A power grid generator unit scheduling optimization method

Abstract

The present invention discloses a method for optimizing the dispatching of power grid generating units, which belongs to the field of power grid dispatching. The present invention is used to solve the problem that environmental factors are usually ignored in the prior art. The present invention predicts the power grid load in a future time period through a neural network model; establishes a power grid power generation system model, constructs an objective function with the lowest operating cost of the power generation system, and the constraint conditions of the power grid power generation system model include the power balance constraint of the power generation system based on the power grid load in a future time period, the maximum and minimum output constraints of each unit, the continuous startup and shutdown time constraints, the unit ramp rate constraint, the energy storage device constraint, and the synchronous condenser constraint; based on the objective function and the constraint conditions, the wolf pack optimization algorithm is used to obtain the optimal unit combination method for power grid dispatching. The dispatching method of the present invention takes into account both economic factors and environmental factors, helps to timely master the power generation state of the power grid, optimize the power grid dispatching, and improve the power generation economy.

Description

Technical Field

[0001] The present invention relates to the field of power grid dispatching, and particularly to a method for optimizing the dispatching of power grid generating units. Background Art

[0002] With the large-scale application of technologies such as new energy power generation, energy storage, electric vehicle charging stations, decentralized charging piles, and microgrids in the distribution network, the traditional distribution network has shifted from a passive and unidirectional power supply mode to an active and bidirectional power supply mode, which has brought many new problems to the coordinated dispatching of the distribution network. (1) Since photovoltaic power generation and wind power generation are easily affected by the environment, their output has the characteristics of randomness, intermittency, and volatility, resulting in an increasing degree of uncertainty in the actual operation and control of the distribution network, and the difficulty of system dispatching and operation is also increasing day by day; (2) The traditional distribution network usually adopts a passive and unidirectional power supply network, and the access and application of a large number of distributed generations have changed the system network structure into a multi-source network with bidirectional power flow, which may cause problems such as reverse power flow in the system, bringing safety risks to the power grid system. At the same time, the phenomenon of wind and light abandonment is widespread in many regions, and there are still great difficulties in the complete consumption of clean energy.

[0003] Coordinated control is for a distribution network with a high penetration of distributed generation in its system, and this distribution network has both active control ability and active management ability at the same time. Coordinated control enables the active participation of the power generation side in the system through coordinated control strategies, and at the same time enables the active response of the user side. In this way, the distribution network begins to actively utilize electric energy from the previous passive acceptance of electric energy. Coordinated control can achieve the full utilization of controllable resources, and researchers believe that it is another effective mode after the microgrid and virtual power plant models to support the grid-connected coordinated control of large-scale distributed generation. By actively controlling various controllable resources, such as controllable loads, energy storage systems, and distribution network grid structures, etc., the optimization of coordinated control can be achieved.

[0004] Regarding the problem of coordinated optimization of unit commitment and unit maintenance, existing technologies all choose to construct a single-objective model, mostly aiming at minimizing the total cost of the power system to achieve the economic operation of the power system. However, social welfare and environmental responsibilities in the actual operation process of the power system are also crucial. Therefore, those skilled in the art are committed to providing a method for optimizing the dispatching of power grid generating units that takes into account both economic factors and environmental factors. Summary of the Invention

[0005] In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present invention is to provide a method for optimizing the dispatching of power grid generating units that can take into account both economic factors and environmental factors.

[0006] The present invention provides a method for optimizing the dispatching of power grid generating units, including the following steps:

[0007] S1. Predict the power grid load for a future time period through a neural network model;

[0008] S2. Establish a power grid generation system model, where the generation system model includes thermal power units, photovoltaic units, wind power units, energy storage devices, and synchronous condensers. Build an objective function with the lowest operating cost of the power generation system. The constraint conditions of the power grid generation system model include power balance constraints of the generation system based on the power grid load for a future time period, maximum output constraints, minimum output constraints, continuous startup time constraints, continuous shutdown time constraints, unit ramp constraints, energy storage device constraints, and synchronous condenser constraints;

[0009] S3. Based on the objective function and constraint conditions, use the wolf pack optimization algorithm to obtain the optimal unit combination method for power grid scheduling.

[0010] Further, the neural network model is an LSTM neural network model.

