Optimized scheduling method based on artificial intelligence algorithm and new energy power generation punishment mechanism

Abstract

The invention relates to an optimal scheduling method based on an artificial intelligence algorithm and a new energy power generation punishment mechanism. Comprising the following steps: 1, constructing an HDEED model, and designing an objective function and constraint conditions of the HDEED model; 2, establishing a photovoltaic or wind power-containing dynamic scheduling system, and obtaining load demands and wind and light output at each moment in the system; step 3, adopting an MTSA algorithm to solve the model, and obtaining a Pareto optimal frontier solution; and 4, selecting a Pareto optimal compromise solution, and checking whether the scheme meets the system power balance constraint or not. A new energy power generation punishment mechanism is introduced, so that the system does not increase extra cost due to absorption of too much new energy power generation power, resource waste due to absorption of too little new energy power generation power by a power grid is avoided, and the scientificity of formulating a scheduling scheme by a decision maker is improved. On the basis of the TSA algorithm, an operation operator is additionally arranged, and the MTSA algorithm suitable for solving the multi-objective optimization problem is constructed. An initialization strategy of the algorithm is improved, and a random position disturbance strategy is introduced.

Description

Technical Field

[0001] The present invention belongs to the field of renewable energy power generation, constructs a hybrid dynamic emission economy (HDEED) model containing a renewable energy power generation penalty mechanism, and proposes a solution method of a novel artificial intelligence algorithm based on a multi-objective quilting swarm (MTSA). Background Art

[0002] As environmental pollution and fossil energy depletion become more and more prominent, people are paying more and more attention to the sustainability and low-carbon nature of the power industry. In addition, the scale of grid-connected renewable energy power generation has increased year by year, which has a significant impact on alleviating the current environmental pollution problems. It is urgent and of great significance to explore the impact of renewable energy on the power system. However, renewable energy power generation represented by photovoltaic and wind power generation has significant volatility and uncertainty, which affects the stable operation of the power grid and causes waste of renewable energy.

[0003] The power industry is a basic industry and pillar industry for the economic development of countries around the world. It is also a key area for fossil energy consumption and pollutant gas emissions. According to statistics, carbon emissions from coal accounted for more than 40% of global carbon dioxide emissions in 2021, and coal-fired power generation is the main component of carbon emissions from coal. Therefore, adjusting the structure of the power industry and developing low-carbon power and new energy technologies are of great significance to promoting the sustainable development of the power industry in various countries and promoting the development and utilization of clean energy. In addition, the power system is a relatively complex and large engineering system, and its optimization problem has always been a key research topic. As the basis and core of power system operation optimization, the optimization scheduling problem is a current research hotspot. Considering the huge scale of the power system, improving the operating efficiency of related systems through optimized scheduling can not only bring significant economic benefits, but also have important value in promoting the transformation and upgrading of the power industry and sustainable development.

[0004] Optimal dispatch of power systems refers to clarifying the start and stop status of units and adjusting the output distribution of each generating unit under specific constraints in order to seek a decision-making plan that makes the system's various indicators optimal. Optimal dispatch in the traditional sense takes the optimal economic indicator as the primary basis for decision-making, also known as economic dispatch. However, with the grid connection of new energy sources, the number of decision variables in the dispatch model continues to increase, the applicability of traditional economic dispatch models has deteriorated, and the resulting dispatch plans are difficult to meet actual needs. To this end, researchers have developed a variety of new dispatch models based on different application scenarios and actual needs.

[0005] According to the time coupling, the power system optimization dispatch is divided into static dispatch and dynamic dispatch. The power system static dispatch only optimizes the active output of the unit at a fixed time, aiming to obtain the optimal output allocation plan of the unit under the premise of minimizing the system's pollutant gas emissions, so as to maximize the economic benefits of the system.

[0006] Static dispatch can only provide a unit output allocation plan for a certain dispatch period, but in practice it is often necessary to formulate a dispatch plan for a period. Dynamic dispatch of power systems solves the limitations of static dispatch by considering the coupling between time sections and changes in load demand. Dynamic dispatch of power systems aims to obtain a decision-making plan that makes the system objectives optimal in a future dispatch cycle under the premise of meeting dynamic constraints such as unit ramp rate.

[0007] From the perspective of the number of optimization objectives, power system optimization dispatching is divided into single-objective optimization dispatching and multi-objective optimization dispatching. Traditional economic optimization dispatching belongs to single-objective optimization dispatching, while multi-objective optimization seeks to make the decision-making plan that makes the system's multiple indicators comprehensively optimal. Emission economic dispatching belongs to multi-objective optimization dispatching, which takes energy conservation and pollution gas emission reduction as the primary decision-making goals of power dispatching, which is more in line with the current development trend of power dispatching. With the increasingly prominent problems of environmental pollution and fossil energy depletion, people are paying more and more attention to the sustainability and low-carbon nature of the power industry. In addition, the scale of new energy power generation connected to the grid has increased year by year, which has a significant impact on alleviating the current environmental pollution problems. It is urgent and of great significance to explore the impact of new energy on the power system.

[0008] In summary, the non-stationary random characteristics of wind and solar power output make traditional dispatching methods unable to meet the actual needs of wind and solar grid connection. In order to improve the utilization rate of wind and solar energy, it is urgent to study new power dispatching theories and solution methods. Summary of the invention

[0009] The technical problem solved by the present invention is: a HDEED model containing a penalty mechanism for renewable energy generation is constructed, and a solution method based on the MTSA algorithm is proposed. The intermittent and random nature of photovoltaic and wind power output will cause the problem of overestimation or underestimation of wind power and photovoltaic output, resulting in additional power generation costs and pollutant gas emissions. The introduction of a penalty mechanism for renewable energy generation will prevent the system from increasing additional costs due to the absorption of too much renewable energy power generation and grid-connected power, nor will it cause resource waste due to the grid absorbing too little renewable energy power generation, thereby improving the scientificity and operability of the scheduling scheme formulated by decision makers. In view of the characteristics of strong coupling, non-convexity, high nonlinearity, and multi-objective optimization of the HDEED optimization problem, the present invention, based on the TSA algorithm, constructs an MTSA algorithm suitable for solving multi-objective optimization problems by adding running operators. At the same time, in order to improve the solution performance of the MTSA algorithm, the initialization strategy of the algorithm is improved, and a random position perturbation strategy is introduced.

