A multi-time scale optimal scheduling method for a nuclear energy-based integrated energy system

Abstract

The present application belongs to the technical field of comprehensive energy system optimal scheduling, and particularly relates to a multi-time scale optimal scheduling method for a comprehensive energy system based on nuclear energy. Step 1: Collect historical power load, photovoltaic power generation data and related meteorological information to lay a foundation for the training of a deep learning prediction model, and use a deep learning algorithm for prediction; Step 2: intelligently adjust the weight of the prediction result to find the best weight combination; Step 3: build a day-ahead scheduling optimization model; Step 4: execute an intra-day rolling scheduling and real-time scheduling strategy. Deep learning includes the Transformer and CNN-LSTM algorithms. The two algorithms are simultaneously run using a parallel architecture. The present application uses a grey wolf optimization algorithm to intelligently adjust the weight of the prediction results of the Transformer and CNN-LSTM. The present application efficiently integrates and schedules various energy resources, including nuclear energy, photovoltaic power generation, pumped storage and molten salt energy storage, to optimize the operating efficiency and cost-effectiveness of the entire energy system.

Description

Technical Field

[0001] This invention belongs to the field of integrated energy system optimization scheduling technology, specifically involving a multi-timescale optimization scheduling method for integrated energy systems based on nuclear energy. Background Technology

[0002] With the growing global demand for clean energy, nuclear energy, with its low-carbon characteristics and high energy density, has become an indispensable energy form, playing a crucial role in ensuring long-term, stable energy supply and achieving environmental goals. However, the efficiency and reliability of power systems urgently need to be further improved through the intelligent integration and peak-shaving of nuclear energy with other energy forms, especially by combining it with renewable energy sources such as wind and solar power, to achieve all-weather energy supply and demand response.

[0003] Within this framework, leveraging deep learning technology for high-precision forecasting of electricity load and photovoltaic power generation is of particular value. By accurately simulating and predicting energy demand, these advanced data analytics tools can significantly improve the responsiveness of energy systems, optimize resource allocation, and ultimately achieve the goals of improving energy efficiency and reducing costs.

[0004] Single-time-scale optimization has significant limitations when dealing with complex energy systems. This approach often fails to adequately adapt to rapid changes in electricity demand and market fluctuations. Implementing multi-time-scale energy dispatch methods can effectively overcome these shortcomings. By comprehensively considering and optimizing energy dispatch across multiple time scales—day-ahead, intraday, and real-time—this method can more comprehensively address the dynamic changes in energy demand and the uncertainties of market conditions. Multi-time-scale strategies provide power system managers with a comprehensive, flexible, and efficient operational framework, significantly improving the economics and stability of the energy system. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-timescale optimization scheduling method for integrated energy systems based on nuclear energy. It utilizes deep learning technology to establish predictive models for load and photovoltaic power generation, accurately predicting future electricity demand and photovoltaic power generation. Based on accurate predictive data, it implements day-ahead optimization, intraday rolling optimization, and real-time scheduling optimization. This method efficiently integrates and schedules various energy resources, including nuclear energy, photovoltaic power generation, pumped storage, and molten salt storage, to optimize the overall operating efficiency and cost-effectiveness of the energy system.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A multi-timescale optimization scheduling method for integrated energy systems based on nuclear energy:

[0008] Step 1: Collect historical power load, photovoltaic power generation data and related meteorological information to lay the foundation for training the deep learning prediction model and use deep learning algorithms for prediction;

[0009] Step 2: Perform intelligent weight adjustment on the prediction results to find the optimal weight combination;

[0010] Step 3: Construct a day-ahead scheduling optimization model;

[0011] Step 4: Execute intraday rolling scheduling and real-time scheduling strategies.

[0012] Deep learning includes Transformer and CNN-LSTM algorithms.

[0013] A parallel architecture is used to allow the two algorithms to run simultaneously.

[0014] The Grey Wolf Optimization Algorithm is applied to intelligently adjust the weights of the prediction results of Transformer and CNN-LSTM.

[0015] By defining an objective function with prediction error as its core, and using the Grey Wolf optimization algorithm to find the weight configuration that minimizes this function, the optimal overall prediction accuracy is achieved.

[0016] A day-ahead scheduling optimization model is constructed with a time scale of 1 hour. Based on day-ahead forecasts of photovoltaic and power loads, the model takes into account the operational limitations and constraints of each device in the integrated energy system, the stable output of nuclear power, the intermittency of photovoltaic power, and the peak-shaving capabilities of molten salt and pumped storage.

[0017] Step 4: Consider the cost of power change penalty terms and dynamically adjust the day-ahead plan based on real-time data and forecast results.

[0018] Step 2: Apply the Grey Wolf Optimization (GWO) algorithm to intelligently adjust the weights of the Transformer and CNN-LSTM prediction results, find the optimal weight combination, define the objective function for the prediction error, and find the weight configuration that minimizes this function using GWO. This includes the following steps:

[0019] (1) Comprehensive prediction model

[0020] Y pred =ω T ·Y T +ω CNN-LSTM ·Y CNN-LSTM (11)

[0021] In the formula, Y pred This is a comprehensive prediction result; Y T This is the prediction result of the Transformer model; Y CNN-LSTMThis is the prediction result of the CNN-LSTM model; ω T and ω CNN-LSTM These are the weights of the two models, optimized by the GWO algorithm;

[0022] (2) Objective function

[0023] Minimize F(ω T ,ω CNN-LSTM (12)

[0024] In the formula, F is the objective function of the prediction error, and the goal of the GWO algorithm is to find the weights ω that minimize F. T and ω CNN-LSTM The GWO algorithm continuously updates the weights ω through an iterative process. T and ω CNN-LSTM In order to find the minimum value of the objective function.

[0025] Step 3: First, establish a day-ahead scheduling optimization model.