[0011] Further, the objective function is:

[0012] ;

[0013] Where i is the number of each unit; is the operating cost of the thermal power unit at time period j; is the operating cost of the wind power unit at time period j; is the operating cost of the photovoltaic unit at time period j; is the charging and discharging cost of the energy storage device at time period j; is the operating cost of the synchronous condenser at time period j.

[0014] Further, the operating cost of the thermal power unit includes energy consumption cost and environmental cost, and the energy consumption cost is:

[0015] ;

[0016] Where, is the energy consumption cost of the unit; is the number of units; is the number of time periods; is the output of the i -th thermal power unit at time period j; is 1 or 0, representing whether the i -th unit outputs power at time period j respectively; is the energy consumption cost of starting the i -th unit within time period j; is the energy consumption cost of the i -th unit's output; is the consumption coefficient of the unit;

[0017] The environmental cost is the total penalty cost of pollutant emissions is:

[0018] ;

[0019] Among them, is the penalty cost for carbon dioxide pollutant emissions; is the penalty cost for sulfur dioxide pollutant emissions;

[0020] The penalty cost for carbon dioxide pollutant emissions and the penalty cost for sulfur dioxide pollutant emissions have the following relationship with the emissions:

[0021] ;

[0022] ;

[0023] Among them, is the emission of carbon dioxide pollutants; is the relationship coefficient between the carbon dioxide pollutant emissions of the i - th thermal power unit and the output of the thermal power unit; is the emission of sulfur dioxide pollutants; is the relationship coefficient between the sulfur dioxide pollutant emissions of the i - th thermal power unit and the unit output.

[0024] Furthermore, the operating cost of the wind turbine is:

[0025] ;

[0026] is the wind power generation cost, which is:

[0027] ;

[0028] is the curtailment cost, which is:

[0029] ;

[0030] Among them, is the coefficient of the wind turbine power generation cost function; is the active power output of the i - th wind turbine in the j - th time period; is the cost coefficient of wind turbine curtailment; is the wind speed in the j - th time period; is the maximum value of.

[0031] Furthermore, the operating cost of the photovoltaic unit is:

[0032] ;

[0033] The power generation cost of the photovoltaic unit is as follows:

[0034]

[0035] The curtailment cost of photovoltaic power is as follows:

[0036] ;

[0037] Among them, is the coefficient of the power generation cost function of the photovoltaic unit; is the active power output of the i-th photovoltaic unit in the j-th time period; is the cost coefficient of curtailment of photovoltaic power; is the solar radiation amount in the j-th time period; is the maximum value of.

[0038] Furthermore, the cost of the synchronous condenser is the cost of reactive power regulation, which is as follows:

[0039] ;

[0040] Among them, is the coefficient of the operating cost function of the synchronous condenser; is the reactive power output of the i-th synchronous condenser in the j-th time period.

[0041] Furthermore, 1) The constraint of the power balance of the power generation system based on the grid load in the next time period is as follows:

[0042] ;

[0043] Among them, , is the power output of wind power in the j-th time period, is the power output of photovoltaic power in the j-th time period, is the power output of thermal power in the j-th time period, is the power output of energy storage in the j-th time period, is the load in the j-th time period;

[0044] When is the case, ; when is the case, ; ;

[0045] 2) The maximum and minimum output constraints of each unit are as follows:

[0046] ;

[0047] Among them, They are the minimum and maximum limits of the output of Unit i respectively. They are the minimum and maximum limits of the reactive power output of Thermal Power Unit i respectively.

[0048] 3) The continuous startup constraint and shutdown time constraint are as follows:

[0049] ;

[0050] 4) The unit ramp-up constraint is as follows:

[0051] ;

[0052] Among them, They are the limits of the output ramp-up and ramp-down speeds of Unit i within a unit time period respectively.

[0053] 5) The energy storage device constraints include SOC constraint, charge-discharge power constraint, and charge-discharge capacity constraint. The SOC constraint is as follows:

[0054] ;

[0055] Among them, is the minimum SOC value required during the operation of the energy storage device, is the maximum SOC value required during the operation of the energy storage device;

[0056] ;

[0057] Among them, is the maximum charging power required during the operation of the energy storage device, is the maximum discharging power required during the operation of the energy storage device;

[0058] The charge-discharge capacity constraint is as follows:

[0059] ;

[0060] Among them, is the number of energy storage units, is the output power of the i-th energy storage device, is the total power that the energy storage device needs to provide for the distribution network, They are the minimum and maximum output powers of the energy storage device respectively;

[0061] 6) The synchronous condenser constraint is the node voltage constraint:

[0062] .

