identical or similar to the present invention were found. Summary of the invention

[0007] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a rolling prediction method for the regulation capacity of the power grid for the diversified regulation needs. Based on data-driven robust optimization technology, a rolling prediction model for the regulation capacity of the park's comprehensive energy system is established. Combined with the characteristics of multi-time scales of park energy and load uncertainty, the park's historical data is fully utilized to realize the prediction of the regulation capacity of the park's comprehensive energy system.

[0008] The present invention solves the practical problem by adopting the following technical solutions:

[0009] A method for rolling prediction of regulation capacity for multi-level regulation needs of a power grid includes the following steps:

[0010] Step 1: Analyze the needs of multi-dimensional power grid regulation and analyze the dynamic characteristics of the park's integrated energy system, and extract key indicators of the park's integrated energy system participating in the dynamic regulation of the power grid;

[0011] Step 2: Real-time monitoring of park load fluctuations and changes in grid control requirements;

[0012] Step 3: Predict the regulation capacity of the park's integrated energy system based on data-driven robust optimization technology;

[0013] Step 4: Based on the prediction results of the dynamic changes in the park load and the regulation capacity, verify the model accuracy and analyze the prediction error.

[0014] Step 5: Based on the error analysis of the prediction model in step 4, adopt a rolling prediction strategy to reduce the prediction error, optimize the prediction model, and complete the prediction of the regulation capacity of the park's integrated energy system.

[0015] Moreover, the key indicators of the park integrated energy system participating in the dynamic control of the power grid in step 1 include:

[0016] 1) Adjust capacity

[0017] The adjustable amplitude of the system's upward and downward flexibility is shown in equations (1) and (2);

[0018]

[0019] Among them, P up —Adjustable amplitude of upward flexibility, kW;

[0020] P down —Adjustable amplitude of downward flexibility, kW;

[0021] x—decision variable;

[0022] Ω—feasible domain;

[0023] P tur —Internal power generation of the park's comprehensive energy system, kW;

[0024] P t —The power consumption capacity of the system in its current operating state, kW.

[0025] The adjustable capacity of the peak-shaving and valley-filling strategy is shown in equations (3), (4), and (5);

[0026]

[0027] Among them, C p(t,τ)—peak shaving capacity at time t and within the next τ hours, kW; α p —Proportional coefficient of peak shaving regulation capability;

[0028] D p (τ) — peak demand function for time period τ, kW;

[0029] τ — rolling time window;

[0030] C v (t,τ)—valley filling regulation capacity at time t and within the next τ hours, kW;

[0031] α v —Proportional coefficient of peak shaving regulation capability;

[0032] D v (τ)—demand trough function for time period τ, kW;

[0033] C(t,τ)—total regulation capacity at time t and within the next τ hours, kW

[0034] 2) Adjust the time

[0035] Regulation time refers to the time required for the entire process from receiving the regulation signal to the actual completion of regulation in the integrated energy system. Its influence effect is detailed in equations (6), (7), and (8);

[0036]

[0037] Where, D(t)—regulation amount of market demand, kWh;

[0038] d i (t)—demand of the i-th energy source at time t, kWh;

[0039] S i (t) — supply of the i-th energy source at time t, kWh;

[0040] ΔS i (t) — adjustment amount, kWh;

[0041] R i (t)—regulation capacity, kW / h.

[0042] In the integrated energy system, the regulation capacity of the system is subject to the combined influence of multiple limiting factors. The regulation capacity constraints are detailed in formula (9);

[0043] ΔS i (t)≤R i (t)(9)

[0044] In an integrated energy system, demand satisfaction constraints are various restrictions and requirements that ensure that the system can effectively meet energy demand. They are often related to energy supply, demand response, and economic constraints. The demand satisfaction constraints are detailed in formula (10);

[0045] S i (t)≥D(t)(10)

[0046] The stability constraint is detailed in formula (11);

[0047] |S i (t)-S i (t-1)|≤C max (11)

[0048] Among them, C max —The maximum supply change allowed per unit time, kWh.

[0049] 3) Start time

[0050] The adjustment capacity, adjustment time and start time are aggregated to form a three-dimensional evaluation model, see formula (12).

[0051] Evaluation(t,h)={C p (t,τ),C v (t,τ),T(t,τ)}(12)

[0052] Among them, C p (t,τ)—peak shaving capacity after τ hours from the start time t, kW;

[0053] C v (t,τ)—valley-filling regulation capacity after τ hours from the start time t, kW;

[0054] T(t,τ)—The adjustment time after τ hours from the start time t.

[0055] The adjustment curve within the rolling cycle is updated once every hour, see formula (13) for details.