[0010] The present invention solves the technical problem by adopting the following scheme:

[0011] The HDEED model with the penalty mechanism for renewable energy generation and the solution method based on the MTSA algorithm includes the following steps:

[0012] Step 1: Construct a HDEED model with a penalty mechanism for renewable energy generation, and design the objective function and constraints of the HDEED model.

[0013] Step 2: Based on the IEEE-10 unit 39-node thermal power system, a dynamic scheduling test system including photovoltaic or wind power is established to obtain the load demand and photovoltaic and wind farm output at each scheduling time in the dynamic scheduling test system.

[0014] Step 3: Use the MTSA algorithm to solve the constructed dynamic scheduling model and obtain the Pareto optimal frontier solution of the dynamic scheduling model.

[0015] Step 4: Select the best compromise solution in the Pareto optimal frontier based on fuzzy theory, and check whether the resulting scheduling scheme meets the system power balance constraint.

[0016] The process of the MTSA algorithm suitable for solving multi-objective optimization problems includes:

[0017] In the process of constructing the optimization and scheduling model of the integrated energy system, the present invention not only considers the cost target of the system, but also considers the pollutant emission target of the system. The optimization and scheduling of the wind-solar-fire integrated energy system containing energy storage units is a strongly coupled, non-convex, highly nonlinear, multi-objective optimization problem, which has high requirements on the convergence performance of the solution algorithm. To this end, the present invention constructs a multi-objective quilt swarm algorithm (MTSA) based on the elite set guidance mechanism.

[0018] The MTSA algorithm mainly includes three parts: the dominance relationship judgment criterion, the elite set guidance mechanism and the population generation. The specific principles are as follows:

[0019] (1) Criteria for determining dominance

[0020] The dominance relationship judgment criterion is the core of the MTSA algorithm. In the MTSA algorithm, due to the introduction of the concept of dominance level, the dominance relationship between two solutions cannot be judged by simply comparing the size of the objective function. It is necessary to combine the concepts of dominance level and crowding distance to realize the judgment of the dominance relationship between solutions. The dominance relationship judgment criterion between solutions in the MTSA algorithm is defined as

[0021]

[0022] Among them, R x rand represents the dominance level of the xth individual; P y rank represents the dominance level of the yth individual; C x represents the crowding distance of the xth individual; C y represents the crowding distance of the y-th individual.

[0023] It can be seen from the above formula that in the MTSA algorithm, when judging the superiority relationship between two solutions, the solution with a smaller dominance level is better than the solution with a larger dominance level; if the dominance levels of the two solutions are the same, the crowding distances of the two solutions are compared, that is, the solution with a larger crowding distance is better than the solution with a smaller crowding distance.

[0024] (2) Elite Collection Guidance Mechanism

[0025] In the elite set guidance mechanism, the crowding distance of the individuals in the quilt group under different dominance levels is first calculated; then, the individual crowding distances are sorted and the top 5% individuals are selected as the elite set; finally, an individual is randomly selected from the elite set as the optimal individual, and the positions of other individuals in the quilt group are updated accordingly. The optimal solution selection method based on the elite set guidance mechanism is more reasonable, avoiding the optimal solution being limited to a single fixed solution, and improving the distribution of non-dominated solutions. The formula for calculating the individual crowding distance in the elite set guidance mechanism is:

[0026]

[0027] Among them, C x represents the crowding distance of the x-th solution; N p Indicates the population size; N o represents the number of targets; f y x+1 and f y x-1 Represent the target values ​​of the x+1th and x-1th individuals respectively; f y max and f y min Represent the maximum and minimum values ​​of the y-th target respectively.

[0028] (3) Population generation

[0029] The population strategy used by the MTSA algorithm is a dual population strategy that includes a parent population and a child population. Before the first iteration of the algorithm, the parent population is initialized and the dominance level of the individuals in the parent population is assigned (level 1 is the highest, level 2 is the second, and level 3 is the lowest). After the first iteration of the algorithm, the child population can be obtained. The parent population and the child population are merged to generate a scale of 2N. p Secondly, the merged population is non-dominated, the dominance level of the population is assigned, the crowding distance of the individuals in the population is calculated, and the next generation of the population is selected.

[0030] The optimal dispatch of power systems under the grid connection of renewable energy is a multi-constrained, nonlinear, and non-convex multi-objective optimization problem. When solving such optimization problems, it is necessary not only to consider the power balance relationship between load, renewable energy, and thermal power, but also to deal with complex constraints, which places high demands on the algorithm's solution performance. To this end, this study introduces a population position initialization method based on a chaotic mapping strategy and a random position perturbation strategy into the MTSA algorithm.

[0031] (1) Population position initialization method based on chaos mapping strategy

[0032] In the process of solving the power system dispatching model, in order to obtain the initial population position with higher quality, speed up the convergence of the algorithm, and reduce the calculation cost of the algorithm, this section uses the randomness and ergodicity of the chaotic mapping strategy to obtain the initial population position:

[0033]

[0034] Among them, c s is a random number in the interval [0, 1]; Represents the maximum boundary of the search space; represents the minimum boundary of the search space; ρ represents the scale of the chaotic mapping; when ρ=4, the model is in a chaotic state.

[0035] (2) Random Position Perturbation Strategy

[0036] The positions of individuals in the tunicate group are dynamically disturbed through the random position perturbation strategy to ensure that the population maintains a high diversity throughout the entire solution process, thereby increasing the probability of the algorithm avoiding local extreme solutions and improving the solution effect of the scheduling model.

[0037] The principle of the random position perturbation strategy is: during the iteration of the MTSA algorithm, the perturbation frequency f is set d By comparing f with a random number r in the interval [0, 1] d The value of and r determines whether to disturb the population position. Specifically, when f d ≤r t When f d >r t When , the individual positions in the quilt group are updated according to the formula.