[0063] Furthermore, in the wolf pack algorithm, the operating cost of each unit is used as the position vector X of the grey wolf individual in the grey wolf optimization algorithm; the fitness function of the grey wolf optimization algorithm is the objective function.

[0064] Compared with the prior art, the present invention has the following technical effects:

[0065] 1. The present invention predicts the short-term power generation of new energy units through the LSTM network. Effective prediction of the new energy grid enables the grid dispatching system to effectively utilize new energy, laying a foundation for unit commitment and active power dispatching.

[0066] 2. The present invention combines the problems of unit commitment and economic issues, and pays more attention to the social welfare issues during grid power generation, laying a foundation for the further popularization of the new energy grid and flexibly adjusting the unit output.

[0067] 3. The present invention adopts a coordinated control method based on the grey wolf optimization algorithm to solve the objective function. The grey wolf optimization algorithm has simpler parameters and faster optimization speed, and can solve the objective function more accurately and quickly, laying a foundation for power plants to formulate power generation plans, realizing the economic and reliable operation of the power system, and taking into account social welfare responsibilities at the same time.

[0068] The optimization method of the present invention can effectively predict the operation state of the grid load; predicting the grid load and formulating the unit output according to the prediction results helps to timely grasp the power generation state of the grid, formulate the power generation plan curve for different unit time periods by the grid, optimize the grid dispatching, and improve the power generation economy; it is of great significance to improve the operation safety, reliability, economy and social welfare of the grid.

[0069] The following will further illustrate the concept, specific structure and technical effects of the present invention in conjunction with the drawings, so as to fully understand the purpose, features and effects of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 is a schematic diagram of the basic flow of the grid unit dispatching optimization method according to a specific embodiment of the present invention;

[0071] Figure 2 is a diagram of the LSTM neural network model according to a specific embodiment of the present invention;

[0072] Figure 3 is a diagram of the grey wolf hierarchy in the grey wolf optimization algorithm according to a specific embodiment of the present invention;

[0073] Figure 4 is a schematic diagram of the principle of the grey wolf optimization algorithm according to a specific embodiment of the present invention;

[0074] Figure 5 is a flowchart of the grey wolf optimization algorithm according to a specific embodiment of the present invention;

[0075] Figure 6The load diagram for the 365th day predicted by the LSTM neural network according to historical data in a specific embodiment of the present invention;

[0076] Figure 7 The load diagram for the 366th day predicted by the LSTM neural network according to historical data in a specific embodiment of the present invention;

[0077] Figure 8 The operation diagram of the unit output based on load prediction by the particle swarm optimization algorithm in a specific embodiment of the present invention;

[0078] Figure 9 The operation diagram of the unit output based on load prediction by the grey wolf optimization algorithm in a specific embodiment of the present invention. Specific Embodiments

[0079] The following specific examples illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0080] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Therefore, only the components related to the present invention are shown in the diagrams, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0081] Some exemplary embodiments of the present invention are described for the purpose of illustration. It should be understood that the present invention can be implemented in other ways not specifically shown in the drawings.

[0082] As Figure 1 shown, in a specific embodiment, a power grid generator unit scheduling optimization method is provided. The scheduling design is carried out according to the power grid load and the operation cost of the generator units in a future time period. The future time period is one day in the future. In other embodiments, it can be set according to actual needs. For example, it can be one week, one month, one quarter, etc. The future time period in the following embodiments is one day, and the principle of the generator unit scheduling optimization method for the rest of the time periods is the same and will not be elaborated here.

[0083] It includes the following steps:

[0084] S1. Predict the power grid load for the next day through the LSTM neural network model;

[0085] Long Short-Term Memory networks - often just called "LSTMs" - are a special kind of RNN, capable of learning long-term dependencies. LSTMs are explicitly designed to avoid the long-term dependency problem. All recurrent neural networks have a form of neural network repeating module chain. In a standard RNN, this repeating module will have a very simple structure, such as a single tanh layer. LSTMs also have this chain-like structure, but the repeating module has a different structure. There are four, rather than one, neural network layers that interact in a very special way.