[0026]

[0027] In the formula, C pumped-storage (t) represents the pumped storage cost in time period t, C molten-salt (t) represents the cost of molten salt energy storage in time period t, C nuclear (t) represents the nuclear power cost in time period t, C PV (t) represents the cost of photovoltaic power, C start-stop (t) is the start-up and shutdown cost, which refers to the additional costs incurred when starting or stopping the equipment, including additional wear and tear and energy consumption, t = 24;

[0028] Nuclear power cost:

[0029] C nuclear (t)=C variable,nuclear ·P gen,nuclear (t) (14)

[0030] In the formula, C variable,nuclear This refers to the unit cost of nuclear power generation, including the initial investment, the converted unit cost over the service life, fuel costs, and unit operation and maintenance costs; P gen,nuclear (t) represents the amount of nuclear power generated during time period t;

[0031] Photovoltaic cost:

[0032] C PV (t)=C variable,PV ·P gen,PV (t) (15)

[0033] In the formula, C variable,PVThis refers to the unit cost of electricity generated by a photovoltaic system, including the initial investment, the converted unit cost over the service life, and the unit operation and maintenance cost, P. gen,PV (t) represents the photovoltaic power generation during time period t;

[0034] Molten salt cost:

[0035] C molten-salt (t)=C variable,molten-salt ·E molten-salt (t) (16)

[0036] In the formula, C variable,molten-salt It is the unit cost of electricity generation, including the initial investment converted to a unit cost over the service life, and the cost of energy conversion and storage. molten-salt (t) represents the change in energy storage over time period t;

[0037] Pumped storage cost:

[0038] C pumped-storage (t)=C variable,pumped-storage ·E pumped-storage (t) (17)

[0039] In the formula, C variable,pumped-storage It is the unit cost of electricity generation, including the initial investment converted to a unit cost over the service life, and the cost of energy conversion and storage. pumped-storage (t) represents the change in energy storage over time period t;

[0040] Start-up and shutdown costs:

[0041]

[0042] In the formula, C start-stop It is the total start-up and shutdown cost of all equipment, C start-stop,nuclear The cost of a single start-up and shutdown of nuclear power equipment, C start-stop,PV The start-up and shutdown cost of photovoltaic equipment, C start-stop,molten-salt The single start-up and shutdown cost of molten salt energy storage equipment, C start-stop,pumped-storage The single start-up and shutdown cost of pumped storage equipment, S nuclear S PV S molten-salt S pumped-storage These are the number of start-ups and shutdowns of nuclear power, photovoltaic, molten salt energy storage, and pumped storage equipment within the considered time period;

[0043] The constraints for optimizing various cost ratios are as follows:

[0044] (1) Power balance constraint

[0045]

[0046] In the formula, Pgen,PV P(t) is the photovoltaic power generation at time t. discharge,molten-salt P(t) is the power of the molten salt storing heat and releasing it into electrical energy at time t. gen,pumped-storage (t) represents the pumped-storage power generation at time t, P load (t) represents the total power of the user load at time t, P charge,molten-salt P(t) is the power of the molten salt energy storage and heat absorption at time t. charge,pumped-storage (t) is the power that the pumped hydro storage begins to store at time t;

[0047] (2) Nuclear power constraint including ramp rate

[0048] Power generation range:

[0049] P min,nuclear ≤P gen,nuclear (t)≤P max,nuclear (20)

[0050] Slope rate:

[0051] |P gen,nuclear (t)-P gen,nuclear (t-1)|≤ΔP max,nuclear (twenty one)

[0052] In the formula, P min,nuclear It is the minimum power generation capacity of a nuclear power plant, P gen,nuclear (t) represents the amount of electricity generated by nuclear power at time t, P max,nuclear It is the maximum power generation of the nuclear power plant, ΔP max,nuclear This is the maximum permissible ramp rate for a nuclear power plant;

[0053] (3) Constraints on photovoltaic power generation

[0054] Power generation is affected by sunlight conditions:

[0055] 0≤P gen,PV (t)≤P max,PV (t) (22)

[0056] (4) Pumped storage constrained by slope rate

[0057] Energy storage range:

[0058] E min,pumped-storage ≤E pumped-storage (t)≤E max,pumped-storage (twenty three)

[0059] Slope rate:

[0060] |P pumped-storage (t)-P pumped-storage (t-1)|≤ΔP max,pumped-storage(twenty four)

[0061] In the formula, E min,pumped-storag It is the minimum energy storage capacity of pumped hydro storage, E pumped-storage (t) represents the pumped hydro storage energy stored in time period t, E max,pumped-storage It is the maximum energy storage capacity of pumped hydro storage, P pumped-storage (t) represents the pumped-storage power generation or discharge during time period t, ΔP max,pumped-storage This is the maximum permissible slope rate for pumped storage hydroelectric power generation.

[0062] (5) Molten salt energy storage constraints:

[0063]

[0064] In the formula, p charge,molten-salt (t): Thermal power of molten salt at time t, p discharge,molten-saltdisc (t): Heat release power of molten salt at time t, σ molten-salt : Self-loss rate of molten salt; η discharge,molten-sal : The exothermic efficiency of molten salt, n charge,molten-sal : Heat charging efficiency of molten salt; Q molten-salt (t): Thermal energy stored in the molten salt at time t, η molten-salt The efficiency of converting the thermal energy of molten salt into electrical energy;

[0065] E min,molten-salt ≤E molten-salt (t)≤E max,molten-salt (27)

[0066] |P molten-salt (t)-P molten-salt (t-1)|≤ΔP max,molten-salt (28)

[0067] In the formula, E min,molten-sal It is the minimum energy storage capacity for molten salt energy storage, E molten-salt (t) represents the energy stored in the molten salt during time interval t, E max,molten-salt It is the maximum energy storage capacity of molten salt energy storage, P molten-salt (t) represents the amount of electricity generated or discharged by the molten salt energy storage during time period t, ΔP max,molten-salt This is the maximum allowable ramp rate for molten salt energy storage;

[0068] (6) The same form of energy cannot be started and stopped simultaneously at the same time t. This is expressed by defining a constraint:

[0069] Regarding nuclear energy:

[0070] start nuclear,t +stop nuclear,t ≤1 (29)

[0071] Regarding photovoltaic energy:

[0072] start PV,t +stop PV,t ≤1 (30)

[0073] For molten salt energy:

[0074] start molten-salt,t +stop molten-salt,t ≤1 (31)

[0075] For pumped hydro storage:

[0076] start pumped-storage,t +stop pumped-storage,t ≤1 (32)

[0077] In the formula, start x,t This indicates that the operation of starting energy form x begins at time t, while stop... x,t This indicates that the operation of stopping energy form x begins at time t. Both of these are 0 or 1 variables. These constraints ensure that at any point in time, each energy form will not be started and stopped simultaneously.

