Multi-time-scale composite energy storage system and energy-saving economic operation method

Abstract

The invention discloses a multi-time-scale composite energy storage system and an energy-saving economic operation method, relates to the technical field of energy storage, and aims to accurately calculate user-side energy consumption resources required by energy storage and energy supply through accurate load prediction calculation in combination with data such as user-side electric energy and heat energy consumption, so that the planning scientificity of the energy storage system is remarkably improved. Besides, the electrochemical energy storage system and the fused salt energy storage system are finely controlled through the intelligent management platform, multi-target optimization control is combined, the output proportion of each energy storage unit is dynamically distributed according to various parameters, the operation stability, high efficiency and energy-saving economy of the energy storage system are greatly improved, stable and reliable operation of the regional comprehensive energy supply system is ensured, and the economic benefit of the regional comprehensive energy supply system is improved. The multi-time-scale composite energy storage system is constructed, effective response to diversity of power grid events is realized, cooperative work is carried out for power grid events with different frequencies, the adaptive capacity to complex power grid working conditions is greatly improved, and the stability and reliability of energy storage supply are powerfully guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage, and specifically, to a multi-time scale composite energy storage system and an energy-saving and economical operation method. Background Art

[0002] In a regional integrated energy supply system, composite energy storage is crucial. It can cope with the instability and volatility of renewable energy, as well as the frequent changes in user loads. Single energy storage is difficult to meet diverse requirements and grid emergencies, while composite energy storage can, through multi-time scale coordination, deeply grasp the dynamic variable operating conditions characteristics, ensure the efficient, reliable, and economical operation of the system, and promote the sustainable development of the energy field.

[0003] In the process of constructing composite energy storage, many factors have become key problems hindering the accurate calculation of capacity. Renewable energy itself is unstable and shows strong volatility with seasons. For example, solar energy changes due to day and night and weather, and wind power generation is randomly affected by wind speed and direction, making the output power difficult to be constant. User loads are disturbed by multiple factors such as production and life and change constantly. These factors are intertwined with each other, making the interaction between renewable energy, multiple energy storages, and user demands become intricate, making it difficult to accurately calculate the required capacity during the construction of composite energy storage, which severely restricts the scientific construction and efficient operation of the composite energy storage system. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-time scale composite energy storage system and an energy-saving and economical operation method to solve the problems raised in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: A multi-time scale composite energy storage system, the multi-time scale composite energy storage system includes a short-time energy storage module, a medium-time energy storage module, and a long-time energy storage module:

[0006] The short-time energy storage module is used to handle the inertial characteristic changes of the power grid from seconds to minutes, including but not limited to lithium-ion batteries, supercapacitors, and flywheel energy storage. The inertial characteristic changes are expressed as the frequency fluctuation response characteristics caused by the dynamic imbalance between the power generation power and the load power within a specific time; the medium-time energy storage module is used to handle the inertial characteristic changes of the power grid from minutes to hours, including but not limited to vanadium redox flow batteries and lead-acid batteries; the long-time energy storage module is used to handle the inertial characteristic changes of the power grid from hours to days, including but not limited to compressed air energy storage, molten salt heat storage, and ice thermal storage. The input and output of the short-time energy storage module are both electrical energy; the input and output of the medium-time energy storage module are both electrical energy; the input of the long-time energy storage module includes but not limited to electrical energy, light energy, and waste heat, and the output of the long-time energy storage module includes but not limited to electrical energy, cold energy, and heat energy.

[0007] The types, advantages and disadvantages, and applicable scenarios of the multi-time scale composite energy storage system are as follows:

[0008]

[0009]

[0010] Furthermore, the multi-time scale composite energy storage system further includes an electrochemical energy storage unit and a molten salt energy storage unit;

[0011] The electrochemical energy storage unit is used for storing and releasing electric energy;

[0012] The molten salt energy storage unit is used for storing and converting thermal energy.

[0013] Furthermore, the energy-saving and economic operation method includes the following steps:

[0014] Step S1: Perform predictive calculation on the load according to the similarity function;

[0015] Step S1-1: Taking m as the time step and starting from time 0 o'clock, divide the time points of 0, 0 + m, 0 + 2m…, 24h every day to construct a historical load database;

[0016] Step S1-2: Establish a fitting relationship between the load at the time point to be analyzed and the loads at each time point before the time point to be analyzed based on the historical load database and the statistical analysis algorithm: X(t)1, X(t)2, …… X(t) n , where X(t) represents the load function at time t, and 1, 2……n represent each day. The accuracy of the prediction model is adjusted by adjusting the value of the time step m. Among them, the smaller the value of the time step m, the greater the accuracy of the prediction model;

[0017] Step S1-3: Perform dynamic prediction on the load based on function similarity. The dynamic prediction calculation formula is as follows:

[0018] Y(t) = a1X(t)1, a2X(t)2, …… a n X(t) n + b;

[0019] In the formula, Y(t) represents the dynamically predicted load value; a1, a2……a n represent the weights of the daily historical loads; n represents the number of days participating in the dynamic load prediction calculation in the historical load database; b represents the offset;

[0020] Step S1-4: Set the load variance threshold S max , sort according to the values of the load variances from large to small, and arrange the daily historical loads a1, a2……a nThe weight is set to 1 / n; set the parameter R = 1 / m; adjust the weights of the daily historical loads a1, a2... a n through the parameter A, and the adjustment process is as follows:

[0021] Add the parameter R to the maximum value of the weight, subtract the parameter R from the minimum value of the weight, and finally make the value of the load variance less than the set S max to end the weight adjustment.