[0086] As Figure 2 shown, there are three gate structures in the LSTM network, namely the forget gate, the input gate, and the output gate.

[0087] Forget gate: At each time step, the value in the memory cell undergoes a process of whether to be forgotten.

[0088] ;

[0089] Input gate: The information input layer, and the switch of the input layer determines whether there is information input to the memory cell at this time step.

[0090] ;

[0091] ;

[0092] Immediately update the value of the cell according to the forget gate and the output gate.

[0093] ;

[0094] Output gate: The information output layer. Whether there is information output from the memory cell at a certain time step depends on the output gate.

[0095] ;

[0096] ;

[0097] LSTMs control the transmission state through gated states, remembering what needs to be remembered for a long time and forgetting unimportant information; unlike ordinary RNNs that can only have one way of memory stacking.

[0098] Input the historical data of the daily load operation of the power grid into the LSTM neural network for training, and utilize the characteristics of the LSTM recurrent network to cyclically predict the load operation state of the future days, thereby reducing the prediction error.

[0099] S2. Establish a power grid power generation system model. The power generation system model includes thermal power units, photovoltaic units, wind turbine units, energy storage devices, and synchronous condensers, and establish an objective function and constraints with the lowest operating cost of the power generation system;

[0100] The objective function is:

[0101] ;

[0102] where i is the number of each unit; j is the jth period of unit operation; is the operating cost of the thermal power unit in the jth period; is the operating cost of the wind turbine unit in the jth period; is the operating cost of the photovoltaic unit in the jth period; is the charging and discharging cost of the energy storage device in the jth period; is the operating cost of the synchronous condenser in the jth period; In a specific embodiment, T = 24 hours, and j represents the jth hour. Therefore, the above objective function is the daily operating comprehensive economic cost function of each unit.

[0103] The operating cost of the thermal power unit includes energy consumption cost and environmental cost. The energy consumption cost is:

[0104] ;

[0105] where, is the energy consumption cost of the unit; is the number of units; i represents the ith thermal power unit, is the number of time periods. In this embodiment, T = 24, representing 24 hours, and j represents the jth hour; is the active power output of the ith thermal power unit in the jth time period; is 1 or 0, representing whether the ith unit is outputting power in the jth time period respectively; is the energy consumption cost of starting the ith unit in the jth time period; is the energy consumption cost of the ith unit's output; is the consumption coefficient of the unit;

[0106] The environmental cost is the total penalty cost of pollutant emissions is:

[0107] ;

[0108] where, is the penalty cost of carbon dioxide pollutant emissions; is the penalty cost of sulfur dioxide pollutant emissions;

[0109] The penalty cost of carbon dioxide pollutant emissions The penalty cost for sulfur dioxide pollutant emissions The relationship with the emission volume is:

[0110] ;

[0111] ;

[0112] Among them, in this embodiment, i represents the i-th thermal power unit, is the number of time periods. In this embodiment, T = 24, indicating 24 hours, and j represents the j-th hour; is the emission volume of carbon dioxide pollutants; is the relationship coefficient between the carbon dioxide pollutant emission volume of the i-th thermal power unit and the output of the thermal power unit; is the emission volume of sulfur dioxide pollutants; is the relationship coefficient between the sulfur dioxide pollutant emission volume of the i-th thermal power unit and the unit output.

[0113] The operating cost of the wind turbine generator set includes the wind power generation cost and the curtailment cost;

[0114]

[0115] is the wind power generation cost; is the curtailment cost.