[0038]

[0039] Among them, s c is the convergence coefficient, which is used to guide the individual convergence direction; s c min( s c min =0.3) is the minimum value of the convergence coefficient; sc max (s c max =0.7) is the maximum convergence coefficient; randn(0, σ 2 ) has a mean of 0 and a variance of σ 2 Gaussian distribution of T best represents the optimal individual position; f best represents the optimal individual fitness value; f T Represents the current individual fitness value.

[0040] The specific steps of the HDEED model of the present invention including the penalty mechanism for renewable energy power generation and the solution method based on the MTSA algorithm include:

[0041] Construct a HDEED model with a penalty mechanism for renewable energy generation, and design the objective function and constraints of the HDEED model.

[0042] In actual situations, due to the intermittent and random nature of photovoltaic and wind power output, there is a certain deviation between the model's predicted output and the actual output, and overestimation or underestimation of wind power and photovoltaic output will bring additional power generation costs and pollutant gas emissions, which will have a negative impact on the dispatch economy and environmental protection. Therefore, the present invention introduces a new energy power generation penalty mechanism into the HDEED model, and systematically explores the impact of photovoltaic and wind power output forecast deviation on the dispatch results from the two aspects of economy and environmental protection.

[0043] 1) Objective function of HDEED model

[0044] At present, countries around the world are setting strategic goals of "carbon peak" and "carbon neutrality". In this context, it is particularly necessary to improve the environmental friendliness of the system. Therefore, when constructing the HDEED model containing new energy power generation and energy storage units, the present invention considers the two indicators of system power generation cost and pollutant gas emissions in the objective function, and explores the dynamic scheduling scheme that optimizes the economic benefits and environmental benefits of the system.

[0045] (1) Power generation cost objective function

[0046] The power generation cost of the power system generally refers to the fuel cost consumed by thermal power units. For the HDEED model, in addition to the fuel cost, it also includes the operation and maintenance costs of photovoltaic power stations and wind farms, as well as the penalty costs caused by prediction errors.

[0047]

[0048] Among them, T dis is a scheduling cycle (24 hours in this paper); t is a scheduling period (1 hour in this paper); N f is the number of thermal power units; represents the minimum active output of the i-th thermal power unit at time t; a i , b i 、c i 、e i 、f i They represent the fuel cost coefficient of the i-th thermal power unit respectively; represents the active power output of the i-th thermal power unit at time t; N pv is the number of photovoltaic power generation units; N w is the number of wind turbines; N soc is the number of energy storage units; represents the active output of the i-th photovoltaic power generation unit at time t; represents the active output of the i-th wind turbine at time t; represents the operation and maintenance cost coefficient of the i-th photovoltaic power generation unit at time t; represents the operation and maintenance cost coefficient of the i-th wind turbine at time t; The penalty cost function for underestimating the output of the i-th wind turbine; The reserve penalty cost function generated for overestimating the output of the i-th wind turbine; The penalty cost function for underestimating the output of the i-th photovoltaic power generation unit; The backup penalty cost function is generated for overestimating the output of the i-th PV unit.

[0049] The mathematical model is as follows:

[0050]

[0051] in, is the operation and maintenance cost coefficient of the i-th wind turbine; e p，w Penalize cost coefficient for wind power output; is the rated output of the i-th wind turbine; f w (w) is the probability density function of wind power output; e r，w is the standby penalty cost coefficient; is the operation and maintenance cost coefficient of the ith photovoltaic power generation unit; e p，pv Penalize cost factor for PV output; is the rated output of the ith photovoltaic power generation unit; f pv (pv) is the probability density function of photovoltaic output; e r，pv The penalty cost coefficient for photovoltaic output reserve; is the charging power of the energy storage unit; is the discharge power of the energy storage unit.

[0052] (2) Pollution gas emission objective function

[0053]

[0054] Among them, α i , β i , γ i ,δ i , η i is the pollutant gas emission coefficient; is the penalty emission function caused by underestimating the output of the i-th wind turbine; is the reserve penalty emission function caused by overestimating the output of the i-th wind turbine; The penalty emission function generated for underestimating the output of the i-th photovoltaic power generation unit; Reserve penalty emission function generated for overestimating the output of the i-th PV unit.

[0055] The mathematical model is

[0056]

[0057] Among them, d p，w Penalize emission factors for wind power output; is the rated output of the i-th wind turbine; f w (w) is the probability density function of wind power output; d r，w Penalize emission factor for wind power output reserve; d p，w Penalize emission factors for photovoltaic output; is the rated output of the ith photovoltaic power generation unit; f pv (pv) is the probability density function of photovoltaic output; d r，pv Penalize emission factors for photovoltaic output.

[0058] 2) Constraints of the HDEED model

[0059] In order to ensure the scientificity and feasibility of the obtained scheduling strategy, without significantly increasing the complexity of the model, this study focuses on the following constraints:

[0060] (1) Power balance constraints

[0061]

[0062] Among them, P t load represents the system load demand at time t; P t loss represents the line loss at time t, and its expression is

[0063]

[0064] Among them, B * ij , B * i , B * is the line loss parameter.

[0065] When solving the dynamic scheduling model, the loss parameter B * i With B * The impact on the system scheduling results is small. To simplify the model calculation, the above two parameters are ignored when solving the model.

[0066] (2) Active power output constraints of thermal power units

[0067]

[0068] in, Indicates the lower limit of active output of the i-th thermal power unit at time t; It represents the upper limit of active output of the i-th thermal power unit at time t.

[0069] (3) Active output constraints of wind farms and photovoltaic power stations

[0070]

[0071] in, and They are the maximum output of wind farm and photovoltaic power station respectively.

[0072] (4) Climbing rate constraint

[0073] The ramp rate constraint refers to the output that each thermal power unit can reduce or increase within a unit time.

[0074]

[0075] Among them, T * (T * =1) indicates the time interval of each period; are the ramp-down rate and ramp-up rate of the i-th thermal power unit respectively.