[0022] Step S2: Calculate the user-side energy resources required for energy storage supply according to the predicted calculation results of the load in combination with the user-side energy consumption data, where the user-side energy consumption data includes user-side electricity and user-side heat energy;

[0023] Step S2-1: Calculate the input amount of energy storage supply resources. The calculation of the input amount of energy storage supply resources uses the following formula:

[0024]

[0025] In the formula, c represents the input amount of energy storage device resources corresponding to the unit charge-discharge amount; M represents the initial input resources of the energy storage supply device; N represents the resources required for the maintenance of each year during the operation of the energy storage device; n represents the service life of the energy storage device; i represents each year, i = 1, 2,..., n; r is the resource benefit depreciation rate; P represents the remaining available resources of the energy storage device after the end of the service life; A represents the total charge-discharge amount of the energy storage device during the whole life cycle;

[0026] Step S2-2: Determine the energy storage capacity of the energy storage power station according to the input amount of energy storage supply resources in combination with the predicted load.

[0027] Step S2-2-1: Calculate the user-side energy resources, and use the following formula for calculation:

[0028] P = E1×p1 + E2×p2 + (E3 + E4))×ε1×c;

[0029] In the formula, P is the user-side energy resources; E1 is the supply amount of municipal energy; E2 is the direct production energy supply amount of distributed energy, E3 is the energy stored in the energy storage device from municipal energy, E4 is the energy stored in the energy storage device from distributed energy, ε1 is the energy conversion efficiency of the energy storage device; p1 is the proportion of municipal energy resources; p2 is the input resources amount converted by the unit production energy of the distributed energy direct production equipment; c is the input resources amount converted by the unit production energy of the energy storage device;

[0030] Step S2-2-2: Calculate the first user-side energy consumption E now according to the user-side energy consumption historical data, and use the following formula for calculation:

[0031] E now= E1 + E2 + (E3 + E4) × ε1;

[0032] Step S2-2-3: Calculate the total user-side energy consumption E based on the predicted load Y(t).

[0033]

[0034] In the formula, e1 represents the weight of the first user-side energy consumption E now ; e2 represents the weight of the user-side energy consumption calculated based on the predicted load.

[0035] Step S2-3: Screen out the minimum value of the user-side energy resource P to determine the values of E1, E2, E3, and E4. Obtain a matrix from the change amount E, and determine the values of E1, E2, E3, and E4 by combining weighted average. Calculate using the following formula:

[0036] E1 = a1 * E 11 + a2 * E 12 +......+ a n * E ln ;

[0037] E2 = b1 * E 21 + b2 * E 22 +......+ b n * E 2n ;

[0038] E3 = c1 * E 31 + c2 * E 32 +......+ c n * E 3n ;

[0039] E4 = d1 * E 41 + d2 * E 42 +......+ d n * E 4n ;

[0040] In the formula, E 11 、E 12 ...E 1n refer to the numerical value of E1, a1, a2……a n are the frequencies of occurrence of E 11 、E 12 ……E 1n respectively; E 21 、E 22 ...E 2n refer to the numerical value of E2, b1, b2……b n are the frequencies of occurrence of E 21 、E 22 ……E2n The frequency of occurrence; E 31 、E 32 ...E 3n Refers to the numerical value of E3, c1, c2... c n Are respectively E 31 、E 32 ……E 3n The frequency of occurrence; E 41 、E 42 ...E 4n Refers to the numerical value of E4, d1, d2... d n Are respectively E 41 、E 42 ……E 4n The frequency of occurrence.

[0041] Step S2-4. During the energy consumption calculation on the user side, it includes the electrical energy calculation and the thermal energy calculation on the user side; during the thermal energy calculation process on the user side, when the distributed data obtained is empty, the corresponding parameters are default set to 0 for the thermal energy calculation on the user side; and during the energy consumption calculation process on the user side, the following conditions need to be met:

[0042] In the electrical energy calculation on the user side, it is necessary to satisfy: M ≥ (E3 + E4) × ε1 × c;

[0043] In the thermal energy calculation on the user side, it is necessary to satisfy: M ≥ (H3 + H4) × ε2 × c.

[0044] Step S3. Build an intelligent management platform to execute the charge and discharge control strategy of the electrochemical energy storage.