[0116] The wind power generation cost is:

[0117] ;

[0118] The curtailment cost is:

[0119] ;

[0120] Among them, i represents the i-th wind turbine generator set, is the number of time periods. In this embodiment, T = 24, indicating 24 hours, and j represents the j-th hour; is the coefficient of the wind turbine generator set power generation cost function; is the active power output of the i-th wind turbine generator set in the j-th hour; is the cost coefficient of the wind turbine generator set curtailment; is the wind speed in the j-th time period; is the maximum value. The operating cost of the photovoltaic unit includes the photovoltaic unit power generation cost and the curtailment cost;

[0121] The operating cost of the photovoltaic unit includes the photovoltaic power generation cost and the curtailment cost;

[0122] ;

[0123] The power generation cost of the photovoltaic unit is as follows:

[0124] ;

[0125] Cost of curtailment of photovoltaic power:

[0126] ;

[0127] where i represents the i-th photovoltaic unit, is the number of time periods. In this embodiment, T = 24, representing 24 hours, and j represents the j-th hour; is the coefficient of the power generation cost function of the photovoltaic unit; is the active power output of the i-th photovoltaic unit in the j-th hour; is the cost coefficient of curtailment of photovoltaic power; is the solar radiation amount in the j-th hour; is the maximum value of.

[0128] The cost of the synchronous condenser is the cost of reactive power regulation, which is:

[0129] .

[0130] where, is the coefficient of the operating cost function of the synchronous condenser; is the reactive power output of the i-th synchronous condenser in the j-th hour;

[0131] Since the power generation cost of the energy storage unit is negligible, there is no cost function, and it mainly includes equipment cost and purchase electricity price cost.

[0132] The constraint conditions of the power grid power generation system model include power generation system power balance constraint based on the power grid load of the next day, maximum output constraint, minimum output constraint, continuous startup time constraint, continuous shutdown time constraint, unit ramp constraint, energy storage device constraint and synchronous condenser constraint;

[0133] 1), The constraint of the power generation system power balance based on the power grid load of the next day is:

[0134] ;

[0135] where, , represents the j-th hour, T = 24 hours, is the active power output of wind power in the j-th hour, is the active power output of photovoltaic power in the j-th hour, is the active power output of thermal power in the j-th hour, is the active power output of the energy storage in the j-th hour time period, is the load of the j-th hour obtained through the LSTM network;

[0136] When then ; when then ; ;

[0137] 2), the maximum constraint and minimum output constraint of each unit are:

[0138] ;

[0139] where are the minimum and maximum limits of the active power output of the i-th unit respectively, are the minimum and maximum limits of the reactive power of the active power output of the i-th thermal power unit respectively;

[0140] 3), the continuous startup constraint and shutdown time constraint are:

[0141] ;

[0142] 4), the unit ramp constraint is:

[0143] ;

[0144] where are the limits of the climbing and descending speeds of the active power output of the i-th unit per unit time period respectively;

[0145] 5), the energy storage device constraints include SOC constraint, charge and discharge power constraint, and charge and discharge capacity constraint. The SOC constraint is:

[0146] ;

[0147] where is the minimum value of SOC required during the operation of the energy storage device, is the maximum value of SOC required during the operation of the energy storage device;

[0148] ;

[0149] where is the maximum charging power required during the operation of the energy storage device, is the maximum discharge power required during the operation of the energy storage device;

[0150] The charge and discharge capacity constraint is:

[0151] ;

[0152] where is the number of energy storage units, is the output power of the i-th energy storage device, is the total power that the distribution network needs the energy storage device to provide, are the minimum and maximum output powers of the energy storage devices respectively;

[0153] 6), the synchronous condenser constraint is the node voltage constraint:

[0154] .

[0155] Among them, is the rated voltage.

[0156] S3. Based on the objective function and the constraint conditions, use the wolf pack optimization algorithm to obtain the optimal unit combination method for power grid dispatching.

[0157] The Grey Wolf Optimization Algorithm (GWO) is inspired by grey wolves. The GWO algorithm simulates the leadership hierarchy and hunting mechanism of grey wolves in nature. Four types of grey wolves, such as α, are used to simulate the leadership hierarchy. In addition, three main steps of hunting are implemented: searching for prey, surrounding prey, and attacking prey.

[0158] In order to mathematically model the social hierarchy of grey wolves when designing the GWO algorithm, we take the optimal solution as the α wolf. Therefore, the second and third best solutions are named β wolf and δ wolf respectively. The remaining candidate solutions are assumed to be ω wolves. In the GWO algorithm, the hunting process is guided by α, β, and δ, and the ω wolves follow these three wolves.