[0056] RAC(t)={Evaluation(t,h)|h∈{0,1,...,23}}(13)

[0057] Moreover, the specific steps of step 2 include:

[0058] (1) Collect park load data:

[0059] According to the control indicators extracted in step 1, the load data of the comprehensive energy system in the park is collected in real time, including load changes of different energy forms such as electricity, heat energy, and cold energy;

[0060] Collect N pairs of samples from the system historical load data to form a data set, as shown in formula (14):

[0061]

[0062] Among them, y i —The system is in (x i ,ξ i ) actual performance under conditions;

[0063] x—decision variable;

[0064] ξ—Uncertain factor.

[0065] (2) Collecting grid control demand data:

[0066] Obtain the real-time changes of the control demand identified in step 1 from the power grid side, including data on frequency regulation, reserve capacity demand, load balancing, etc.;

[0067] N pairs of samples are collected from the data of power grid regulation demand to form a data set, as shown in formula (15):

[0068]

[0069] Among them, f i —Grid frequency regulation requirements;

[0070] r i —spare capacity requirements;

[0071] p i —Load balancing requirements.

[0072] (3) Synchronize park and power grid data:

[0073] In order to synchronize the park load data with the grid control demand data to ensure the consistency of the timeliness of the two data, an interpolation function is used to map the park load data at time t to the time point of the grid control demand, as shown in Equations (16) and (17).

[0074] X i (t)≈Γ D (t,X i )(16)

[0075] T i (t)≈Γ S (t,T i )(17)

[0076] The synchronized dataset is:

[0077] X(t)=[X1(t),X2(t),...,X3(t)](18)

[0078] T(t)=[T1(t),T2(t),...,T N (t)](19)

[0079] where t∈{t1,t2,...,t n} is the time point after all synchronization.

[0080] In order to ensure the timeliness and consistency of the data, the time difference between the load data and the control demand data after synchronization is required to be as small as possible, see formula (20) for details.

[0081]

[0082] Moreover, the specific steps of step 3 include:

[0083] (1) Define the regulation capacity of the park's integrated energy system and use the data set in step 2 to form a data set of system regulation capacity;

[0084] The regulation capacity of the park integrated energy system is assumed to be the ability of the system to maintain performance under uncertain conditions, which can be expressed by formula (21):

[0085] P(t)=f(x,ξ,X(t))-AC tar (twenty one)

[0086] Among them, x—decision variable;

[0087] ξ—uncertain factor;

[0088] f(·)—actual performance of the system;

[0089] AC tar — the desired level of regulatory capability;

[0090] P(t)—the actual regulation capability of the system.

[0091] (2) Building a data-driven model:

[0092] The data set in step 2 is used to form a training set for building a data-driven model, as shown in formula (22):

[0093]

[0094] Fit the data-driven model through the training set, see formula (23) for details

[0095] P a (t)=θ·f(W2σ(W1D(t)+b1)+b2)(23)

[0096] Among them, P a (t) — the system’s predicted regulatory capabilities;

[0097] θ — model parameters;

[0098] W1 — weight matrix of the first layer;

[0099] b1 — bias vector of the first layer;

[0100] σ(·)—activation function, which makes the model have nonlinear expression ability;

[0101] W2 — weight matrix of the second layer;

[0102] b2 — Bias vector for the second layer.

[0103] (3) Aggregate data-driven models and robust optimization models:

[0104] Robust optimization technology is used to incorporate the uncertainty of load and power generation fluctuations into the prediction model to ensure the accuracy and stability of the prediction results under uncertain conditions;

[0105] The goal of robust optimization is to minimize the maximum difference performance under all uncertainties. The objective function of the robust optimization problem is shown in Equation (24);

[0106]

[0107] Among them, H γn —Probability distribution of all samples;

[0108] —Probability distribution of fuzzy sets;

[0109] On this basis, 1-normal and ∞-normal are used as constraints to limit the probability distribution of the turnout scenario in the integrated energy system, as shown in formula (25);

[0110]

[0111] in, —1 Maximum allowable deviation value under the constraint of the criterion;

[0112] —∞The maximum allowable deviation value under the criterion constraint.

[0113] The scenario probability satisfies the following confidence constraints, see equations (26) and (27).

[0114]

[0115] Among them, N γn —Total amount of historical data;

[0116] H0 γn —initial probability value;

[0117] e—Euler number.

[0118] Moreover, the specific steps of step 4 include:

[0119] (1) Verify model accuracy

[0120] Use the dataset in step 2 as the basis to form an independent dataset Verify the prediction performance of the model, see formula (28) for details.