[0045] Step S3-1. Real-time monitor the voltage, current, temperature and charge and discharge state parameters of the electrochemical energy storage by using distributed optical fiber, current sensor and voltage sensor;

[0046] Step S3-2. Dynamically adjust the charge and discharge through the intelligent management platform. The specific process is as follows:

[0047] Step S3-2-1. Charge and discharge period division: Based on the peak-valley electricity price in the power market, charge during the valley period of the electricity price and discharge during the peak period;

[0048] Step S3-2-2. Charge rate control: Set the maximum charge percentage according to the state of health (SOH) and state of charge (SOC) of the battery, and dynamically adjust the charge rate in combination with the load prediction data and the battery temperature monitoring results. When the temperature exceeds the threshold, start the heat dissipation strategy;

[0049] Step S3-2-3. Discharge control: During the peak period of power demand, control the discharge amount in combination with the real-time load and prediction data, and set the minimum capacity threshold of the electrochemical energy storage to avoid deep discharge;

[0050] Step S3-2-4, Closed-loop control of charge and discharge temperature: Establish a mathematical model of temperature and charging current, and dynamically adjust the current to maintain the battery within the optimal operating temperature range.

[0051] The intelligent management platform also includes an intelligent control method for the molten salt energy storage system, specifically:

[0052] Construct and train a neural network model. Input layer parameters: current molten salt temperature T(t), energy storage capacity E(t), grid load demand D(t), real-time electricity price P(t), and predicted light intensity W(t);

[0053] Hidden layer design: Include the first hidden layer and the second hidden layer;

[0054] The hidden layer design is as follows: Receive the 5 parameters of the input layer. The number of nodes N1 in the hidden layer: The number of nodes is calculated according to the formula as follows:

[0055]

[0056] In the formula, N i represents the number of nodes in the input layer, 5 layers; N o represents the number of nodes in the output layer, 2; a is a constant, and its value range is an integer between 1 and 10; The ReLU activation function g(x) is used for nonlinear transformation, and the function calculation formula is as follows:

[0057] f(x) = (1 + e -x ) -1 ;

[0058] Output layer instruction: Generate control instructions for energy storage power and discharge power. Activation function: The Sigmoid function f(x) is used for the hidden layer, and the function calculation formula is as follows:

[0059] g(x) = max(0, x);

[0060] Collect historical operation data of the molten salt energy storage system, including time series data of input variables and output variables, and preprocess the data, including data cleaning, normalization or standardization.

[0061] Furthermore, the intelligent management platform performs multi-objective optimization control, including:

[0062] Use the preprocessed data to train the neural network model. Set hyperparameters through the mean squared error MSE of the loss function combined with the optimization algorithm Adam. The hyperparameters include learning rate, batch size, and number of iterations. Adjust the weights and biases of the neural network through the backpropagation algorithm to regulate the output error of the neural network model. The calculation formula of the mean squared error MSE of the loss function is as follows:

[0063]

[0064] In the formula, n represents the number of samples, yi is the actual value, is the predicted value of the neural network model;

[0065] The Adam algorithm is used to adjust the weight matrices W1, W2 and the bias vectors b1, b2 to minimize the loss function; the Adam algorithm, as an optimization algorithm, is used to adjust the weight matrices W1, W2 and the bias vectors b1, b2 to minimize the loss function. The parameters of the Adam algorithm include the learning rate α (ranging from 0.001 to 0.01), β1 (with a value of 0.9), β2 (with a value of 0.999), and ε (with a value of 1e - 8).

[0066] Hyperparameters: learning rate (ranging from 0.001 to 0.01), batch size (ranging from 32 to 256), number of iterations (determined according to the performance of the validation set).

[0067] According to the power grid event type, the state of the energy storage device, and the external environment parameters, dynamically allocate the output ratios of short - term, medium - term, and long - term energy storage units.

[0068] By means of the collaboration of energy storage modules with different time scales, effectively cope with the diversity of power grid events. The load forecasting calculation and the accurate calculation of the user - side energy - using resources improve the scientific nature of the energy storage system planning. The intelligent management platform's fine control of the electrochemical energy storage and molten salt energy storage systems, combined with multi - objective optimization, greatly improves the stability, efficiency, and energy - saving economy of the energy storage system operation, ensuring the stable and reliable operation of the regional integrated energy supply system.

[0069] Compared with the prior art, the beneficial effects of the present invention are:

[0070] 1. By constructing a multi - time - scale composite energy storage system, including short - term, medium - term, and long - term energy storage modules, each module collaborates for different - frequency power grid events, can effectively cope with the diversity of power grid events, and compared with a single energy storage system, greatly improves the adaptability to complex power grid conditions, ensuring the stability and reliability of power supply.

[0071] 2. Through accurate load forecasting calculation, combined with data such as the electrical energy and thermal energy used on the user side, accurately calculate the user - side energy - using resources required for energy storage power supply, significantly improve the scientific nature of the energy storage system planning, avoid the problem of energy storage capacity calculation caused by the instability of renewable energy and the variability of user loads, and help the scientific construction and efficient operation of the composite energy storage system.