[0159] In this embodiment, the method for obtaining the optimal unit combination method through the grey wolf algorithm includes:

[0160] S31. Initialize the population, the convergence factor and the coefficient vectors A and C. The population initialization includes the initialization of the population size, the number of iterations, and the position of each grey wolf. In this embodiment, the position of each grey wolf is the power load distribution, and the dimension of each grey wolf is the number of units in the power system; the power grid operation cost value is the fitness function value, that is, the value of the objective function determined in the above embodiment;

[0161] S32. Calculate the fitness of each grey wolf individual and sort them, and save the top 3 wolves with the best fitness, which are the α wolf, the β wolf, and the δ wolf in turn;

[0162] S33. Calculate the distance between each grey wolf and the α wolf, the β wolf, and the δ wolf according to the following formula;

[0163] Gray wolves can identify the location of prey and surround them. After the gray wolves identify the location of the prey, the β wolves and δ wolves, under the leadership of the α wolf, guide the wolf pack to surround the prey. The mathematical model for an individual gray wolf to track the location of prey is described as follows:

[0164] ;

[0165] Where, , and represent the distances between α, β, and δ and other individuals respectively; , and represent the current positions of α, β, and δ respectively; , , are random vectors, is the position of the current gray wolf.

[0166] S34. Update the positions of the α wolf, β wolf, and δ wolf according to the following formula;

[0167] ;

[0168] The above formula defines the step size, direction, and the final position of ω towards α, β, and δ in the wolf pack respectively.

[0169] S35. Update the convergence factor and coefficients A and C according to the following formula;

[0170] ;

[0171] ;

[0172] The above formula represents the distance between an individual and the prey and the position update formula of the gray wolf. Where, is the current iteration number, and are coefficient vectors, and are the position vectors of the prey and the gray wolf respectively, is the convergence factor, which linearly decreases from 2 to 0 with the number of iterations, and take random numbers between [0, 1] in modulus.

[0173] When the prey stops moving, the gray wolves complete the hunting process by attacking. To simulate approaching the prey, the value of is gradually decreased, so the fluctuation range of The value also varies within the interval [- , . When 's value is within the interval, the next position of the grey wolf can be anywhere between its current position and the prey position. When , the wolf pack attacks the prey (falls into local optimum). When , the grey wolf separates from the prey and hopes to find a more suitable prey (global optimum).

[0174] The GWO algorithm also has another component to help discover new solutions. Since is a random value between [0, 2]. represents the random weight of the position of the wolf on the prey, indicating a large influence weight, and vice versa, indicating a small influence weight. This helps the GWO algorithm to behave more randomly and supports exploration, while avoiding falling into local optimum during the optimization process. Additionally, different from , decreases non-linearly. In this way, from the initial iteration to the final iteration, it provides a global search in the decision space. When the algorithm falls into local optimum and it is difficult to jump out, 's randomness plays a very important role in avoiding local optimum, especially in the final iteration where a global optimum solution needs to be obtained.

[0175] S36. Update the fitness values and positions of the alpha wolf, beta wolf, and delta wolf;

[0176] S37. Determine whether the maximum number of iterations is reached. If so, stop the iteration and output the optimal unit commitment and fitness function value. If not, repeat step S33.

[0177] To verify the rationality and economy of the grey wolf optimization algorithm results in the present invention, the particle swarm optimization algorithm will be used for comparison.

[0178] The basic principle of the particle swarm optimization algorithm is as follows:

[0179] Assume is the position of the i-th particle in the n-dimensional space during the k-th iteration, is the d-th component of particle i. Then, the velocity and position update process of the particle during the k + 1-th iteration are shown in the following formula.

[0180]

[0181]

[0182] In the formula, w is the inertia weight coefficient, c1 、c 2 is the learning factor, pbest i is the individual optimum of particle i, is its position at the d-th dimension during the k-th iteration process, gbest is the population optimum, is its position at the d-th dimension during the k-th iteration process, r 1 、r 2 is a random number uniformly distributed on [0, 1].

[0183] w adopts an adaptive linearly decreasing manner, that is:

[0184]

[0185] In the formula, k max is the maximum number of iterations, w ini 、w end are the initial inertia weight coefficient value and the inertia weight coefficient value at the last iteration respectively. Generally, w ini = 0.9, w end = 0.4.