[0076] (5) Positive and negative spinning reserve constraints

[0077] Traditionally, spinning reserve constraints are used to deal with the adverse effects of unit shutdown and load forecast errors by setting a certain reserve capacity. However, for wind and solar grid-connected power systems, it is necessary not only to deal with the impact of load forecast errors, but also to deal with the impact of uncertainty in photovoltaic and wind power output. Positive and negative spinning reserve constraints are the most direct way to deal with wind and solar uncertainty.

[0078] Positive spinning reserve constraints:

[0079]

[0080] in, represents the positive spinning reserve capacity provided by the i-th thermal power unit at time t; r l1 It represents the positive spinning reserve capacity factor of the thermal power unit in response to unit outage or load forecast error; r w1 It represents the positive spinning reserve capacity factor of thermal power units to cope with wind power output forecast error; r p1 It represents the positive spinning reserve capacity factor of thermal power units to cope with the photovoltaic output forecast error.

[0081] Negative spinning reserve constraints:

[0082]

[0083] in, represents the negative spinning reserve capacity provided by the i-th thermal power unit at time t; r l2 It represents the negative spinning reserve capacity factor of the thermal power unit in response to unit outage or load forecast error; r w2 It represents the negative spinning reserve capacity factor of thermal power units to cope with the wind power output forecast error; r p2 It represents the negative spinning reserve capacity factor of thermal power units to cope with the photovoltaic output forecast error.

[0084] (6) Charge and discharge power constraints of energy storage units

[0085]

[0086] Among them, p t char，max With P t dischar，max They are the maximum charging power and maximum discharging power of the energy storage unit respectively.

[0087] In addition, the energy storage unit cannot be charged and discharged at the same time during operation, so it should also meet

[0088]

[0089] (7) Charge ratio constraints of energy storage units

[0090]

[0091] Among them, Q t soc，max is the upper limit of charge ratio; Q t soc，min is the lower limit of charge ratio; Q t soc is the charge ratio of the energy storage unit at time t, calculated as follows:

[0092]

[0093] Among them, Q t r is the remaining power of the energy storage unit; Q t max It is the maximum storage capacity of the energy storage unit.

[0094] In addition, in order to ensure the sustainability of charging and discharging when the energy storage unit participates in scheduling, the sum of the charging power of the energy storage unit should be equal to the sum of the discharging power within a scheduling cycle, that is,

[0095]

[0096] 3) Conceptual scheme of HDEED multi-objective optimization

[0097] The HDEED model constructed above belongs to a multi-objective minimization problem. Its purpose is to explore the optimal load distribution of the generator set based on the unit operation characteristic equation under given load levels and wind and solar grid conditions, and to seek a static dispatching scheme that optimizes the economic and environmental benefits of the system. Based on the analysis of 1) and 2), the HDEED mathematical model description formula taking into account the penalty mechanism for renewable energy generation is as follows, and its specific mathematical model is shown in Table A1.

[0098]

[0099] Among them, O cost With O emission They are the power generation cost target and pollutant gas emission target of the dynamic scheduling model taking into account the penalty mechanism for renewable energy power generation; E and I refer to the equality constraint and inequality constraint of the dynamic scheduling model respectively.

[0100] Table A1. Relevant parameter settings of IEEE-10 unit test system

[0101]

[0102] (1) Based on the IEEE-10 unit 39-node thermal power system, a dynamic scheduling test system containing photovoltaic or wind power is established to obtain the load demand and the output of photovoltaic and wind farms at each scheduling time in the dynamic scheduling test system. Most of the current studies are based on the IEEE-10 unit 39-node thermal power system to establish a dynamic scheduling test system containing photovoltaic or wind power. To this end, the present invention is based on a photovoltaic power station and a wind farm in China, combined with the IEEE-10 unit thermal power system to establish a dynamic scheduling test system, and verify the constructed dynamic scheduling model and solution algorithm. The relevant parameter settings of the IEEE-10 unit 39-node thermal power system in the test system, the pollutant gas emission parameters and the transmission loss parameters are shown in Tables A2 and A3 below.

[0103] Table A2. Pollutant gas emission parameters for IEEE-10 unit test system

[0104]

[0105]

[0106] Table A3. Transmission loss parameters of IEEE-10 unit test system

[0107]

[0108] (2) Use the MTSA algorithm to solve the constructed dynamic scheduling model and obtain the Pareto solution set of the dynamic scheduling model

[0109] The constructed dynamic scheduling model is solved using the MTSA algorithm. When the algorithm termination condition is reached, the individuals in the quilt group are the solutions in the Pareto solution set of the dynamic scheduling model.

[0110] The objective function of the dynamic scheduling model is defined as the fitness function F of the MTSA algorithm, that is,

[0111] F(p i )=[O cls t (pi), O emission (p i )]

[0112] The specific steps of the MTSA algorithm suitable for solving multi-objective optimization problems are:

[0113] (1) Initialize MTSA algorithm parameters, such as population size, maximum number of iterations, etc. Based on a wind farm and a photovoltaic power station in China, a dynamic scheduling test system is established in combination with the IEEE-10 unit 39-node thermal power system.

[0114] (2) Obtain the grid-connected power of photovoltaic and wind power in the next scheduling cycle and construct the initial population, that is,

[0115]

[0116] Among them, T dis Indicates the scheduling period, which is 24 hours.

[0117] (3) Divide the parent population into dominance levels and calculate the crowding distance of individuals in the population.

[0118] In the MTSA algorithm, due to the introduction of the concept of dominance level, the dominance relationship between two solutions cannot be judged by simply comparing the size of the objective function. It is necessary to combine the concepts of dominance level and crowding distance to realize the judgment of the dominance relationship between solutions. The judgment criterion of the dominance relationship between solutions in the MTSA algorithm is defined as

[0119]

[0120] Among them, R x rank represents the dominance level of the xth individual; P y rnnt represents the dominance level of the yth individual; C x represents the crowding distance of the xth individual; C y represents the crowding distance of the y-th individual.