[0078] Step 4: Implement intraday rolling scheduling and real-time scheduling strategies. Then, with the objective function of minimizing the rolling operating cost of the integrated energy system considering energy storage change penalties, dynamically adjust the day-ahead plan based on real-time data and forecast results. Intraday optimization is based on day-ahead optimization, with the objective function of operating cost and power change penalty cost. Optimize scheduling within the rolling time domain, which is set to 4 hours. Each rolling cycle is optimized in 15-minute units, and only the scheduling strategy for the upcoming 15 minutes is fixed and implemented.

[0079]

[0080] In the formula, C op (t) is the operating cost in the rolling time domain, including the power generation cost, fuel cost and start-up cost of all equipment, and T is the total number of optimization periods;

[0081]

[0082] In the formula, C penalty,total (t) represents the penalty cost for the total power change at time t, C penalty,nuclear (t), C penalty,PV (t), C penalty,pumped-storage (t), C penalty,molten-salt (t) represents the power change penalty cost of nuclear power, photovoltaic, pumped storage, and molten salt energy storage equipment at time t, respectively.

[0083] (1) Power variation cost penalty function for nuclear power equipment:

[0084] C penalty,nuclear (t)=k nuclear ·(P gen,nuclear (t)-P gen,nuclear (t-1)) 2 (35)

[0085] In the formula, P gen,nuclear (t) refers to the amount of electricity generated by the nuclear power plant at time t, k nuclear It is the penalty factor for power variation in nuclear power equipment.

[0086] (2) Power variation cost penalty function for photovoltaic equipment:

[0087] C penalty,PV (t)=k PV ·(P gen,PV (t)-P gen,PV (t-1)) 2 (36)

[0088] In the formula, P gen,PV (t) refers to the power generation of the photovoltaic device at time t, k PV It is the penalty coefficient for power variation in photovoltaic equipment.

[0089] (3) Power change cost penalty function for pumped storage equipment:

[0090] C penalty,pumped-storage (t)=k pumped-storage ·(P pumped-storage (t)-P pumped-storage (t-1)) 2 (37)

[0091] In the formula, P pumped-storage (t) refers to the power generation of the pumped storage hydroelectric power station at time t, k pumped-storage It is the power change penalty coefficient of pumped storage equipment.

[0092] (4) Power variation cost penalty function for molten salt energy storage equipment

[0093] C penalty,molten-salt (t)=k molten-salt ·(P molten-salt (t)-P molten-salt (t-1)) 2 (38)

[0094] In the formula, P molten-salt (t) refers to the power generation of the pumped storage hydroelectric power station at time t, k molten-salt It is the power change penalty coefficient of pumped storage equipment.

[0095] Intraday scheduling also needs to meet the constraints of equipment power, and the same as the previous day; the real-time scheduling strategy is further fine-tuned on the basis of intraday rolling optimization, using a time granularity of 5 minutes to carefully correct and update the results of intraday optimization. The objective function of real-time scheduling focuses on minimizing the sum of power fluctuation rates of all energy equipment, thereby achieving a stable and efficient energy supply and ensuring the stability and reliability of the power system.

[0096] The overall objective function for real-time optimization can be expressed as:

[0097]

[0098] In the formula, F real-time-opt The objective function represents the real-time optimization, N represents the total number of time periods considered, α1 is the equipment power adjustment coefficient, and ΔP gen,nuclear,i P represents the change in nuclear power output during time period i. max,nuclear,i ΔP represents the maximum power variation capacity of nuclear power. gen,pumped-storage,i P represents the power change of pumped storage in time period i. max,pumped-storage,i ΔP represents the maximum power change capacity of pumped storage. gen,molten-salt,i P represents the power change of molten salt energy storage in time period i. max,molten-salt,i This represents the maximum power variation capacity of molten salt energy storage. The purpose of this function is to minimize the total power variation rate of nuclear power, pumped hydro storage, and molten salt energy storage devices during real-time operation, in order to reduce power system fluctuations and maintain a stable energy supply.

[0099] The beneficial effects achieved by this invention are as follows:

[0100] This invention utilizes deep learning algorithms, such as Transformer and CNN-LSTM, to achieve high-precision prediction of electricity load and photovoltaic power generation. This accurate predictive capability enables energy systems to respond more effectively to changes in demand, thereby improving overall dispatch efficiency and resource utilization.

[0101] By implementing day-ahead, intraday, and real-time multi-timescale scheduling, this invention overcomes the limitations of traditional single-timescale optimization methods. This multi-level optimization strategy can more comprehensively address the dynamic changes in energy demand and the uncertainties in market conditions, thereby enhancing the stability and reliability of the overall energy system.

[0102] In the current stage, this invention optimizes the operation schedule of various energy devices through forecasting and planning, effectively reducing unnecessary operating costs. This advance planning ensures the rational allocation of resources, while reducing energy waste and improving the economic efficiency of the system.

[0103] Through intraday rolling optimization and real-time scheduling, this invention can promptly respond to deviations and emergencies in actual operation, thereby reducing frequent fluctuations in equipment power. This flexible adjustment mechanism helps protect equipment, extend its service life, and maintain the stable operation of the power system. Attached Figure Description

[0104] Figure 1 This is a simplified diagram of the multi-timescale scheduling method for a nuclear-based integrated energy system according to the present invention.

[0105] Figure 2 This invention is a multi-timescale scheduling framework for a nuclear-powered integrated energy system.

[0106] Figure 3 This is a flowchart of the different stages of the multi-timescale optimization scheduling strategy of this invention. Detailed Implementation

[0107] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0108] A multi-timescale optimization scheduling method for a nuclear energy-based integrated energy system includes the following steps:

[0109] Step 1: By collecting historical power load, photovoltaic power generation data, and relevant meteorological information, the foundation for training the deep learning prediction model is laid. Deep learning techniques, including Transformer and CNN-LSTM algorithms, are used for prediction, and a parallel architecture is adopted to enable the two algorithms to run simultaneously, thereby significantly improving computational efficiency.