[0072] 3. The present invention implements fine control over the electrochemical energy storage and molten salt energy storage systems through an intelligent management platform. Combining multi-objective optimal control, it dynamically allocates the output power ratios of each energy storage unit based on the grid event type, the state of the energy storage device, and external environmental parameters, greatly improving the stability, efficiency, and energy-saving economy of the energy storage system operation, and ensuring the stability of the regional integrated energy supply system. Description of the Drawings

[0073] Figure 1 It is a multi-time scale composite energy flow schematic diagram of a multi-time scale composite energy storage system of the present invention;

[0074] Figure 2 It is a process schematic diagram of a multi-time scale composite energy storage system of the present invention;

[0075] Figure 3 It is a daily historical load weight adjustment iterative calculation flow chart of a multi-time scale composite energy storage system of the present invention;

[0076] Figure 4 It is a mathematical fitting schematic of load Q - frequency f of a multi-time scale composite energy storage system of the present invention. Detailed Embodiments

[0077] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0078] The present invention provides a technical solution, a multi-time scale composite energy storage system, and the multi-time scale composite energy storage system includes a short-term energy storage module, a medium-term energy storage module, and a long-term energy storage module:

[0079] The short-term energy storage module is used to handle the inertial characteristic changes of the power grid from seconds to minutes;

[0080] The medium-term energy storage module is used to handle the inertial characteristic changes of the power grid from minutes to hours;

[0081] The long-term energy storage module is used to handle the inertial characteristic changes of the power grid from hours to days.

[0082] The electrochemical energy storage unit is used for the storage and release of electric energy;

[0083] The molten salt energy storage unit is used for the storage and conversion of thermal energy.

[0084] The energy-saving and economic operation method includes the following steps:

[0085] 1. Perform load forecasting calculation:

[0086] Load forecasting is the basis for comprehensive energy storage scheduling. To improve the accuracy of load forecasting, a load forecasting calculation is proposed.

[0087] Step S1. Perform load forecasting calculation according to the similarity function.

[0088] Step S1-1. Taking m as the time step and starting from time 0 o'clock, divide the time points of 0, 0 + m, 0 + 2m,..., 24h every day to construct a historical load database.

[0089] Step S1-2. Based on the historical load database and combined with the statistical analysis algorithm, establish the fitting relationship between the load at the time point to be analyzed and the loads at each time point before the time point to be analyzed: X(t)1, X(t)2,..., X(t) n , where X(t) represents the load function at time t, and 1, 2,..., n represent each day. By adjusting the value of the time step m, the accuracy of the prediction model is adjusted. Among them, the smaller the value of the time step m, the greater the accuracy of the prediction model.

[0090] Step S1-3. Perform dynamic load forecasting based on function similarity. The dynamic forecasting calculation formula is as follows:

[0091] Y(t) = a1X(t)1, a2X(t)2,..., a n X(t) n +b;

[0092] In the formula, Y(t) represents the dynamically predicted load value; a1, a2,..., a n represent the weights of the daily historical loads; n represents the number of days participating in the dynamic load forecasting calculation in the historical load database; b represents the offset.

[0093] Step S1-4. Set the load variance threshold S max , sort according to the values of the load variances from large to small, and set the weights of the daily historical loads a1, a2,..., a n to 1 / n; set the parameter R = 1 / m; adjust the weights of the daily historical loads a1, a2,..., a n through the parameter A. The adjustment process is as follows:

[0094] Add the parameter R to the maximum value of the weights, subtract the parameter R from the minimum value of the weights, and finally make the value of the load variance less than the set S max to end the weight adjustment.

[0095] (1) A method for predicting the daily load interval Q(τ) at a certain time point through the historical load data at that time point is proposed.

[0096] The historical load data at time τ is basically normally distributed. Based on the historical load data at time τ, function fitting is performed to establish a mathematical model between the load magnitude and the occurrence frequency. The type of mathematical function is determined based on the standard deviation, the fitting effect of the function curve and the scatter plot. The determined mathematical model is used as the basic function for load prediction, thereby eliminating some problems with poor prediction generality caused by drastic load changes that may be due to special scenarios. Through this method, the fitting formula of load Q - frequency f at each time point τ is obtained, and then the minimum load and maximum load at each time point are determined. The load Q - frequency f fitting is as Figure 4 shown.

[0097] (2) Perform load prediction based on the similarity function;

[0098] The platform first automatically retrieves the historical load data of electricity consumption, and then performs dynamic load prediction based on the load at the prediction base point;

[0099] ① Taking m as the time step and starting from 0 o'clock, historical load databases are established for each working day at time points such as 0, 0 + m, 0 + 2m…, 24h;

[0100] ② Based on the historical load database, a statistical analysis algorithm is used to establish a function fitting relationship between the load at the time point to be analyzed and the loads at each time point before the time point to be analyzed. For example, the mathematical relationship between the loads at 0 + m and 0 o'clock, the mathematical relationship between the loads at 0 + 2m and 0 + m, 0 o'clock; the mathematical relationship between the loads at 0 + 3m and 0 + 2m, 0 + m, 0 o'clock; and so on. In this way, a mathematical function model between the load at the time point to be predicted and the loads at each time point before the time point to be predicted is established. For the historical load data of each day in the prediction reference period, X(t)1, X(t)2, ……X(t)n are established, where X(t) represents the load function at time t, and 1, 2…n represent each day. The accuracy of the prediction model is adjusted by adjusting the size of the time step m. The smaller m is, the greater the prediction accuracy.