[0121] It can be seen from the above formula that in the MTSA algorithm, when judging the superiority relationship between two solutions, the solution with a smaller dominance level is better than the solution with a larger dominance level; if the dominance levels of the two solutions are the same, the crowding distances of the two solutions are compared, that is, the solution with a larger crowding distance is better than the solution with a smaller crowding distance.

[0122] The calculation formula of individual crowding distance in the elite set guidance mechanism is:

[0123]

[0124] Among them, C x represents the crowding distance of the x-th solution; N p Indicates the population size; N o represents the number of targets; f y x+1 and f y x-1 Represent the target values ​​of the x+1th and x-1th individuals respectively; f y max and f y min Represent the maximum and minimum values ​​of the y-th target respectively.

[0125] (4) Sort the crowding distances of the individuals and select the top 5% of the individuals as the elite set; then randomly select an individual from the elite set as the optimal individual, and use this to update the positions of other individuals in the tunicate.

[0126] During the iteration of the MTSA algorithm, the disturbance frequency f is set d By comparing f with a random number r in the interval [0, 1] d The value of and r determines whether to disturb the population position. Specifically, when f d ≤r tWhen f d >r t When , the individual positions in the quilt group are updated according to the following formula.

[0127]

[0128] Among them, s c is the convergence coefficient, which is used to guide the individual convergence direction; s c min (s c min =0.3) is the minimum value of the convergence coefficient; s c max (s c max =0.7) is the maximum convergence coefficient; randn(0, σ 2 ) has a mean of 0 and a variance of σ 2 Gaussian distribution of T best represents the optimal individual position; f best represents the optimal individual fitness value; f T Represents the current individual fitness value.

[0129] (5) After obtaining the offspring population from step 4, merge the parent population and the offspring population, divide the dominance level of the individuals in the merged population, and calculate the individual crowding distance. Based on the dominance level and crowding distance, generate the next generation population and calculate the Fit of the individuals in the population.

[0130] (6) Determine whether the termination condition of the MTSA algorithm is met. If the termination condition is met, terminate the algorithm and obtain the Pareto optimal frontier of the dynamic scheduling model; otherwise, repeat steps 3 to 6 until the termination condition is met.

[0131] The beneficial effects of the present invention are:

[0132] (1) Based on the TSA algorithm, the present invention constructs the MTSA algorithm suitable for solving multi-objective optimization problems by adding a running operator. In order to improve the solving performance of the MTSA algorithm, the initialization strategy of the algorithm is improved, and a random position perturbation strategy is introduced.

[0133] (2) Compared with existing multi-objective optimization algorithms, the MTSA algorithm can obtain the most competitive solution. For multi-objective test functions, the Pareto frontier obtained by the algorithm has good distribution and diversity.

[0134] (3) For the HDEED problem, the system pollutant gas emissions and power generation costs obtained by the MTSA algorithm are both optimal, indicating that for a dynamic scheduling model with strong coupling and strong nonlinear characteristics that takes into account the penalty mechanism for renewable energy power generation, the scheduling scheme obtained by the MTSA algorithm takes into account both system economy and environmental protection.

[0135] (4) The penalty costs for overestimating or underestimating the output of photovoltaic and wind power are higher than the operation and maintenance costs of wind farms and photovoltaic power stations, and the penalty costs account for about 2.1% of the total costs, but at the same time they also improve the feasibility of the scheduling plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0136] Figure 1 It is a flow chart of solving the HDEED problem based on the MTSA algorithm and including a new energy power generation penalty mechanism according to the present invention.

[0137] Figure 2 It is the economic benefit of scheduling scheme 1 at each scheduling time obtained by each algorithm in Example 1 and the environmental benefit of scheduling scheme 2 obtained.

[0138] Figure 3 This is the detailed information of the scheduling scheme obtained by each algorithm in Example 1.

[0139] Figure 4 It is the output of photovoltaic and wind farms at each scheduling moment in the dynamic scheduling test system of the present invention.

[0140] Figure 5 This is the detailed information of the scheduling solutions obtained by each algorithm in Example 2.

[0141] Figure 6 It is the economic benefit of the scheduling scheme 1 at each scheduling time obtained by each algorithm in Example 2 and the environmental benefit of the scheduling scheme 2 obtained.

[0142] Figure 7 It is the Pareto optimal frontier of the dynamic scheduling model under different wind-solar grid-connected scales obtained by the MTSA algorithm in Example 3.

[0143] Figure 8 It is the system power generation cost and pollutant gas emission in each time period under different wind-solar grid-connected scales obtained by the MTSA algorithm in Example 3.

[0144] Fig. 9 It is the system power generation cost and pollutant gas emission at each scheduling time obtained by the MTSA algorithm in Example 3. DETAILED DESCRIPTION

[0145] In order to illustrate the technical solution of the present invention, specific embodiments will be described below in conjunction with the accompanying drawings.

[0146] The present invention discloses an HDEED model containing a penalty mechanism for renewable energy power generation based on an MTSA algorithm solving method. The HDEED model containing a penalty mechanism for renewable energy power generation is constructed, the objective function and constraint conditions of the HDEED model are designed, a dynamic scheduling test system containing photovoltaic or wind power is established based on a 39-node thermal power system of an IEEE-10 unit, the constructed dynamic scheduling model is solved by using the MTSA algorithm, the Pareto optimal frontier solution of the dynamic scheduling model is obtained, and finally the best compromise solution in the Pareto optimal frontier is selected based on fuzzy theory, and it is checked whether the obtained scheduling scheme meets the system power balance constraint.

[0147] Figure 1 The solution process of the HDEED problem with a penalty mechanism for renewable energy generation constructed based on the MTSA algorithm of the present invention is as follows: first, the MTSA algorithm parameters are initialized, and a dynamic scheduling test system is established based on a certain wind farm and a certain photovoltaic power station in China, combined with the IEEE-10 unit 39-node thermal power system → the photovoltaic and wind power grid-connected power in the next scheduling cycle is obtained, and the initial population is constructed → the objective function of the dynamic scheduling model is defined as the fitness function F of the MTSA algorithm → the dominance level of the parent population is divided, and the crowding distance of individuals in the population is calculated → whether the termination condition of the MTSA algorithm is met. If the algorithm termination condition is met, the algorithm is terminated and the Pareto optimal frontier of the dynamic scheduling model is obtained; otherwise, steps 4 to 5 are repeated until the termination condition is met → the best compromise solution in the Pareto optimal frontier is selected based on fuzzy theory, and the obtained scheduling scheme is checked to see whether it meets the system power balance constraint.