[0186] To verify the rationality and effectiveness of the scheduling method provided by the present invention, this example uses the load data of a certain area in Heilongjiang as an example, uses the LSTM neural network for prediction, and the grey wolf optimization algorithm for optimization calculation.

[0187] There are many influencing factors for predicting the power generation of new energy grid units. In the method steps of predicting the load based on the LSTM network, the load is predicted in different time periods. Similar to the standard recurrent neural network, starting from t = 1, the forward propagation process of the input sequence x with a length of T is calculated, and the update equation is recursively applied while increasing. Starting from t = T, the backpropagation process is calculated using the gradient descent method to complete the reverse parameter tuning of the network. In this case, the number of iterations is set to 300 times. To accelerate the optimization speed, the initial learning rate is 0.005. To avoid falling into local optimum, the learning rate is updated after 125 iterations, and the learning rate decay factor is set to 0.2. To enhance the accuracy of the prediction results and reduce errors, the prediction results will be verified once every five iterations.

[0188] Rely on historical data to predict the daily grid load for 366 days, Figure 6 and Figure 7They respectively represent the power grid load prediction results and the actual power grid load on the 365th and 366th days in the power grid load prediction results based on LSTM. In the figure, the load prediction data is basically close to the true value, but there are still certain errors. Considering the limited test data, if the historical data can be increased, the accuracy can be improved, and considering the physical model of the power grid will further reflect the advantages of the LSTM network. Increasing the amount of test data can also more accurately predict the power grid load. The accurate prediction of the new energy power generation power lays the foundation for the implementation of unit grouping and active power dispatching.

[0189] In the present invention, a new energy power grid is constructed, which includes a thermal power unit, a wind power unit, a photovoltaic unit, and two energy storage units. The population size of the grey wolf optimization algorithm is set to 50, and the number of iterations is set to 500. The operating capacity constraints of each unit are shown in Table 1:

[0190] Table 1. Operating capacity constraints of each unit

[0191]

[0192] The time-of-use electricity price is 0.288, 0.288, 0.288, 0.288, 0.288, 0.288, 0.777, 0.777, 1.231, 1.231, 1.231, 0.777, 0.777, 1.231, 1.231, 0.777, 0.777, 0.777, 1.231, 1.231, 1.231, 0.288, 0.288, 0.288.

[0193] In this case, the grey wolf optimization algorithm is used to solve the objective function according to the load prediction results of the LSTM neural network. The results are as Figure 9 shown. At the same time, this case will be compared with the particle swarm optimization algorithm, as Figure 9 shown, to verify the rationality and economy of the results of this case. The comparison results with the particle swarm are shown in Table 2:

[0194] Table 2. Comparison results with the particle swarm

[0195]

[0196] Through comparison, we can see that the grey wolf optimization algorithm not only has a lower total cost than the particle swarm optimization algorithm, but also has a faster iteration speed, thus proving the accuracy and superiority of the model and algorithm adopted in this case, as well as the rationality of the method of this case.