[0121]

[0122] (2) Analysis of prediction error

[0123] The prediction error is specified as formula (29);

[0124] E(t)=P a (t)-P(t)(29)

[0125] Where, E(t)—prediction error;

[0126] P(t)—the actual regulation capacity of the system at time t;

[0127] P a (t)—The system predictive regulation capability at time t.

[0128] Moreover, the specific steps of step 5 include:

[0129] (1) Control algorithm for setting system regulation capability

[0130] After obtaining the prediction error, the PID controller is used to determine the control input, as shown in formula (30).

[0131]

[0132] Among them, K p — proportional gain;

[0133] K i —Integral gain;

[0134] K d — differential gain;

[0135] C(t)—control quantity.

[0136] (2) Rolling Forecast Adjustment Capability

[0137] By adopting a rolling forecast strategy, the forecast model can be dynamically adjusted according to the forecast results of the regulation capacity and historical data;

[0138] Set the forecast demand and historical demand data as formula (31);

[0139] H(t)={P a (t-1),P a (t-2),...,P a (tn)}(31)

[0140] Among them, P a (t)—the system prediction and regulation capability at time t;

[0141] H(t)—Historical data of the system's predictive regulation capability.

[0142] In the process of adjusting capacity rolling prediction, the system updates the prediction results at each time step t. The prediction function is realized using a weighted regression model, as shown in formula (32).

[0143]

[0144] Among them, P a (t+k)—The predicted value at time t+k.

[0145] Advantages and beneficial effects of the present invention:

[0146] 1. Improve the overall stability of the power grid: Based on the load fluctuation of the park's integrated energy system and the power grid regulation requirements (steps 1 and 2), the present invention uses data-driven robust optimization technology to predict the park's regulation capacity (step 3). This method can effectively improve the support capacity of the park's integrated energy system for the power grid, help the power grid to smooth out load fluctuations, and enhance stability.

[0147] 2. Improve the energy utilization rate of the park: Data-driven robust optimization technology can realize the prediction of the park load (step 3), provide a reference for the optimization of the park scheduling strategy, improve the rationality and efficiency of energy utilization within the park, and avoid energy waste.

[0148] 3. Improve the accuracy of prediction: This method relies on data-driven robust optimization technology and a rolling forecast strategy (step 4). It continuously uses the latest and historical data of the park load to improve the prediction accuracy in an environment with high load fluctuations and energy uncertainty.

[0149] 4. Improve response speed: By analyzing the error of the prediction results and the accuracy of the model (step 5), the rolling prediction method is used to adjust the prediction model, so that the park can maintain excellent responsiveness and adaptability in a dynamic environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0150] Figure 1 is a process flow chart of the present invention;

[0151] Figure 2 It is a processing flow chart of the data-driven robust optimization technology of the present invention. DETAILED DESCRIPTION

[0152] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings:

[0153] A rolling prediction method for regulation capacity of power grid facing multiple regulation needs, such as Figure 1 and Figure 2 As shown, the following steps are included:

[0154] A method for rolling prediction of regulation capacity for multi-level regulation needs of a power grid includes the following steps:

[0155] Step 1: Analyze the needs of multi-dimensional power grid regulation and analyze the dynamic characteristics of the park's integrated energy system, and extract key indicators of the park's integrated energy system participating in the dynamic regulation of the power grid;

[0156] The specific steps of step 1 include:

[0157] (1) Extracting key indicators for park regulation

[0158] The park establishes a regulation curve through real-time monitoring and data analysis, and adjusts resource allocation or operation strategy according to the diversified regulation needs of the power grid to maintain efficient operation and resource optimization. By analyzing the regulation curve of the integrated energy system, accurate prediction of the system regulation capability can be achieved, so that the regulation capability of the park's integrated energy system can match the diversified regulation needs of the power grid. The present invention aggregates the regulation capacity (peak shaving, valley filling), regulation time (according to market demand) and start time (24 hours, one per hour) to form a regulation curve to evaluate the regulation capability of the park's integrated energy system and provide support for regulation capability prediction.

[0159] 4) Adjust capacity

[0160] The adjustable capacity of the park's integrated energy system participating in the dynamic regulation of the power grid is mainly measured by the adjustable amplitude and the peak shaving and valley filling regulation capacity. The adjustable amplitude refers to the maximum value of the park's power generation that can be increased or decreased within a given operating time while meeting the system's own needs. It is determined by the system's current operating point and the maximum and minimum outputs under the system's power consumption constraints. The adjustable amplitude of the system's upward and downward flexibility is shown in equations (1) and (2).