[0101] ③ Dynamic load prediction

[0102] Dynamic load prediction based on function similarity:

[0103] Y(t) = a1X(t)1, a2X(t)2, ……a n X(t) n +b;

[0104] a1 + a2 + …… + a n = 1;

[0105] Wherein, a1, …… an are the weights of daily historical loads, n is the number of days for historical data statistics, and b is the offset; the principle of the above formula is actually to first establish a functional relationship between a certain time point within a day and other time points to obtain X(t)1, X(t)2, …… X(t)n, and then on this basis, synthesize the functional relationships of several days, and adjust the accuracy of the prediction function by controlling the weights (a1, …… an);

[0106] According to as Figure 3 shown, based on indicators such as variance, by adjusting the weights, determine the fitting mathematical function between the load and time in the time period before the prediction reference point. Based on the process shown in the following figure, adjust the weight size, and through iterative calculation, gradually determine the weight sizes of a1, …… an, and then determine the prediction function, and determine the predicted load at future time points based on the prediction function.

[0107] Step S2: Calculate the user-side energy resources required for energy supply of the energy storage according to the prediction calculation result of the load in combination with the user-side energy consumption data, where the user-side energy consumption data includes user-side electricity energy and user-side heat energy;

[0108] Step S2-1: Calculate the input amount of energy storage energy supply resources, and the input amount of energy storage energy supply resources is calculated using the following formula:

[0109]

[0110] Wherein, c represents the input amount of energy storage device resources corresponding to the unit charge-discharge amount; M represents the initial input resources of the energy storage energy supply device; N represents the resources required for maintenance in each year during the operation of the energy storage device; n represents the service life of the energy storage device; i represents each year, i = 1, 2, …, n; r is the resource benefit depreciation rate; P represents the remaining available resources of the energy storage device after the end of the service life; A represents the total charge-discharge amount of the energy storage device during the whole life cycle;

[0111] Step S2-2: Determine the energy storage capacity of the energy storage power station according to the input amount of energy storage energy supply resources in combination with the predicted load;

[0112] Step S2-2-1: Calculate the user-side energy resources, and calculate using the following formula:

[0113] P = E1×p1 + E2×p2 + (E3 + E4)×ε1×c;

[0114] Wherein, P is the energy consumption resources on the user side; E1 is the municipal energy supply; E2 is the direct energy production supply of distributed energy, E3 is the energy stored in the energy storage device from municipal energy, E4 is the energy stored in the energy storage device from distributed energy, and ε1 is the energy conversion efficiency of the energy storage device; p1 is the proportion of municipal energy resources; p2 is the converted input resource quantity per unit energy production of the distributed energy direct energy production equipment; c is the converted input resource quantity per unit energy production of the energy storage device;

[0115] Step S2-2-2: Calculate the first user-side energy consumption E based on the historical energy consumption data on the user side now , and calculate it using the following formula:

[0116] E now = E1 + E2 + (E3 + E4) × ε1;

[0117] Step S2-2-3: Calculate the total user-side energy consumption E by combining the predicted load Y(t).

[0118]

[0119] Wherein, e1 represents the weight of the first user-side energy consumption E now ; e2 represents the weight of the user-side energy consumption calculated based on the predicted load.

[0120] Step S2-3: Screen out the minimum value of the user-side energy consumption resources P to determine the values of E1, E2, E3, and E4. Obtain a matrix through the change amount E, and combine the weighted average to determine the values of E1, E2, E3, and E4. Calculate using the following formula:

[0121] E1 = a1*E 11 + a2*E 12 + …… + a n * E 1n ;

[0122] E2 = b1*E 21 + b2*E 22 +...... + b n * E 2n ;

[0123] E3 = c1*E 31 + c2*E 32 +...... + c n * E 3n ;

[0124] E4 = d1*E 41 + d2*E 42 +...... + d n * E 4n ;

[0125] In the formula, E 11 , E 12 ...E 1n represents the numerical value of E1, and a1, a2... a n are respectively the frequencies of occurrence of E 11 , E 12 ...E 1n ; E 21 , E 22 ...E 2n represents the numerical value of E2, and b1, b2... b n are respectively the frequencies of occurrence of E 21 , E 22 ...E 2n ; E 31 , E 32 ...E 3n represents the numerical value of E3, and c1, c2... c n are respectively the frequencies of occurrence of E 31 , E 32 ...E 3n ; E 41 , E 42 ...E 4n represents the numerical value of E4, and d1, d2... d n are respectively the frequencies of occurrence of E 41 , E 42 ...E 4n the frequency of occurrence.

[0126] Step S2-4: During the energy consumption calculation on the user side, it includes the user-side electric energy calculation and the user-side heat energy calculation; during the user-side heat energy calculation process, when the distributed data obtained is empty, the corresponding parameters are default set to 0 for the user-side heat energy calculation; and during the user-side energy consumption calculation process, the following conditions need to be met:

[0127] In the user-side electric energy calculation, it is necessary to satisfy: M ≥ (E3 + E4) × ε1 × c;

[0128] In the user-side heat energy calculation, it is necessary to satisfy: M ≥ (H3 + H4) × ε2 × c.