[0110] Step 2: Apply the Gray Wolf Optimization Algorithm to intelligently adjust the weights of the Transformer and CNN-LSTM prediction results. This step aims to find the optimal solution, i.e., the best weight combination, to reduce prediction error and improve the overall reliability and accuracy of the prediction results. By defining an objective function with prediction error as the core, and using the Gray Wolf Optimization Algorithm to find the weight configuration that minimizes this function, the best overall prediction accuracy is achieved.

[0111] Step 3: Construct a day-ahead scheduling optimization model with a time scale of 1 hour. Based on day-ahead forecasts of photovoltaic and electricity loads, the model comprehensively considers the operational limitations and constraints of each device in the integrated energy system. Furthermore, the model specifically considers the stable output of nuclear power, the intermittency of photovoltaic power, and the peak-shaving capabilities of molten salt hydroelectric power and pumped storage, ensuring the overall efficiency and cost-effectiveness of the system.

[0112] Step 4: Implement intraday rolling dispatch and real-time dispatch strategies. This step considers the cost of power change penalties and dynamically adjusts the day-ahead schedule based on real-time data and forecast results. This strategy ensures the flexibility and responsiveness of the power system at both the intraday and real-time levels, further improving the overall operating efficiency and reliability of the system.

[0113] As nuclear energy's share in the global power industry steadily increases, the need for optimizing integrated energy systems centered on nuclear power is becoming increasingly urgent. Traditionally, power industry system optimization has focused on conventional energy sources such as thermal power plants and gas turbines. However, nuclear energy must now be considered to fully leverage its low-carbon and high-efficiency characteristics. In-depth research into integrated energy systems based on nuclear energy means taking into account the stability and safety of nuclear power plants, as well as their synergistic effects with other renewable energy sources such as solar power. Figure 1 and Figure 2 As shown, the present invention provides a multi-timescale scheduling method for a nuclear energy-based integrated energy system, which includes the following steps:

[0114] Step 1: By collecting historical electricity load, photovoltaic power generation data, and related meteorological information, rich training data is provided for the model, helping to capture the complex relationships between these variables and laying the foundation for training the deep learning prediction model. Deep learning techniques, including Transformer and CNN-LSTM algorithms, are used for prediction. The Transformer algorithm and CNN-LSTM capture patterns in time series data through self-attention mechanisms and convolutional layers, which helps improve prediction accuracy. The Transformer's self-attention mechanism can focus on all locations in the input data, thus capturing global dependencies. CNN-LSTM combines the spatial feature extraction capabilities of CNNs with the long-term dependency capture capabilities of LSTMs in time series analysis. A parallel architecture is adopted, allowing the two algorithms to run simultaneously, significantly improving computational efficiency.

[0115] The Transformer algorithm is primarily based on a self-attention mechanism, and its core formula can be summarized as follows:

[0116] (1) Calculation of self-attention:

[0117]

[0118] In the formula, Q, K, and V are the query, key, and value matrices, respectively, and d K is the dimension of the key vector.

[0119] (2) Multi-Head Attention:

[0120] MultiHead(Q,K,V)=Concat(head1,...,head h W O (2)

[0121]

[0122] In the formula, i represents the number of multi-head attention mechanisms; and It is a learnable weight matrix.

[0123] CNN-LSTM combines the features of Convolutional Neural Networks (CNN) and Long Short-Term Memory Networks (LSTM). Its core formula can be summarized as follows:

[0124] (1) Convolution operation in CNN:

[0125] f(x)=x*w+b (4)

[0126] In the formula, * denotes the convolution operation, w is the convolution kernel, and b is the bias term.

[0127] (2) LSTM section:

[0128] Forgotten Gate: f t =σ(W f ·[h t-1 ,x t ]+b f (5)

[0129] Input gate: i t =σ(W i ·[h t-1 ,x t ]+b i (6)

[0130] Output gate: o t =σ(W o ·[h t-1 ,x t ]+b o (7)

[0131] New memory unit:

[0132] Final memory unit:

[0133] Final hidden state: h t =o t *tanh(C t(10)

[0134] In the formula, σ represents the sigmoid activation function, which compresses the input to between 0 and 1, representing the proportion of information retained; W f This is the weight matrix of the forget gate; h t-1 It is the hidden state of the previous time step; x t This is the input for the current time step; b f It is the bias term of the forget gate.

[0135] Step 2: Apply the Grey Wolf Optimization (GWO) algorithm to intelligently adjust the weights of the Transformer and CNN-LSTM prediction results. By finding the optimal solution (i.e., the best weight combination), prediction errors are reduced, and the reliability and accuracy of the overall prediction results are improved. This process involves adjusting the relative weights of the prediction results of the two algorithms to achieve the best overall prediction accuracy. The core of this process is to define an objective function (a function of prediction error) and then use GWO to find the weight configuration that minimizes this objective function.

[0136] It generally includes the following steps:

[0137] (1) Comprehensive prediction model

[0138] Y pred =ω T ·Y T +ω CNN-LSTM ·Y CNN-LSTM (11)

[0139] In the formula, Y pred This is a comprehensive prediction result; Y T This is the prediction result of the Transformer model; Y CNN-LSTM This is the prediction result of the CNN-LSTM model; ω T and ω CNN-LSTM These are the weights of the two models, optimized by the GWO algorithm.

[0140] (2) Objective function

[0141] Minimize F(ω T ,ω CNN-LSTM (12)

[0142] In the formula, F is an objective function, usually a function of the prediction error (such as mean squared error). The goal of the GWO algorithm is to find the weights ω that minimize F. T and ω CNN-LSTM The GWO algorithm continuously updates the weights ω through an iterative process. T and ω CNN-LSTM In order to find the minimum value of the objective function.