[0161]

[0162]

[0163] Among them, P up —Adjustable amplitude of upward flexibility, kW;

[0164] P down —Adjustable amplitude of downward flexibility, kW;

[0165] x—decision variable;

[0166] Ω—feasible domain;

[0167] P tur —Internal power generation of the park's comprehensive energy system, kW;

[0168] P t —The power consumption capacity of the system in its current operating state, kW.

[0169] The participation of the park's integrated energy system in power grid regulation involves two control strategies: peak shaving and valley filling. Peak shaving reduces the pressure on the power grid by reducing the load during peak periods, while valley filling balances the load of the power grid by increasing the load during valley periods. The adjustable capacity of the peak shaving and valley filling strategy is shown in equations (3), (4), and (5).

[0170]

[0171] Among them, C p (t,τ)—peak shaving capacity at time t and within the next τ hours, kW;

[0172] α p —Proportional coefficient of peak shaving regulation capability;

[0173] D p (τ) — peak demand function for time period τ, kW;

[0174] τ — rolling time window;

[0175] C v (t,τ)—valley filling regulation capacity at time t and within the next τ hours, kW;

[0176] α v —Proportional coefficient of peak shaving regulation capability;

[0177] D v (τ)—demand trough function for time period τ, kW;

[0178] C(t,τ)—total regulation capacity at time t and within the next τ hours, kW

[0179] 5) Adjust the time

[0180] Regulation time refers to the time required for the entire process from receiving the regulation signal to the actual completion of the regulation of the integrated energy system. Shortening the regulation time can not only improve the response speed and flexibility of the system, but also optimize the energy efficiency and enhance the overall performance and reliability of the system. Market demand is an important factor affecting the regulation time, and its influence effect is detailed in equations (6), (7), and (8).

[0181]

[0182] ΔS i (t) = S i (t)-S i (t-1) (7)

[0183]

[0184] Where, D(t)—regulation amount of market demand, kWh;

[0185] d i (t)—demand of the i-th energy source at time t, kWh;

[0186] S i (t) — supply of the i-th energy source at time t, kWh;

[0187] ΔS i (t) — adjustment amount, kWh;

[0188] R i (t)—regulation capacity, kW / h.

[0189] In an integrated energy system, the regulation capacity of the system is subject to the combined influence of multiple limiting factors, which may affect the flexibility, response speed and overall efficiency of the system. The regulation capacity constraint is detailed in equation (9).

[0190] ΔS i (t)≤R i (t) (9)

[0191] In an integrated energy system, demand satisfaction constraints are various restrictions and requirements that ensure that the system can effectively meet energy demand, which are often related to energy supply, demand response, and economic constraints. The demand satisfaction constraints are detailed in formula (10).

[0192] S i (t)≥D(t) (10)

[0193] The stability constraint of the system regulation capability ensures that the system can maintain its stability and performance when subjected to external disturbances or changes. In terms of energy supply, it should be ensured that the system's energy supply matches the system regulation requirements to avoid large supply changes. The stability constraint is detailed in Equation (11).

[0194] |S i (t)-S i (t-1)|≤C max (11)

[0195] Among them, C max —The maximum supply change allowed per unit time, kWh.

[0196] 6) Start time

[0197] The start time of the rolling forecast refers to the time when the system starts a new regulation cycle at a fixed time every hour within the regulation cycle. That is, at a fixed time every hour, the system starts regulation measures to adjust energy production, distribution or consumption to match the current demand forecast or actual demand.

[0198] The adjustment capacity, adjustment time and start time are aggregated to form a three-dimensional evaluation model, see formula (12).

[0199] Evaluation(t,h)={C p (t,τ),C v (t,τ),T(t,τ)}(12)

[0200] Among them, C p (t,τ)—peak shaving capacity after τ hours from the start time t, kW;

[0201] C v (t,τ)—valley-filling regulation capacity after τ hours from the start time t, kW;

[0202] T(t,τ)—The adjustment time after τ hours from the start time t.

[0203] The adjustment curve within the rolling cycle is updated once every hour, see formula (13) for details.

[0204] RAC(t)={Evaluation(t,h)|h∈{0,1,...,23}}(13)

[0205] Step 2: Real-time monitoring of park load fluctuations and changes in grid control requirements;

[0206] The specific steps of step 2 include:

[0207] (1) Collect park load data:

[0208] According to the control index extracted in step 1, the load data of the integrated energy system in the park is collected in real time, including load changes of different energy forms such as electricity, heat energy, and cooling energy. N pairs of samples are collected from the historical load data of the system to form a data set, as shown in formula (14):

[0209]

[0210] Among them, y i —The system is in (x i ,ξ i ) actual performance under conditions;

[0211] x—decision variable;

[0212] ξ—Uncertain factor.