[0148] (1) Embodiment 1

[0149] Dynamic dispatch model without wind and solar grid-connected power and energy storage units

[0150] Table 1 shows the solution results of the dynamic scheduling model in Example 1 using MTSA, MSSA and MOMVO algorithms.

[0151] Table 1. Results of the dynamic scheduling model obtained by each algorithm in Example 1

[0152]

[0153] Scheme 1 in Table 1 is the boundary solution that only considers economic efficiency, Scheme 2 is the boundary solution that only considers environmental protection, and Scheme 3 is the best compromise solution selected based on fuzzy theory.

[0154] As shown in Table 1, for Scheme 1, if only the system economy is considered, the system power generation cost obtained by the MTSA algorithm is 2.47×10 6 $, which is lower than the 2.51×10 of the MOMVO algorithm. 6$2.48×10 with the MSSA algorithm 6 If only the environmental protection index of the system is considered, the MTSA algorithm also obtains the best scheduling result, and the minimum pollutant gas emission is 2.92×10 5 In addition, for the best compromise solution, the system power generation cost obtained by MTSA and MSSA algorithms is 2.51×10 6 $, but the system pollutant gas emissions obtained by the MTSA algorithm are the lowest, which are 9000lb and 6000lb lower than those of the MOMVO algorithm and MSSA algorithm respectively, indicating that the economic and environmental indicators of the scheduling scheme obtained by the MTSA algorithm are the best, which can provide decision makers with a reliable and scientific decision-making basis.

[0155] Figure 2 The economic benefits of the scheduling scheme 1 and the environmental benefits of the scheduling scheme 2 obtained by each algorithm at each scheduling time in Example 1 are shown. Figure 2 It can be seen that if only the economy of the system is considered, the scheduling scheme obtained by the MTSA algorithm has the best economy. At each scheduling time, the operating cost of the scheduling scheme obtained by the MTSA algorithm is less than that of the MOMVO and MSSA algorithms. If only the environmental protection of the system is considered, the pollutant gas emissions at each scheduling time obtained by the MTSA algorithm are lower than those of other algorithms. The results show that the scheduling scheme obtained by the MOTSAD algorithm can maximize the economy and environmental protection of the scheduling scheme while meeting the diverse needs of decision makers.

[0156] Table 2 shows the distribution of unit output in each time period corresponding to the scheduling scheme obtained by the MTSA algorithm in Example 1. As can be seen from Table 2, the output fluctuations of units U1 to U4 in each time period are large. The reason is that units U1 to U4 undertake peak load regulation tasks to meet the load requirements at each scheduling time during the dynamic scheduling process; units U5 to U10 are basically in full load power generation at each scheduling time.

[0157] Table 2. Results of the dynamic scheduling model obtained by each algorithm in Example 1

[0158]

[0159] Figure 3 The detailed information of the scheduling scheme obtained by each algorithm in Example 1 is displayed, including the output of each unit in each time period and the line loss in each time period. Figure 3Each stacked bar chart represents the sum of the outputs of all units at the scheduling moment, and the bars of different colors represent the outputs of different units. At each scheduling moment, the total output of the units obtained by each algorithm can meet the sum of the load demand and the line loss, indicating that the scheduling schemes corresponding to the best compromise solutions obtained by each algorithm meet the system power balance constraints. In addition, at different scheduling moments, the line loss varies, and its value fluctuates between 18MW and 93MW. In particular, during peak load periods, the line loss value even exceeds the output of smaller units. Therefore, the impact of line loss must be considered when solving the dynamic scheduling model, otherwise the resulting scheduling scheme cannot truly reflect the actual operating conditions of the system.

[0160] (2) Embodiment 2

[0161] Dynamic dispatch model with different scales of wind and solar grid-connected power but without energy storage units

[0162] Table 3 shows the solution results of the dynamic dispatch model in Example 2 using the MTSA, MSSA and MOMVO algorithms, where the wind and solar grid-connected power is as follows: Figure 4 as shown in .

[0163] Table 3. Results of the dynamic scheduling model obtained by each algorithm in Example 2

[0164]

[0165] As shown in Table 3, for the boundary solution, if only the system economic indicators are considered, the lowest system power generation cost obtained by the MTSA algorithm is 2.28×10 6 $, which is 5×10 lower than that of MOMVO and MSSA algorithms respectively. 4 $, 2×10 4 $; If only the environmental protection of the system is considered, the minimum system pollutant gas emission obtained by the MTSAD algorithm is 2.49×10 5 lb. For the best compromise solution, the system pollutant gas emissions and power generation costs obtained by the MTSA algorithm are both optimal, which are 2.55×10 5 lb and 2.33×10 6 In summary, the MTSA algorithm still has a strong solution performance for the dynamic scheduling model with strong coupling and strong nonlinear characteristics and the penalty mechanism of renewable energy generation, and the resulting scheduling scheme takes into account both system economy and environmental protection.

[0166] Table 4 shows the distribution of unit output in each time period corresponding to the scheduling scheme obtained by the MTSA algorithm in Example 2. From the data in Table 4, it can be seen that the output fluctuations of units U1 to U4 in each time period are relatively large. The reason is that units U1 to U4 undertake peak load regulation tasks to meet the load requirements at each scheduling time during the dynamic scheduling process; units U5 to U10 are basically in full load power generation at each scheduling time.