[0143] Step 3: Figure 3 The flowchart shown illustrates the day-ahead scheduling, intraday rolling, and real-time adjustment phases. First, a day-ahead scheduling optimization model is established with a time scale of 1 hour, based on day-ahead forecasts of photovoltaic and power loads, while also considering the operational limitations and constraints of various devices within the integrated energy system. Based on these parameters, the model considers the stable output of nuclear power, the intermittency of photovoltaic power, and the peak-shaving capabilities of molten salt hydroelectric power and pumped storage to ensure the overall efficiency and cost-effectiveness of the system's operation.

[0144]

[0145] In the formula, C pumped-storage (t) represents the pumped storage cost in time period t, C molten-salt (t) represents the cost of molten salt energy storage in time period t, C nuclear (t) represents the nuclear power cost in time period t, C PV (t) represents the cost of photovoltaic power, C start-stop (t) is the start-up and shutdown cost, which mainly refers to the additional costs incurred when starting or shutting down the equipment. This usually includes additional wear and tear, energy consumption, etc., and t = 24.

[0146] Nuclear power cost:

[0147] C nuclear (t)=C variable,nuclear ·P gen,nuclear (t) (14)

[0148] In the formula, C variable,nuclear This refers to the unit cost of nuclear power generation, including the initial investment, the converted unit cost over the service life, fuel costs, and unit operation and maintenance costs; P gen,nuclear (t) represents the amount of nuclear power generated during time period t.

[0149] Photovoltaic cost:

[0150] C PV (t)=C variable,PV ·P gen,PV (t) (15)

[0151] In the formula, C variable,PV This refers to the unit cost of electricity generated by a photovoltaic system, including the initial investment, the converted unit cost over the service life, and the unit operation and maintenance cost, P. gen,PV (t) represents the photovoltaic power generation during time period t.

[0152] Molten salt cost:

[0153] C molten-salt (t)=C variable,molten-salt ·E molten-salt (t) (16)

[0154] In the formula, C variable,molten-salt It is the unit cost of electricity generation, including the initial investment converted to unit cost over the service life, and the cost of energy conversion and storage. molten-salt (t) represents the change in energy storage during time period t.

[0155] Pumped storage cost:

[0156] C pumped-storage (t)=C variable,pumped-storage ·E pumped-storage (t) (17)

[0157] In the formula, C variable,pumped-storage It is the unit cost of electricity generation, including the initial investment converted to a unit cost over the service life, and the cost of energy conversion and storage. pumped-storage (t) represents the change in energy storage during time period t.

[0158] Start-up and shutdown costs:

[0159]

[0160] In the formula, C start-stop It is the total start-up and shutdown cost of all equipment, C start-stop,nuclear The cost of a single start-up and shutdown of nuclear power equipment, C start-stop,PV The start-up and shutdown cost of photovoltaic equipment, C start-stop,molten-salt The single start-up and shutdown cost of molten salt energy storage equipment, C start-stop,pumped-storage The single start-up and shutdown cost of pumped storage equipment, S nuclear S PV S molten-salt S pumped-storage These refer to the number of start-ups and shutdowns of nuclear power, photovoltaic, molten salt energy storage, and pumped storage equipment within the considered time period.

[0161] The constraints for optimizing various cost ratios are as follows:

[0162] (1) Electricity balance constraint:

[0163]

[0164] In the formula, P gen,PV P(t) is the photovoltaic power generation at time t. discharge,molten-salt P(t) is the power of the molten salt storing heat and releasing it into electrical energy at time t. gen,pumped-storage (t) represents the pumped-storage power generation at time t, P load (t) represents the total power of the user load at time t, P charge,molten-salt P(t) is the power of the molten salt energy storage and heat absorption at time t. charge,pumped-storage(t) is the power at which the pumped hydro storage begins to store energy at time t.

[0165] (2) Nuclear power constraints (including ramp rate)

[0166] Power generation range:

[0167] P min,nuclear ≤P gen,nuclear (t)≤P max,nuclear (20)

[0168] Slope rate:

[0169] |P gen,nuclear (t)-P gen,nuclear (t-1)|≤ΔP max,nuclear (twenty one)

[0170] In the formula, P min,nuclear It is the minimum power generation capacity of a nuclear power plant, P gen,nuclear (t) represents the amount of electricity generated by nuclear power at time t, P max,nuclear It is the maximum power generation of the nuclear power plant, ΔP max,nuclear It is the maximum permissible ramp rate for a nuclear power plant.

[0171] (3) Constraints on photovoltaic power generation (weather has a greater impact, and slope rate constraints are usually not applicable)

[0172] Power generation is affected by sunlight conditions, typically as follows:

[0173] 0≤P gen,PV (t)≤P max,PV (t) (22)

[0174] (4) Pumped storage constraints (including slope ratio)

[0175] Energy storage range:

[0176] E min,pumped-storage ≤E pumped-storage (t)≤E max,pumped-storage (twenty three)

[0177] Slope rate:

[0178] |P pumped-storage (t)-P pumped-storage (t-1)|≤ΔP max,pumped-storage (twenty four)

[0179] In the formula, E min,pumped-storag It is the minimum energy storage capacity of pumped hydro storage, E pumped-storage (t) represents the pumped hydro storage energy stored in time period t, E max,pumped-storage It is the maximum energy storage capacity of pumped hydro storage, P pumped-storage (t) represents the pumped-storage power generation or discharge during time period t, ΔPmax,pumped-storage It is the maximum permissible slope rate for pumped storage.

[0180] (5) Molten salt energy storage constraints:

[0181]

[0182] E molten-salt (t)=Q molten-saltt (t).η molten-salt (26)

[0183] In the formula, p charge,molten-salt (t): Thermal power of molten salt at time t, p discharge,molten-saltdisc (t): Heat release power of molten salt at time t, σ molten-salt : Self-loss rate of molten salt; η discharge,molten-sal : The exothermic efficiency of molten salt, n charge,molten-sal : Heat charging efficiency of molten salt; Q molten-salt (t): Thermal energy stored in the molten salt at time t, η molten-salt The efficiency of converting the thermal energy of molten salt into electrical energy;

[0184] E min,molten-salt ≤E molten-salt (t)≤E max,molten-salt (27)

[0185] |P molten-salt (t)-P molten-salt (t-1)|≤ΔP max,molten-salt (28)

[0186] In the formula, E min,molten-sal It is the minimum energy storage capacity for molten salt energy storage, E molten-salt (t) represents the energy stored in the molten salt during time interval t, E max,molten-salt It is the maximum energy storage capacity of molten salt energy storage, P molten-salt (t) represents the amount of electricity generated or discharged by the molten salt energy storage during time period t, ΔP max,molten-salt It is the maximum permissible ramp rate for molten salt energy storage.