[0213] (2) Collecting grid control demand data:

[0214] The real-time changes of the control demand identified in step 1 are obtained from the power grid side, including data on frequency regulation, spare capacity demand, load balancing, etc. N pairs of samples are collected from the data on power grid control demand to form a data set, as shown in formula (15):

[0215]

[0216] Among them, f i —Grid frequency regulation requirements;

[0217] r i —spare capacity requirements;

[0218] p i —Load balancing requirements.

[0219] (3) Synchronize park and power grid data:

[0220] In order to synchronize the park load data with the grid control demand data and ensure the consistency of the timeliness of the two data, an interpolation function is used to map the park load data at time t to the time point of the grid control demand, as shown in Equations (16) and (17) for details.

[0221] X i (t)≈Γ D (t,X i )(16)

[0222] T i (t)≈Γ S (t,T i )(17)

[0223] The synchronized dataset is:

[0224] X(t)=[X1(t),X2(t),...,X3(t)](18)

[0225] T(t)=[T1(t),T2(t),...,T N (t)](19)

[0226] where t∈{t1,t2,...,t n} is the time point after all synchronization.

[0227] In order to ensure the timeliness and consistency of the data, the time difference between the load data and the control demand data after synchronization is required to be as small as possible, see formula (20) for details.

[0228]

[0229] Step 3: Predict the regulation capacity of the park's integrated energy system based on data-driven robust optimization technology;

[0230] The specific steps of step 3 include:

[0231] (1) Define the regulation capability of the park's integrated energy system, and use the data set in step 2 to form a data set of the system's regulation capability: Regulation capability usually refers to the ability of the park to operate smoothly under load fluctuations, changes in power generation, and changes in energy storage status. The regulation capability of the park's integrated energy system is set as the system's ability to maintain performance under uncertain conditions, which can be expressed by formula (21):

[0232] P(t)=f(x,ξ,X(t))-AC tar (twenty one)

[0233] Among them, x—decision variable;

[0234] ξ—uncertain factor;

[0235] f(·)—actual performance of the system;

[0236] AC tar — the desired level of regulatory capability;

[0237] P(t)—the actual regulation capability of the system.

[0238] (2) Building a data-driven model:

[0239] The data set in step 2 is used to form a training set for building a data-driven model, as shown in formula (22):

[0240]

[0241] Fit the data-driven model through the training set, see formula (23) for details

[0242] P a (t)=θ·f(W2σ(W1D(t)+b1)+b2)(23)

[0243] Among them, P a (t) — the system’s predicted regulatory capabilities;

[0244] — model parameters;

[0245] W1 — weight matrix of the first layer;

[0246] b1 — bias vector of the first layer;

[0247] σ(·)—activation function, which makes the model have nonlinear expression ability;

[0248] W2 — weight matrix of the second layer;

[0249] b2 — Bias vector for the second layer.

[0250] (3) Aggregate data-driven models and robust optimization models:

[0251] Robust optimization technology is used to incorporate the uncertainty of load and power generation fluctuations into the prediction model to ensure the accuracy and stability of the prediction results under uncertain conditions.

[0252] Data-Driven Robust Optimization (DRRO) provides a cutting-edge method for integrated energy system scheduling that does not require a preset probability distribution. Traditional robust optimization relies on assumptions about the probability distribution of uncertainty, which may lead to overly conservative decisions due to the limitations of model assumptions. In contrast, DRRO constructs fuzzy sets by extracting statistical characteristics from historical data, which can effectively simulate and respond to the worst probability distribution patterns, thereby improving the system's robustness to extreme uncertain scenarios. The mathematical transformation of the DRRO model integrates fuzzy set construction and historical data analysis into the optimization framework, making optimization decisions more accurate and effective when faced with a variety of uncertain scenarios.

[0253] The goal of robust optimization is to minimize the maximum difference performance under all uncertainties. The objective function of the robust optimization problem is constructed as shown in formula (24).

[0254]

[0255] Among them, H γn —Probability distribution of all samples;

[0256] —Probability distribution of fuzzy sets;

[0257] Theoretically, the feasible range of scenario probabilities It can be arbitrary. However, in order to make it closer to the actual operation data and ensure the volatility of the probability distribution within a reasonable range, the initial probability distribution value of each discrete scenario is usually taken as the central value. On this basis, 1-normal and ∞-normal are used as constraints to limit the probability distribution of the branching scenario in the integrated energy system, as shown in formula (25).

[0258]

[0259] in, —1 Maximum allowable deviation value under the constraint of the criterion;

[0260] —∞The maximum allowable deviation value under the criterion constraint.