[0167] Table 4. Results of the dynamic scheduling model obtained by each algorithm in Example 2

[0168]

[0169] Figure 5 The detailed information of the dispatching scheme obtained by each algorithm in Example 2 is shown, including the output of each unit in each time period and the line loss in each time period. Figure 5 It can be seen that the dispatching schemes corresponding to the best compromise solutions obtained by each algorithm all meet the power balance constraints of the system, that is, the sum of the output of each unit and the renewable energy power generation in each period within a dispatching cycle is equal to the sum of the load demand and network loss in that period, which verifies the effectiveness of the obtained strategy.

[0170] Figure 6 The economic benefits of the scheduling scheme 1 and the environmental benefits of the scheduling scheme 2 obtained by each algorithm at each scheduling time in Example 2 are shown. Figure 6 It can be seen that at each dispatching time, the system power generation cost and pollutant gas emissions obtained by the MTSA algorithm are lower than those of other algorithms. It can also be seen from the partial enlarged figure that the system pollutant gas emissions obtained by the MTSA algorithm are smaller, indicating that the dispatching scheme obtained by the MTSA algorithm can maximize the environmental protection of the system while ensuring the economic benefits of the system.

[0171] Example 3

[0172] Dynamic dispatch model including wind and solar grid-connected power and energy storage units

[0173] Table 5 shows the solution results of the dynamic scheduling model under different wind and solar grid connection scales in Example 3 using the MTSA algorithm.

[0174] Table 5. Results of dynamic dispatch model solutions for different grid-connected scales of wind power and photovoltaic power generation

[0175]

[0176] As can be seen from Table 5, with the continuous improvement of the grid penetration rate, the system power generation cost and pollutant gas emissions are significantly reduced, indicating that the higher the grid penetration rate, the more obvious the improvement effect on the economic and environmental benefits of the system. However, the higher the wind and solar grid penetration rate is, the better it is. When the penetration rate is close to 30%, the dispatch model cannot be solved. This is because the rotating reserve capacity of the thermal power unit can no longer meet the fluctuations in photovoltaic and wind power output, that is, it cannot meet the positive and negative rotating reserve constraints of the system. In the process of constructing the dynamic dispatch model, the positive and negative rotating reserve constraints are used to deal with the impact of overestimation or underestimation of wind and solar output on the system, so that sufficient spare capacity is reserved in the formulated dispatch plan to deal with the situation where the future photovoltaic or wind power output forecast value deviates from the actual value. It can be seen that while improving the wind and solar grid penetration rate, a variety of factors should be considered comprehensively, especially the impact of the uncertainty and volatility of photovoltaic and wind power output on the system.

[0177] Figure 7 The Pareto optimal frontier obtained by solving the dynamic dispatch model under different wind and solar grid connection scales in Example 3 using the MTSA algorithm is shown. Figure 7 As shown in the figure, as the scale of grid connection changes, the system power generation cost and pollutant gas emissions also change accordingly, and the larger the scale of grid connection, the more obvious the improvement of system economic and environmental protection indicators. After wind and solar grid connection, wind power and photovoltaic power directly act on the system power balance constraint, change the output distribution of thermal power units, and thus affect the system power generation cost and pollutant gas emissions, but it is also necessary to consider the adverse effects of the volatility of photovoltaic and wind power output.

[0178] Figure 8 The system power generation cost and pollutant gas emissions at different wind and solar grid connection scales based on the MTSA algorithm are shown. Figure 8 It can be seen that compared with the dynamic dispatch model without renewable energy generation, the system power generation cost and pollutant gas emissions in each dispatch period are lower when renewable energy generation is included. The reason for the above results is that photovoltaic power stations and wind farms do not emit any pollutants during the power generation process. In addition, if the fixed investment of photovoltaic power stations and wind farms is ignored, only the lower operation and maintenance costs need to be considered, which greatly improves the economic and environmental benefits of the system after wind and solar power are connected to the grid.

[0179] Table 6 shows the system power generation costs and pollutant gas emissions obtained when each algorithm solves three dynamic scheduling models in Example 3, where model A is a dynamic scheduling model without wind and solar grid connection, model B is a dynamic scheduling model with wind and solar grid connection, and model C is a dynamic scheduling model that takes into account the penalty mechanism for renewable energy power generation.

[0180] Table 6. Solution results of three dynamic scheduling models

[0181]

[0182] As shown in Table 6, the solution results of Model B and Model C obtained by each algorithm are better than those of Model A, indicating that wind and solar grid connection has a positive effect on improving the economic and environmental benefits of the system, further verifying the conclusions obtained in the previous article. The solution results of Model C obtained by each algorithm are worse than those of Model B. This is because Model C imposes corresponding penalties on the overestimation or underestimation of renewable energy power generation from the aspects of system economy and environmental protection, resulting in an increase in system pollutant gas emissions and power generation costs. Taking into account the penalty mechanism for renewable energy power generation, the system will not increase additional costs due to the absorption of too much renewable energy power generation grid connection power, nor will it cause resource waste due to the absorption of too little renewable energy power generation power by the power grid, thereby improving the scientificity and operability of the scheduling plan formulated by decision makers.

[0183] Table 7 shows the calculation results of the system’s penalty cost and penalty emissions after taking into account the penalty mechanism for renewable energy generation.

[0184] Table 7. Calculation results of penalty costs and penalty emissions of the system after the introduction of the penalty mechanism

[0185]

[0186] As shown in Table 7, the penalty costs of overestimating or underestimating the output of photovoltaic and wind power are higher than the operation and maintenance costs of wind farms and photovoltaic power stations, and the penalty costs account for about 2.1% of the total costs. Considering that the power system is a relatively large system, the penalty costs caused by the forecast deviation of renewable energy power generation cannot be ignored. Although photovoltaic power stations and wind farms do not produce polluting gas emissions when generating electricity, the penalty emissions caused by the forecast deviation of photovoltaic and wind power output have a direct impact on the environmental benefits of the system, which cannot be ignored.

[0187] Fig. 9 The system power generation cost and pollutant gas emissions at each scheduling time obtained by the MTSA algorithm in Example 3 are shown. Fig. 9 It can be seen that the system power generation cost and pollutant gas emissions of model A are higher than those of model B and model C. When the new energy power generation penalty mechanism is introduced, the system economic and environmental protection indicators deteriorate. This is because the penalty costs generated by the volatility of photovoltaic and wind power output have a negative impact on the system economic and environmental benefits, but at the same time it also improves the feasibility of the decision-making plan.