[0187] (6) The same form of energy cannot be started and stopped at the same time t. This is expressed by defining a constraint.

[0188] Regarding nuclear energy:

[0189] start nuclear,t +stop nuclear,t ≤1 (29)

[0190] For photovoltaic (PV) energy:

[0191] start PV,t +stopPV,t ≤1 (30)

[0192] For molten salt energy:

[0193] start molten-salt,t +stop molten-salt,t ≤1 (31)

[0194] For pumped-storage energy:

[0195] start pumped-storage,t +stop pumped-storage,t ≤1 (32)

[0196] In the formula, start x,t This indicates that the operation of starting energy form x begins at time t, while stop... x,t This indicates the start and stop of the operation of energy form x at time t, and both are variables of 0 or 1. These constraints ensure that at any given time, each of the above energy forms will not be started and stopped simultaneously.

[0197] Step 4: Implement intraday rolling scheduling and real-time scheduling strategies, and then dynamically adjust the day-ahead plan based on real-time data and forecast results, with the objective function being the lowest rolling operating cost of the integrated energy system considering energy storage change penalties.

[0198] Intraday optimization builds upon day-ahead optimization, using operating costs and penalty costs for power changes as objective functions to optimize scheduling within a rolling time domain. The penalty cost for power changes is incorporated because frequent adjustments can place additional stress on the grid and generating equipment; therefore, the extra costs arising from rapid power fluctuations need to be considered. The rolling time domain is set to 4 hours, with optimization performed in 15-minute units per rolling cycle. This ensures timely adjustments and flexibility of the strategy, while only fixing and implementing the scheduling strategy for the upcoming 15 minutes. This approach effectively balances the needs of cost control and system response.

[0199]

[0200] In the formula, C op (t) represents the operating cost over the rolling time domain, including the power generation cost, fuel cost, and start-up and shutdown costs of all equipment. T is the total number of optimization periods.

[0201]

[0202] In the formula, C penalty,total (t) represents the penalty cost for the total power change at time t, C penalty,nuclear (t), C penalty,PV (t), Cpenalty,pumped-storage (t), C penalty,molten-salt (t) represents the penalty cost for power change at time t for nuclear power, photovoltaic, pumped storage, and molten salt energy storage equipment, respectively.

[0203] (1) Power variation cost penalty function for nuclear power equipment:

[0204] C penalty,nuclear (t)=k nuclear ·(P gen,nuclear (t)-P gen,nuclear (t-1)) 2 (35)

[0205] In the formula, P gen,nuclear (t) refers to the amount of electricity generated by the nuclear power plant at time t, k nuclear It is the penalty factor for power variation in nuclear power equipment.

[0206] (2) Power variation cost penalty function for photovoltaic equipment:

[0207] C penalty,PV (t)=k PV ·(P gen,PV (t)-P gen,PV (t-1)) 2 (36)

[0208] In the formula, P gen,PV (t) refers to the power generation of the photovoltaic device at time t, k PV It is the penalty coefficient for power variation in photovoltaic equipment.

[0209] (3) Power change cost penalty function for pumped storage equipment:

[0210] C penalty,pumped-storage (t)=k pumped-storage ·(P pumped-storage (t)-P pumped-storage (t-1)) 2 (37)

[0211] In the formula, P pumped-storage (t) refers to the power generation of the pumped storage hydroelectric power station at time t, k pumped-storage It is the power change penalty coefficient of pumped storage equipment.

[0212] (4) Power variation cost penalty function for molten salt energy storage equipment

[0213] C penalty,molten-salt (t)=k molten-salt ·(P molten-salt (t)-P molten-salt (t-1)) 2 (38)

[0214] In the formula, P molten-salt (t) refers to the power generation of the pumped storage hydroelectric power station at time t, k molten-salt It is the power change penalty coefficient of pumped storage equipment.

[0215] Intraday scheduling also needs to meet equipment power constraints, similar to those of the previous day. The real-time scheduling strategy further fine-tunes the intraday rolling optimization, using a 5-minute time granularity to meticulously correct and update the results of the intraday optimization. The core objective of this strategy is to ensure that the integrated energy system operates both economically and efficiently. Real-time scheduling focuses on refined management, minimizing the fluctuation rate of energy equipment output to avoid system instability. Within this framework, the objective function of real-time scheduling focuses on minimizing the sum of the power fluctuation rates of all energy equipment, thereby achieving a stable and efficient energy supply and ensuring the stability and reliability of the power system.

[0216] The overall objective function for real-time optimization can be expressed as:

[0217]

[0218] In the formula, F real-time-opt This represents the overall objective function for real-time optimization, where N represents the total number of time periods considered, 1 is the equipment power adjustment coefficient, and ΔP gen,nuclear,i P represents the change in nuclear power output during time period i. max,nuclear,i ΔP represents the maximum power variation capacity of nuclear power. gen,pumped-storage,i P represents the power change of pumped storage in time period i. max,pumped-storage,i ΔP represents the maximum power change capacity of pumped storage. gen,molten-salt,i P represents the power change of molten salt energy storage in time period i. max,molten-salt,i This represents the maximum power variation capacity of molten salt energy storage. The purpose of this function is to minimize the total power variation rate of nuclear power, pumped hydro storage, and molten salt energy storage devices during real-time operation, in order to reduce power system fluctuations and maintain a stable energy supply.

Claims

1. A multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system, characterized in that: Step 1: Collect historical power load, photovoltaic power generation data and related meteorological information to lay the foundation for training the deep learning prediction model and use deep learning algorithms for prediction; Step 2: Perform intelligent weight adjustment on the prediction results to find the optimal weight combination; Step 3: Construct a day-ahead scheduling optimization model; Step 4: Execute intraday rolling scheduling and real-time scheduling strategies.

2. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 1, characterized in that: Deep learning includes Transformer and CNN-LSTM algorithms.

3. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 2, characterized in that: A parallel architecture is used to allow the two algorithms to run simultaneously.

4. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 2, characterized in that: The Grey Wolf Optimization Algorithm is applied to intelligently adjust the weights of the prediction results of Transformer and CNN-LSTM.

5. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 4, characterized in that: By defining an objective function with prediction error as its core, and using the Grey Wolf optimization algorithm to find the weight configuration that minimizes this function, the optimal overall prediction accuracy is achieved.

6. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 1, characterized in that: A day-ahead scheduling optimization model is constructed with a time scale of 1 hour. Based on day-ahead forecasts of photovoltaic and power loads, the model takes into account the operational limitations and constraints of each device in the integrated energy system, the stable output of nuclear power, the intermittency of photovoltaic power, and the peak-shaving capabilities of molten salt and pumped storage.

7. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 1, characterized in that: Step 4: Consider the cost of power change penalty terms and dynamically adjust the day-ahead plan based on real-time data and forecast results.

8. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 2, characterized in that: Step 2: Apply the Grey Wolf Optimization (GWO) algorithm to intelligently adjust the weights of the Transformer and CNN-LSTM prediction results, find the optimal weight combination, define the objective function for the prediction error, and find the weight configuration that minimizes this function using GWO. This includes the following steps: (1) Comprehensive prediction model Y pred =ω T ·Y T +ω CNN-LSTM ·Y CNN-LSTM (11) In the formula, Y pred This is a comprehensive prediction result; Y T This is the prediction result of the Transformer model; Y CNN-LSTM This is the prediction result of the CNN-LSTM model; ω T and ω CNN-LSTM These are the weights of the two models, optimized by the GWO algorithm; (2) Objective function Minimize F(ω T ,oh CNN-LSTM ) (12) In the formula, F is the objective function of the prediction error, and the goal of the GWO algorithm is to find the weights ω that minimize F. T and ω CNN-LSTM The GWO algorithm continuously updates the weights ω through an iterative process. T and ω CNN-LSTM In order to find the minimum value of the objective function.

9. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 6, characterized in that: Step 3: First, establish a day-ahead scheduling optimization model. In the formula, C pumped-storage (t) represents the pumped storage cost in time period t, C molten-salt (t) represents the cost of molten salt energy storage in time period t, C nuclear (t) represents the nuclear power cost in time period t, C PV (t) represents the cost of photovoltaic power, C start-stop (t) is the start-up and shutdown cost, which refers to the additional costs incurred when starting or stopping the equipment, including additional wear and tear and energy consumption, t = 24; Nuclear power cost: C nuclear (t)=C variable,nuclear ·P gen,nuclear (t) (14) In the formula, C variable,nuclear This refers to the unit cost of nuclear power generation, including the initial investment, the converted unit cost over the service life, fuel costs, and unit operation and maintenance costs; P gen,nuclear (t) represents the amount of nuclear power generated during time period t; Photovoltaic cost: C PV (t)=C variable,PV ·P gen,PV (t) (15) In the formula, C variable,PV This refers to the unit cost of electricity generated by a photovoltaic system, including the initial investment, the converted unit cost over the service life, and the unit operation and maintenance cost, P. gen,PV (t) represents the photovoltaic power generation during time period t; Molten salt cost: C molten-salt (t)=C variable,molten-salt ·E molten-salt (t) (16) In the formula, C variable,molten-salt It is the unit cost of electricity generation, including the initial investment converted to a unit cost over the service life, and the cost of energy conversion and storage. molten-salt (t) represents the change in energy storage over time period t; Pumped storage cost: C pumped-storage (t)=C variable,pumped-storage ·E pumped-storage (t) (17) In the formula, C variable,pumped-storage It is the unit cost of electricity generation, including the initial investment converted to a unit cost over the service life, and the cost of energy conversion and storage. pumped-storage (t) represents the change in energy storage over time period t; Start-up and shutdown costs: C start-stop =C start-stop,nuclear ·S nuclear +C start-stop,PV ·S PV +C start-stop,molten-salt ·S molten-salt +C start-stop,pumped-storage ·S pumped-storage (18) In the formula, C start-stop It is the total start-up and shutdown cost of all equipment, C start-stop,nuclear The cost of a single start-up and shutdown of nuclear power equipment, C start-stop,PV The start-up and shutdown cost of photovoltaic equipment, C start-stop,molten-salt The single start-up and shutdown cost of molten salt energy storage equipment, C start-stop,pumped-storage The single start-up and shutdown cost of pumped storage equipment, S nuclear S PV S molten-salt S pumped-storage These are the number of start-ups and shutdowns of nuclear power, photovoltaic, molten salt energy storage, and pumped storage equipment within the considered time period; The constraints for optimizing various cost ratios are as follows: (1) Power balance constraint In the formula, P gen,PV P(t) is the photovoltaic power generation at time t. discharge,molten-salt P(t) is the power of the molten salt storing heat and releasing it into electrical energy at time t. gen,pumped-storage (t) represents the pumped-storage power generation at time t, P load (t) represents the total power of the user load at time t, P charge,molten-salt P(t) is the power of the molten salt energy storage and heat absorption at time t. charge,pumped-storage (t) is the power that the pumped hydro storage begins to store at time t; (2) Nuclear power constraint including ramp rate Power generation range: P min,nuclear ≤P gen,nuclear (t)≤P max,nuclear (20) Slope rate: |P gen,nuclear (t)-P gen,nuclear (t-1)|≤ΔP max,nuclear (21) In the formula, P min,nuclear It is the minimum power generation capacity of a nuclear power plant, P gen,nuclear (t) represents the amount of electricity generated by nuclear power at time t, P max,nuclear It is the maximum power generation of the nuclear power plant, ΔP max,nuclear This is the maximum permissible ramp rate for a nuclear power plant; (3) Constraints on photovoltaic power generation Power generation is affected by sunlight conditions: 0≤P gen,PV (t)≤P max,PV (t) (22) (4) Pumped storage constrained by slope rate Energy storage range: E min,pumped-storage ≤E pumped-storage (t)≤E max,pumped-storage (23) Slope rate: |P pumped-storage (t)-P pumped-storage (t-1)|≤ΔP max,pumped-storage (24) In the formula, E min,pumped-storag It is the minimum energy storage capacity of pumped hydro storage, E pumped-storage (t) represents the pumped hydro storage energy stored in time period t, E max,pumpe-d stora It is the maximum energy storage capacity of pumped hydro storage, P pumped-storage (t) represents the pumped-storage power generation or discharge during time period t, ΔP max,pumped-storage This is the maximum permissible slope rate for pumped storage hydroelectric power generation. (5) Molten salt energy storage constraints: In the formula, p charge,molten-salt (t): The heat charge power of the molten salt at time t. The heat release power of the molten salt at time t, σ molten-salt : Self-loss rate of molten salt; η discharge,molten-sal : The exothermic efficiency of molten salt, n charge,molten-sal : Heat charging efficiency of molten salt; Q molten-salt (t): Thermal energy stored in the molten salt at time t, η molten-salt The efficiency of converting the thermal energy of molten salt into electrical energy; E min,molten-salt ≤E molten-salt (t)≤E max,molten-salt (27) |P molten-salt (t)-P molten-salt (t-1)|≤ΔP max,molten-salt (28) In the formula, E min,molten-sal It is the minimum energy storage capacity for molten salt energy storage, E molten-salt (t) represents the energy stored in the molten salt during time interval t, E max,molten-salt It is the maximum energy storage capacity of molten salt energy storage, P molten-salt (t) represents the amount of electricity generated or discharged by the molten salt energy storage during time period t, ΔP max,molten-salt This is the maximum allowable ramp rate for molten salt energy storage; (6) The same form of energy cannot be started and stopped simultaneously at the same time t. This is expressed by defining a constraint: Regarding nuclear energy: start nuclear,t +stop nuclear,t ≤1 (29) Regarding photovoltaic energy: start PV,t +stop PV,t ≤1 (30) For molten salt energy: start molten-salt,t +stop molten-salt,t ≤1 (31) For pumped hydro storage: start pumped-storage,t +stop pumped-storage,t ≤1 (32) In the formula, start x,t This indicates that the operation of starting energy form x begins at time t, while stop... x,t This indicates that the operation of stopping energy form x begins at time t. Both of these are 0 or 1 variables. These constraints ensure that at any point in time, each energy form will not be started and stopped simultaneously.