[0197] The above embodiments are only illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A method for optimizing the dispatching of power grid generating units, characterized in that, it includes the following steps: S1. Predict the power grid load in a future time period through an LSTM neural network model; S2. Establish a power grid power generation system model. The power generation system model includes thermal power units, photovoltaic units, wind turbine units, energy storage devices and synchronous condensers. Establish an objective function with the lowest operating cost of the power generation system. The constraint conditions of the power grid power generation system model include power generation system power balance constraints based on the power grid load in a future time period, maximum output constraints, minimum output constraints, continuous startup time constraints, continuous shutdown time constraints, unit ramp constraints, energy storage device constraints and synchronous condenser constraints; The objective function is: ； where \(i\) is the number of each unit; \(j\) is the \(j\)-th period of unit operation; is the operating cost of thermal power units in the \(j\)-th period; is the operating cost of wind power units in the \(j\)-th period; is the operating cost of photovoltaic units in the \(j\)-th period; is the charge and discharge cost of energy storage equipment in the \(j\)-th period; is the operating cost of synchronous condensers in the \(j\)-th period; The operating cost of the thermal power unit includes energy consumption cost and environmental cost. The energy consumption cost is: ； Among them, is the energy consumption cost of the unit; is the number of units; is the number of time periods; is the output of the i-th thermal power unit in the j-th time period; is 1 or 0, representing whether the i-th unit outputs power in the j-th time period respectively; is the energy consumption cost for starting the i-th unit in the j-th time period; is the energy consumption cost for the output of the i-th unit; is the consumption coefficient of the unit; The environmental cost is the total penalty cost for pollutant emissions It is: ； Among them, is the penalty cost for carbon dioxide pollutant emissions; is the penalty cost for sulfur dioxide pollutant emissions; The penalty cost for carbon dioxide pollutant emissions and the penalty cost for sulfur dioxide pollutant emissions have the following relationship with the emission volume: ； ； Among them, is the emission of carbon dioxide pollutants; is the relationship coefficient between the carbon dioxide pollutant emission of the i thermal power unit and the output of the thermal power unit; is the emission of sulfur dioxide pollutants; is the relationship coefficient between the sulfur dioxide pollutant emission of the i thermal power unit and the unit output; The synchronous condenser cost is the reactive power call cost, which is: ； Among them, is the coefficient of the operating cost function of the synchronous condenser; is the reactive power output of the i-th synchronous condenser in the j-th time period; The synchronous condenser constraint is the node voltage constraint: ； Among them, is the rated voltage; S3. Based on the objective function and constraint conditions, use the wolf pack optimization algorithm to obtain the optimal unit combination method for power grid dispatching.

2. The method for optimizing the dispatching of power grid generating units according to claim 1, characterized in that, The operating cost of the wind turbine unit is: ； The cost of wind power generation is as follows: ； The curtailment cost is as follows: ； Among them, is the coefficient of the power generation cost function of the wind turbine; is the active power output of the i-th wind turbine in the j-th time period; is the cost coefficient of wind curtailment of the wind turbine; is the wind speed in the j-th time period; is the maximum value; The operating cost of the photovoltaic unit includes the power generation cost and the cost of light curtailment of the photovoltaic unit.

3. The method for optimizing the dispatching of power grid generating units according to claim 1, characterized in that, The operating cost of the photovoltaic unit is: ； For the power generation cost of the photovoltaic unit, it is as follows: ; The curtailment cost is as follows: ； Among them, is the coefficient of the power generation cost function of the photovoltaic unit; is the active power output of the i-th photovoltaic unit in the j-th time period; is the cost coefficient of the curtailment of photovoltaic power; is the solar radiation amount in the j-th time period; is the maximum value of.

4. The method for optimizing the dispatching of power grid generating units according to claim 1, characterized in that, 1) The constraint of the power generation system power balance based on the power grid load in a future time period is: ； Among them, , is the output of wind power in the j-th time period, is the output of photovoltaic power in the j-th time period, is the output of thermal power in the j-th time period, is the output of energy storage in the j-th time period, is the load in the j-th time period; When then ; When then ; 2) The maximum constraint and minimum output constraints of each unit are: ； Among them, are the minimum and maximum limits of the output of unit i respectively, are the minimum and maximum limits of the reactive power of the output of thermal power unit i respectively; 3) The continuous startup constraint and shutdown time constraint are: ； 4) The unit ramp constraint is: ； Among them, They are the limits of the output climbing and descending speeds of the i unit within a unit time period, respectively. 5) The energy storage device constraints include SOC constraint, charge and discharge power constraint, charge and discharge capacity constraint. The SOC constraint is: ； Among them, is the minimum SOC value required during the operation of the energy storage device, is the maximum SOC value required during the operation of the energy storage device; ； Among them, is the maximum charging power required during the operation of the energy storage device, is the maximum discharging power required during the operation of the energy storage device; The charge and discharge capacity constraint is: ； Among them, is the number of energy storage units, is the output power of the i-th energy storage device, is the total power that the distribution network needs the energy storage device to provide, are the minimum and maximum output powers of the energy storage device respectively.

5. The method for optimizing the dispatching of power grid generating units according to claim 1, characterized in that, The wolf pack optimization algorithm uses the operating cost of each unit as the position vector X of the gray wolf individual in the gray wolf optimization algorithm; The fitness function of the gray wolf optimization algorithm is the objective function.

Images