[0261] The scenario probability satisfies the following confidence constraints, see equations (26) and (27).

[0262]

[0263] Among them, N γn —Total amount of historical data;

[0264] H0 γn —initial probability value;

[0265] e—Euler number.

[0266] Step 4: Based on the prediction results of the dynamic changes in the park load and the regulation capacity, verify the model accuracy and analyze the prediction error.

[0267] The specific steps of step 4 include:

[0268] (1) Verify model accuracy

[0269] Verifying the accuracy of the model is a critical step to ensure that the model works effectively in real-world applications. By using methods such as cross-validation and training / test set splits, combined with appropriate evaluation metrics, the performance of the model can be fully evaluated. Continuous model tuning and testing can improve the predictive power and reliability of the model, ensuring its effectiveness in real-world applications.

[0270] Use the dataset in step 2 as the basis to form an independent dataset Verify the prediction performance of the model, see formula (28) for details.

[0271]

[0272] (2) Analysis of prediction error

[0273] Prediction error refers to the difference between the model prediction value and the actual observation value. It is one of the important indicators for evaluating the performance of the prediction model. By analyzing the prediction error, we can understand the prediction accuracy of the model, find potential problems, and optimize the model. The prediction error is defined as formula (29).

[0274] E(t)=P a (t)-P(t)(29)

[0275] Where, E(t)—prediction error;

[0276] P(t)—the actual regulation capacity of the system at time t;

[0277] P a (t)—The system predictive regulation capability at time t.

[0278] Step 5: Based on the error analysis of the prediction model in step 4, adopt a rolling prediction strategy to reduce the prediction error, optimize the prediction model, and complete the prediction of the regulation capacity of the park's integrated energy system.

[0279] In this embodiment, the specific steps of step 5 include:

[0280] (1) Control algorithm for setting system regulation capability

[0281] After obtaining the prediction error, the PID controller is used to determine the control input. See equation (30) for details.

[0282]

[0283] Among them, K p — proportional gain;

[0284] K i —Integral gain;

[0285] K d — differential gain;

[0286] C(t)—control quantity.

[0287] (2) Rolling Forecast Adjustment Capability

[0288] By adopting a rolling forecast strategy, the forecast model can be dynamically adjusted based on the forecast results of the regulation capacity and historical data.

[0289] Set the forecast demand and historical demand data as formula (31).

[0290] H(t)={P a (t-1),P a (t-2),...,P a (tn)}(31)

[0291] Among them, P a (t)—the system prediction and regulation capability at time t;

[0292] H(t)—Historical data of the system's predictive regulation capability.

[0293] In the process of adjusting capacity rolling prediction, the system updates the prediction results at each time step t. The prediction function is realized using a weighted regression model, as shown in formula (32).

[0294]

[0295] Among them, P a (t+k)—The predicted value at time t+k.

[0296] The working principle of the present invention is:

[0297] The rolling prediction method of regulation capacity for the diversified regulation needs of the power grid is based on the diversified regulation needs of the power grid, and the regulation capacity of the park is rolled out. The protection focus is on the prediction method and evaluation indicators. First, the diversified regulation needs of the power grid are analyzed and the dynamic characteristics of the park's integrated energy system are analyzed, and the key indicators of the park's integrated energy system participating in the dynamic regulation of the power grid are extracted; further, the load fluctuations of the park and the changes in the regulation needs of the power grid are monitored; further, based on the data-driven robust optimization technology, the load, power generation and energy storage of the park's integrated energy system are predicted to evaluate the regulation capacity of the park; further, according to the dynamic changes of the park's load and the prediction results of the regulation capacity, the accuracy and prediction error of the prediction model are analyzed; further, the prediction model is optimized based on the rolling prediction method to continuously improve the prediction accuracy and regulation effect.

[0298] In the application scenario facing the diversified control needs of the power grid, the present invention takes into account the regulation capacity of the park to achieve the support of the external power grid by the comprehensive energy within the park, promote the interaction between the park and the power grid, enhance the stability of the power grid and improve the energy utilization rate of the park.

[0299] The present invention proposes a rolling prediction method for the regulation capacity of a park's integrated energy system based on data-driven robust optimization technology. It makes full use of the park's historical load data and, taking into account grid demand fluctuations and park load uncertainties, performs a rolling prediction of the regulation capacity of the park's integrated energy system, optimizes energy management and improves operational efficiency.

[0300] The present invention combines the characteristics of multi-time scales and load uncertainty of park energy and adopts a rolling prediction strategy for the park's multi-energy regulation. According to the latest data of the park's energy system and the prediction results of the regulation capacity, and considering the dynamic regulation requirements of the external power grid and the load changes of the internal comprehensive energy system, the park's regulation capacity prediction model is adjusted and updated in real time to match the flexible regulation requirements of the power grid.