10. The multi-timescale optimal scheduling method for a nuclear energy-based integrated energy system according to claim 1, characterized in that: Step 4: Implement intraday rolling scheduling and real-time scheduling strategies, and then dynamically adjust the day-ahead plan based on real-time data and forecast results, with the objective function being the lowest rolling operating cost of the integrated energy system considering energy storage change penalties. Intraday optimization builds upon day-ahead optimization, using operating costs and penalty costs for power changes as objective functions. It optimizes scheduling within a rolling time domain, which is set to 4 hours. Each rolling cycle is optimized in 15-minute units, with the scheduling strategy for the upcoming 15 minutes fixed and implemented only. In the formula, C op (t) is the operating cost in the rolling time domain, including the power generation cost, fuel cost and start-up cost of all equipment, and T is the total number of optimization periods; In the formula, C penalty,total (t) represents the penalty cost for the total power change at time t, C penalty,nuclear (t), C penalty,PV (t), C penalty,pumped-storage (t), C penalty,molten-salt (t) represents the power change penalty cost of nuclear power, photovoltaic, pumped storage, and molten salt energy storage equipment at time t, respectively. (1) Power variation cost penalty function for nuclear power equipment: C penalty,nuclear (t)=k nuclear ·(P gen,nuclear (t)-P gen,nuclear (t-1)) 2 (35) In the formula, P gen,nuclear (t) refers to the amount of electricity generated by the nuclear power plant at time t, k nuclear It is the penalty factor for power variation in nuclear power equipment. (2) Power variation cost penalty function for photovoltaic equipment: C penalty,PV (t)=k PV ·(P gen,PV (t)-P gen,PV (t-1)) 2 (36) In the formula, P gen,PV (t) refers to the power generation of the photovoltaic device at time t, k PV It is the penalty coefficient for power variation in photovoltaic equipment. (3) Power change cost penalty function for pumped storage equipment: C penalty,pumped-storage (t)=k pumped-storage ·(P pumped-storage (t)-P pumped-storage (t-1)) 2 (37) In the formula, P pumped-storage (t) refers to the power generation of the pumped storage hydroelectric power station at time t, k pumped-storage It is the power change penalty coefficient of pumped storage equipment. (4) Power variation cost penalty function for molten salt energy storage equipment C penalty,molten-salt (t)=k molten-salt ·(P molten-salt (t)-P molten-salt (t-1)) 2 (38) In the formula, P molten-salt (t) refers to the power generation of the pumped storage hydroelectric power station at time t, k molten-salt It is the power change penalty coefficient of pumped storage equipment. Intraday scheduling also needs to meet the constraints of equipment power, and the same as the previous day; the real-time scheduling strategy is further fine-tuned on the basis of intraday rolling optimization, using a time granularity of 5 minutes to carefully correct and update the results of intraday optimization. The objective function of real-time scheduling focuses on minimizing the sum of power fluctuation rates of all energy equipment, thereby achieving a stable and efficient energy supply and ensuring the stability and reliability of the power system. The overall objective function for real-time optimization can be expressed as: In the formula, F real-time-opt The objective function represents the real-time optimization, N represents the total number of time periods considered, α1 is the equipment power adjustment coefficient, and ΔP gen,nuclear,i P represents the change in nuclear power output during time period i. max,nuclear,i ΔP represents the maximum power variation capacity of nuclear power. gen,pumped-storage,i P represents the power change of pumped storage in time period i. max,pumped-storage,i ΔP represents the maximum power change capacity of pumped storage. gen,molten-salt,i P represents the power change of molten salt energy storage in time period i. max,molten-salt,i This represents the maximum power variation capacity of molten salt energy storage. The purpose of this function is to minimize the total power variation rate of nuclear power, pumped hydro storage, and molten salt energy storage devices during real-time operation, in order to reduce power system fluctuations and maintain a stable energy supply.

Images