[0129] Step S3: Build an intelligent management platform to execute the electrochemical energy storage charge and discharge control strategy.

[0130] Step S3-1: Real-time monitor the voltage, current, temperature and charge and discharge state parameters of the electrochemical energy storage by collecting them in real time through distributed optical fibers, current sensors and voltage sensors;

[0131] Step S3-2: Dynamically adjust the charging and discharging through the intelligent management platform. The specific process is as follows:

[0132] Step S3-2-1, Charge and Discharge Period Division: Based on the peak-valley electricity price in the power market, charge during the valley period of the electricity price and discharge during the peak period.

[0133] Step S3-2-2, Charge Rate Control: Set the maximum charging percentage according to the state of health (SOH) and state of charge (SOC) of the battery, dynamically adjust the charging rate in combination with the load prediction data and the battery temperature monitoring results, and activate the heat dissipation strategy when the temperature exceeds the threshold.

[0134] Step S3-2-3, Discharge Control: During the peak period of power demand, control the discharge amount in combination with the real-time load and prediction data, and set the minimum capacity threshold of the electrochemical energy storage to avoid deep discharge.

[0135] Step S3-2-4, Charge and Discharge Temperature Closed-Loop Control: Establish a mathematical model of the temperature and charging current, and dynamically adjust the current to maintain the battery within the optimal operating temperature range.

[0136] The intelligent management platform also includes an intelligent control method for the molten salt energy storage system, specifically:

[0137] Construct and train a neural network model. Input layer parameters: current molten salt temperature T(t), energy storage E(t), grid load demand D(t), real-time electricity price P(t), and predicted light intensity W(t).

[0138] Hidden layer design: Include the first hidden layer and the second hidden layer.

[0139] The hidden layer design is as follows: Receive 5 parameters from the input layer. The number of nodes in the hidden layer N1: The number of nodes is calculated according to the formula as follows:

[0140]

[0141] In the formula, N i represents the number of nodes in the input layer, 5 layers; N o represents the number of nodes in the output layer, 2; a is a constant, and its value range is an integer between 1 and 10; Use the ReLU activation function g(x) for non-linear transformation, and the function calculation formula is as follows:

[0142] f(x)=(1+e -x ) -1 ;

[0143] Output layer instruction: Generate control instructions for the energy storage power and discharge power. Activation function: The Sigmoid function f(x) is used for the hidden layer, and the function calculation formula is as follows:

[0144] g(x)=max(0,x);

[0145] Collect the historical operation data of the molten salt energy storage system, including the time series data of input variables and output variables, and preprocess the data, including data cleaning, normalization or standardization.

[0146] The intelligent management platform performs multi-objective optimization control, including:

[0147] Use the preprocessed data to train a neural network model. Set hyperparameters through the mean squared error (MSE) of the loss function combined with the Adam optimization algorithm. The hyperparameters include the learning rate, batch size, and number of iterations. Adjust the weights and biases of the neural network through the backpropagation algorithm to regulate the output error of the neural network model. The calculation formula of the mean squared error (MSE) of the loss function is as follows:

[0148]

[0149] In the formula, n represents the number of samples, yi is the actual value, is the predicted value of the neural network model;

[0150] The Adam algorithm is used to adjust the weight matrices W1, W2 and the bias vectors b1, b2 to minimize the loss function; the Adam algorithm, as an optimization algorithm, is used to adjust the weight matrices W1, W2 and the bias vectors b1, b2 to minimize the loss function. The parameters of the Adam algorithm include the learning rate α (the value ranges from 0.001 to 0.01), β1 (the value is 0.9), β2 (the value is 0.999), and ε (the value is 1e-8).

[0151] Hyperparameters: learning rate (the value ranges from 0.001 to 0.01), batch size (the value ranges from 32 to 256), number of iterations (determined according to the performance of the validation set).

[0152] According to the grid event type, the state of the energy storage device, and the external environment parameters, dynamically allocate the output ratios of short-term, medium-term, and long-term energy storage units;

[0153] Energy storage power adjustment strategy: When it is predicted that the future electricity price is low and the molten salt energy storage level is low, increase the energy storage power Pc for charging to store energy using low-cost electricity. When it is predicted that the future electricity price is high but the molten salt energy storage is full or nearly full, reduce or stop the energy storage power to avoid unnecessary energy waste.

[0154] It is obvious to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. A multi-time-scale composite energy storage system, characterized in that: The multi-time-scale composite energy storage system includes a short-time energy storage module, a medium-time energy storage module, and a long-time energy storage module: The short-time energy storage module is used to handle the inertial characteristic changes of the power grid from seconds to minutes. The inertial characteristic changes are expressed as the frequency fluctuation response characteristics caused by the dynamic imbalance between the power generation power and the load power within a specific time; The medium-time energy storage module is used to handle the inertial characteristic changes of the power grid from minutes to hours; The long-time energy storage module is used to handle the inertial characteristic changes of the power grid from hours to days.