[0301] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0302] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0303] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0304] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

Claims

1. A rolling prediction method for regulation capacity facing the multi-level regulation needs of power grid, characterized by: The following steps are involved: Step 1: Analyze the needs of multi-dimensional power grid regulation and analyze the dynamic characteristics of the park's integrated energy system, and extract key indicators of the park's integrated energy system participating in the dynamic regulation of the power grid; Step 2: Real-time monitoring of park load fluctuations and changes in grid control requirements; Step 3: Predict the regulation capacity of the park's integrated energy system based on data-driven robust optimization technology; Step 4: Verify the model accuracy and analyze the prediction error based on the prediction results of the park load dynamic change and regulation capacity; Step 5: Based on the error analysis of the prediction model in step 4, adopt a rolling prediction strategy to reduce the prediction error, optimize the prediction model, and complete the prediction of the regulation capacity of the park's integrated energy system.

2. The method for rolling prediction of regulation capacity for multi-dimensional regulation needs of power grid according to claim 1 is characterized by: The key indicators of the park integrated energy system participating in the dynamic control of the power grid in step 1 include: 1) Adjust capacity The adjustable amplitude of the system's upward and downward flexibility is shown in equations (1) and (2); Among them, P up —Adjustable amplitude of upward flexibility, kW; P down —Adjustable amplitude of downward flexibility, kW; x—decision variable; Ω—feasible domain; P tur —Internal power generation of the park's comprehensive energy system, kW; P t —Power consumption capacity of the system in current operation state, kW; The adjustable capacity of the peak-shaving and valley-filling strategy is shown in equations (3), (4), and (5); Among them, C p (t,τ)—peak shaving capacity at time t and within the next τ hours, kW; α p —Proportional coefficient of peak shaving regulation capability; D p (τ) — peak demand function for time period τ, kW; τ — rolling time window; C v (t,τ)—valley filling regulation capacity at time t and within the next τ hours, kW; α v —Proportional coefficient of peak shaving regulation capability; D v (τ)—demand trough function for time period τ, kW; C(t,τ)—total regulation capacity at time t and within the next τ hours, kW 2) Adjust the time Regulation time refers to the time required for the entire process from receiving the regulation signal to the actual completion of the regulation of the integrated energy system. Its influence effect is detailed in equations (6), (7), and (8); ΔS i (t)=S i (t)-S i (t-1) (7) Where, D(t)—regulation amount of market demand, kWh; d i (t)—demand of the i-th energy source at time t, kWh; S i (t) — supply of the i-th energy source at time t, kWh; ΔS i (t) — adjustment amount, kWh; R i (t)—regulation capacity, kW / h; In the integrated energy system, the regulation capacity of the system is subject to the combined influence of multiple limiting factors. The regulation capacity constraints are detailed in formula (9); ΔS i (t)≤R i (t) (9) In an integrated energy system, demand satisfaction constraints are various restrictions and requirements that ensure that the system can effectively meet energy demand. They are often related to energy supply, demand response, and economic constraints. The demand satisfaction constraints are detailed in formula (10); S i (t)≥D(t) (10) The stability constraint is shown in formula (11); |S i (t)-S i (t-1)|≤C max (11) Among them, C max —The maximum supply change allowed per unit time, kWh; 3) Start time The adjustment capacity, adjustment time and start time are aggregated to form a three-dimensional evaluation model, see formula (12); Evaluation(t,h)={C p (t,τ),C v (t,τ),T(t,τ)} (12) Among them, C p (t,τ)—peak shaving capacity after τ hours from the start time t, kW; C v (t,τ)—valley-filling regulation capacity after τ hours from the start time t, kW; T(t,τ)—the adjustment time after τ hours from the start time t; The regulation curve within the rolling cycle is updated once every hour, see formula (13) for details; RAC(t)={Evaluation(t,h)|h∈{0,1,...,23}} (13) 3. The method for rolling prediction of regulation capacity for multi-dimensional regulation needs of power grid according to claim 1 is characterized by: The specific steps of step 2 include: (1) Collect park load data: According to the control indicators extracted in step 1, the load data of the comprehensive energy system in the park is collected in real time, including load changes of different energy forms such as electricity, heat energy, and cold energy; Collect N pairs of samples from the system historical load data to form a data set, as shown in formula (14): Among them, y i —The system is in (x i ,ξ i ) actual performance under conditions; x—decision variable; ξ—uncertain factor; (2) Collecting grid control demand data: Obtain the real-time changes of the control demand identified in step 1 from the power grid side, including data on frequency regulation, reserve capacity demand, load balancing, etc.; N pairs of samples are collected from the data of power grid regulation demand to form a data set, as shown in formula (15): Among them, f i —Grid frequency regulation requirements; r i —spare capacity requirements; p i —Load balancing requirements; (3) Synchronize park and power grid data: In order to synchronize the park load data with the grid control demand data to ensure the consistency of the timeliness of the two data, an interpolation function is used to map the park load data at time t to the time point of the grid control demand, as shown in equations (16) and (17). X i (t)≈Γ D (t,X i )(16) T i (t)≈Γ S (t,T i )(17) The synchronized dataset is: X(t)=[X1(t),X2(t),...,X3(t)] (18) T(t)=[T1(t),T2(t),...,T N (t)] (19) where t∈{t1,t2,...,t n } is the time point after all synchronization; In order to ensure the timeliness and consistency of the data, the time difference between the load data and the control demand data after synchronization is required to be as small as possible, as shown in formula (20); 4. The method for rolling prediction of regulation capacity for multi-dimensional regulation needs of power grid according to claim 1 is characterized by: The specific steps of step 3 include: (1) Define the regulation capability of the integrated energy system of the park, and use the data set in step 2 to form a data set of system regulation capability; The regulation capacity of the park integrated energy system is assumed to be the ability of the system to maintain performance under uncertain conditions, which can be expressed by formula (21): P(t)=f(x,ξ,X(t))-AC tar (21) Among them, x—decision variable; ξ—uncertain factor; f(·)—actual performance of the system; AC tar — the desired level of regulatory capability; P(t)—the actual regulation capability of the system; (2) Building a data-driven model: The data set in step 2 is used to form a training set for building a data-driven model, as shown in formula (22): Fit the data-driven model through the training set, see formula (23) for details P a (t)=θ·f(W2σ(W1D(t)+b1)+b2) (23) Among them, P a (t) — the system’s predicted regulatory capabilities; — model parameters; W1 — weight matrix of the first layer; b1 — bias vector of the first layer; σ(·)—activation function, which makes the model have nonlinear expression ability; W2 — weight matrix of the second layer; b2 — bias vector of the second layer; (3) Aggregate data-driven models and robust optimization models: Robust optimization technology is used to incorporate the uncertainty of load and power generation fluctuations into the prediction model to ensure the accuracy and stability of the prediction results under uncertain conditions; The goal of robust optimization is to minimize the maximum difference performance under all uncertainties. The objective function of the robust optimization problem is shown in formula (24); Among them, H γn —Probability distribution of all samples; —Probability distribution of fuzzy sets; On this basis, 1-normal and ∞-normal are used as constraints to limit the probability distribution of the turnout scenario in the integrated energy system, as shown in formula (25); in, —1 Maximum allowable deviation value under the constraint of the criterion; —∞ Maximum allowable deviation value under the criterion constraint; The scenario probability satisfies the following confidence constraints, see equations (26) and (27); Among them, N γn —Total amount of historical data; H0 γn —initial probability value; e—Euler number.