2. A multi-time-scale composite energy storage system according to claim 1, wherein: Both the energy storage input form and the energy storage output form controlled by the short-time energy storage module are electric energy; Both the energy storage input form and the energy storage output form controlled by the medium-time energy storage module are electric energy; The energy storage input forms controlled by the long-time energy storage module include electric energy, light energy, and waste heat, and the energy storage output forms controlled by the long-time energy storage module include electric energy, cold energy, and heat energy.

3. A multi-time-scale composite energy storage system according to claim 1, characterized in that: The multi-time-scale composite energy storage system further includes an electrochemical energy storage unit and a molten salt energy storage unit; The electrochemical energy storage unit is used for the storage and release of electric energy; The molten salt energy storage unit is used for the storage and conversion of heat energy.

4. An energy-saving and economic operation method, applied to a multi-time-scale composite energy storage system according to any one of claims 1-3, characterized in that: The energy-saving and economic operation method includes the following steps: Step S1: Perform predictive calculation on the load according to the similarity function; Step S2: Calculate the user-side energy consumption resources required for energy storage supply according to the predictive calculation results of the load in combination with the user-side energy consumption data. The user-side energy consumption data includes user-side electricity consumption and user-side heat consumption; Step S3: Build an intelligent management platform to execute the electrochemical energy storage charge and discharge control strategy.

5. An energy-saving and economic operation method according to claim 4, characterized in that: In step S1, the prediction of the load according to the similarity function includes the following steps: Step S1-1: Divide the daily time points of 0, 0 + m, 0 + 2m…, 24h with m as the time step and starting from time 0 to construct a historical load database; Step S1-2: Based on the historical load database, establish the fitting relationship of the load function between the time point to be analyzed and the load at each time point before the time point to be analyzed by combining the statistical analysis algorithm: X(t)1, X(t)2, …… X(t) n , where X(t) represents the load function at time t, and 1, 2, ……, n represent each day. The prediction model accuracy is adjusted by adjusting the value of the time step m. Among them, the smaller the value of the time step m, the greater the prediction model accuracy; Step S1-3: Perform dynamic prediction on the load based on function similarity. The dynamic prediction calculation formula is as follows: Y(t) = a1X(t)1, a2X(t)2,..., a n X(t) n + b; Wherein, Y(t) represents the dynamically predicted load value; a1, a2... a n represents the weight of the daily historical load; n represents the number of days participating in the dynamic load prediction calculation in the historical load database; b represents the offset; Step S1-4: Set the load variance threshold S max , sort the values of the load variance in descending order, and set the weights of the daily historical loads a1, a2... a n to 1 / n; set the parameter R = 1 / m; adjust the weights of the daily historical loads a1, a2... a n through the parameter A. The adjustment process is as follows: Add the parameter R to the maximum value of the weights and subtract the parameter R from the minimum value of the weights, ultimately making the value of the load variance less than the set S max End the weight adjustment when this occurs.

6. The energy-saving and economical operation method according to claim 5, characterized in that: The analysis and calculation of the energy storage capacity includes the following steps: Step S2-1: Calculate the input amount of energy storage supply resources. The calculation of the input amount of energy storage supply resources uses the following formula: In the formula, c represents the input amount of energy storage device resources corresponding to the unit charge and discharge amount; M represents the initial input resources of the energy storage supply device; N represents the resources required for maintenance in each year during the operation of the energy storage device; n represents the service life of the energy storage device; i represents each year, i = 1, 2,..., n; r is the resource benefit loss rate; P represents the remaining available resources of the energy storage device after the end of its service life; A represents the total charge and discharge amount of the energy storage device during its entire life cycle; Step S2-2: Determine the energy storage capacity of the energy storage power station according to the input amount of energy storage supply resources in combination with the predicted load; Step S2-2-1: Calculate the user-side energy consumption resources using the following formula: P = E1×p1 + E2×p2 + (E3 + E4)×ε1×c; Wherein, P is the energy consumption resource on the user side; E1 is the energy supply of municipal energy consumption; E2 is the direct energy production supply of distributed energy; E3 is the energy stored in the energy storage device from municipal energy consumption; E4 is the energy stored in the energy storage device from distributed energy consumption; ε1 is the energy conversion efficiency of the energy storage device; p1 is the proportion of municipal energy resources; p2 is the converted input resource quantity per unit energy production of the distributed energy direct energy production equipment; c is the converted input resource quantity per unit energy production of the energy storage device; Step S2-2-2: Calculate the first user-side energy consumption E based on the historical energy consumption data of the user side now , which is calculated using the following formula: E now = E1 + E2 + (E3 + E4) × ε1; Step S2-2-3: Calculate the total energy consumption E on the user side by combining the predicted load Y(t). Wherein, e1 represents the weight of the first user-side energy consumption E now ; e2 represents the weight of the user-side energy consumption calculated according to the predicted load.

7. An energy-saving economic operation method according to claim 6, characterized in that: In step S2, it further includes: Step S2-3: Screen out the minimum value of the energy consumption resource P on the user side to determine the values of E1, E2, E3, and E4. Obtain a matrix through the change amount E, and combine weighted average to determine the values of E1, E2, E3, and E4. Calculate using the following formula: E1 = a1 * E 11 + a2 * E 12 +......+ a n * E 1n ; E2 = b1 * E 21 + b2 * E 22 +...... + b n * E 2n ; E3 = c1 * E 31 + c2 * E 32 +......+ c n * E 3n ; E4 = d1*E 41 + d2*E 42 +......+ d n *E 4n ; where E 11 、E 12 ...E 1n represent the numerical values of E1, and a1, a2... a n are the frequencies of occurrence of E 11 、E 12 ……E 1n respectively; E 21 、E 22 ...E 2n represent the numerical values of E2, and b1, b2... b n are the frequencies of occurrence of E 21 、E 22 ……E 2n respectively; E 31 、E 32 ...E 3n represent the numerical values of E3, and c1, c2... c n are the frequencies of occurrence of E 31 、E 32 ……E 3n respectively; E 41 、E 42 ...E 4n represent the numerical values of E4, and d1, d2... d n are the frequencies of occurrence of E 41 、E 42 ……E 4n respectively; Step S2-4: In the process of calculating the energy consumption on the user side, it includes the calculation of the electrical energy on the user side and the calculation of the thermal energy on the user side; in the process of calculating the thermal energy on the user side, when the distributed data obtained is empty, the corresponding parameters are default set to 0 for the calculation of the thermal energy on the user side; and in the process of calculating the energy consumption on the user side, the following conditions need to be met: In the calculation of the electrical energy on the user side, it is necessary to meet: M ≥ (E3 + E4) × ε1 × c; In the calculation of the thermal energy on the user side, it is necessary to meet: M ≥ (H3 + H4) × ε2 × c.

8. A method for energy-saving and economic operation according to claim 4, characterized in that: The intelligent management platform constructed in step S3 executes the electrochemical energy storage charge and discharge control strategy, specifically including: Step S3-1: Real-time monitor the voltage, current, temperature, and charge and discharge state parameters of the electrochemical energy storage through distributed optical fibers, current sensors, and voltage sensors. Step S3-2: Dynamically adjust charging and discharging through the intelligent management platform. The specific process is as follows: Step S3-2-1: Division of charging and discharging periods: Based on the peak-valley electricity price in the power market, charge during the valley period of the electricity price and discharge during the peak period. Step S3-2-2: Charging rate control: Set the maximum charging percentage according to the state of health (SOH) and state of charge (SOC) of the battery, and dynamically adjust the charging rate in combination with the load prediction data and the battery temperature monitoring results. When the temperature exceeds the threshold, start the heat dissipation strategy. Step S3-2-3: Discharge control: During the peak period of power demand, control the discharge amount in combination with the real-time load and prediction data, and set the minimum capacity threshold of the electrochemical energy storage to avoid deep discharge. Step S3-2-4: Closed-loop control of charging and discharging temperature: Establish a mathematical model of temperature and charging current, and dynamically adjust the current to maintain the battery within the optimal operating temperature range.

9. An energy-saving economic operation method according to claim 8, characterized in that: In step S3, the intelligent management platform also includes an intelligent control method for the molten salt energy storage system, specifically: Construct and train a neural network model. Input layer parameters: the current molten salt temperature T(t), the stored energy E(t), the grid load demand D(t), the real-time electricity price P(t), and the predicted light intensity W(t); Hidden layer design: It includes the first hidden layer and the second hidden layer; Hidden layer design is as follows: Receive the 5 parameters of the input layer. The number of nodes in the hidden layer N1: The number of nodes is calculated according to the following formula: where N i represents the number of nodes in the input layer, which is 5 layers; N o represents the number of nodes in the output layer, which is 2; a is a constant, and its value range is an integer between 1 and 10; the ReLU activation function g(x) is used for non-linear transformation, and the function calculation formula is as follows: f(x) = (1 + e -x ) -1 ; Output layer instruction: Generate energy storage power and discharge power control instructions. Activation function: The Sigmoid function f(x) is used for the hidden layer, and the function calculation formula is as follows: g(x) = max(0, x); Collect historical operation data of the molten salt energy storage system, including time series data of input variables and output variables, and preprocess the data, including data cleaning, normalization or standardization.

10. A method for energy-saving and economic operation according to claim 9, characterized in that: The intelligent management platform performs multi-objective optimization control, including: Train the neural network model using the preprocessed data, set hyperparameters through the mean squared error MSE of the loss function combined with the Adam optimization algorithm. The hyperparameters include learning rate, batch size, and number of iterations. Adjust the weights and biases of the neural network through the backpropagation algorithm to regulate the output error of the neural network model. The calculation formula of the mean squared error MSE of the loss function is as follows: Where n represents the number of samples, yi is the actual value, and is the predicted value of the neural network model; The Adam algorithm is used to adjust the weight matrices W1, W2 and bias vectors b1, b2 to minimize the loss function; Dynamically allocate the output ratios of short-term, medium-term, and long-term energy storage units according to the grid event type, energy storage device status, and external environment parameters.

Images