5. The method for rolling prediction of regulation capacity for multi-dimensional regulation needs of power grid according to claim 1 is characterized by: The specific steps of step 4 include: (1) Verify model accuracy Use the dataset in step 2 as the basis to form an independent dataset Verify the model prediction performance, see formula (28) for details; (2) Analysis of prediction error The prediction error is specified as formula (29); E(t)=P a (t)-P(t) (29) Among them, E(t)—prediction error; P(t)—the actual regulation capability of the system at time t; P a (t)—The system predictive regulation capability at time t.

6. The method for rolling prediction of regulation capacity for multi-dimensional regulation needs of power grid according to claim 1 is characterized by: The specific steps of step 5 include: (1) Control algorithm for setting system regulation capability After obtaining the prediction error, the PID controller is used to determine the control input, as shown in formula (30); Among them, K p — proportional gain; K i —Integral gain; K d — differential gain; C(t)—control quantity; (2) Rolling Forecast Adjustment Capability By adopting a rolling forecast strategy, the forecast model can be dynamically adjusted according to the forecast results of the regulation capacity and historical data; Set the forecast demand and historical demand data as formula (31); H(t)={P a (t-1),P a (t-2),...,P a (t-n)} (31) Among them, P a (t)—the system prediction and regulation capability at time t; H(t)—historical data of the system’s predictive regulation capability; In the process of adjusting capacity rolling prediction, the system updates the prediction results at each time step t; the prediction function is realized using a weighted regression model, as shown in formula (32); Among them, P a (t+k)—The predicted value at time t+k.

Images