[0188] For the HDEED problem, the system pollutant gas emissions and power generation costs obtained by the MTSA algorithm in three different cases are optimal, indicating that for a dynamic scheduling model with strong coupling and strong nonlinear characteristics that takes into account the penalty mechanism for renewable energy power generation, the scheduling scheme obtained by the MTSA algorithm takes into account both system economy and environmental protection.

[0189] Although the HDEED model constructed by the present invention has good feasibility and scalability, it still has the following limitations. First, the present invention only considers two new energy sources when constructing the HDEED model, and does not consider the grid connection of other clean power sources such as hydropower; second, the convergence stability and convergence stability of the MTSA algorithm need to be further improved to cope with more complex HDEED models.

[0190] It should be emphasized that the embodiments described in the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific embodiments. Any other embodiments derived by those skilled in the art based on the technical solution of the present invention also fall within the scope of protection of the present invention.

[0191] Any matters not described in the present invention are applicable to the prior art.

Claims

1. An optimization scheduling method based on artificial intelligence algorithm and renewable energy power generation penalty mechanism, comprising the following steps: Step 1: Construct a HDEED model with a penalty mechanism for renewable energy generation, and design the objective function and constraints of the HDEED model. Step 2: Based on the IEEE-10 unit 39-node thermal power system, a dynamic scheduling test system including photovoltaic or wind power is established to obtain the load demand and photovoltaic and wind farm output at each scheduling time in the dynamic scheduling test system. Step 3: Use the MTSA algorithm to solve the constructed dynamic scheduling model and obtain the Pareto optimal frontier solution of the dynamic scheduling model. Step 4: Select the best compromise solution in the Pareto optimal frontier based on fuzzy theory, and check whether the resulting scheduling scheme meets the system power balance constraint.

2. The optimization scheduling method according to claim 1, characterized in that: The process of the MTSA algorithm suitable for solving multi-objective optimization problems includes: In the process of constructing the optimization and scheduling model of the integrated energy system, the present invention not only considers the cost target of the system, but also considers the pollutant emission target of the system. The optimization and scheduling of the wind-solar-fire integrated energy system containing energy storage units is a strongly coupled, non-convex, highly nonlinear, multi-objective optimization problem, which has high requirements on the convergence performance of the solution algorithm. To this end, the present invention constructs a multi-objective quilt swarm algorithm (MTSA) based on the elite set guidance mechanism. The MTSA algorithm mainly includes three parts: the dominance relationship judgment criterion, the elite set guidance mechanism and the population generation. The specific principles are as follows: (1) Criteria for determining dominance The dominance relationship judgment criterion is the core of the MTSA algorithm. In the MTSA algorithm, due to the introduction of the concept of dominance level, the dominance relationship between two solutions cannot be judged by simply comparing the size of the objective function. It is necessary to combine the concepts of dominance level and crowding distance to realize the judgment of the dominance relationship between solutions. The dominance relationship judgment criterion between solutions in the MTSA algorithm is defined as Among them, R x rank represents the dominance level of the xth individual; P y rank represents the dominance level of the yth individual; C x represents the crowding distance of the xth individual; C y represents the crowding distance of the y-th individual. It can be seen from the above formula that in the MTSA algorithm, when judging the superiority relationship between two solutions, the solution with a smaller dominance level is better than the solution with a larger dominance level; if the dominance levels of the two solutions are the same, the crowding distances of the two solutions are compared, that is, the solution with a larger crowding distance is better than the solution with a smaller crowding distance. (2) Elite Collection Guidance Mechanism In the elite set guidance mechanism, the crowding distance of the individuals in the quilt group under different dominance levels is first calculated; then, the individual crowding distances are sorted and the top 5% individuals are selected as the elite set; finally, an individual is randomly selected from the elite set as the optimal individual, and the positions of other individuals in the quilt group are updated accordingly. The optimal solution selection method based on the elite set guidance mechanism is more reasonable, avoiding the optimal solution being limited to a single fixed solution, and improving the distribution of non-dominated solutions. The formula for calculating the individual crowding distance in the elite set guidance mechanism is: Among them, C x represents the crowding distance of the x-th solution; N p Indicates the population size; N o Indicates the number of targets; and Represent the target values ​​of the x+1th and x-1th individuals respectively; f y max and f y min Represent the maximum and minimum values ​​of the y-th target respectively. (3) Population generation The population strategy used by the MTSA algorithm is a dual population strategy that includes a parent population and a child population. Before the first iteration of the algorithm, the parent population is initialized and the dominance level of the individuals in the parent population is assigned (level 1 is the highest, level 2 is the second, and level 3 is the lowest). After the first iteration of the algorithm, the child population can be obtained. The parent population and the child population are merged to generate a scale of 2N. p Secondly, the merged population is non-dominated, the dominance level of the population is assigned, the crowding distance of the individuals in the population is calculated, and the next generation of the population is selected.

3. The prediction method according to claim 1, characterized in that: In order to improve the solving performance of the MTSA algorithm, the algorithm initialization strategy was improved and a random position perturbation strategy was introduced. The optimal dispatch of the power system under the grid connection of renewable energy is a multi-constrained, nonlinear, and non-convex multi-objective optimization problem. When solving such optimization problems, it is necessary not only to consider the power balance relationship between load, renewable energy, and thermal power, but also to deal with complex constraints, which places high requirements on the solving performance of the algorithm. To this end, this study introduced a population position initialization method based on a chaotic mapping strategy and a random position perturbation strategy into the MTSA algorithm. (1) Population position initialization method based on chaos mapping strategy In the process of solving the power system dispatching model, in order to obtain the initial population position with higher quality, speed up the convergence of the algorithm, and reduce the calculation cost of the algorithm, this section uses the randomness and ergodicity of the chaotic mapping strategy to obtain the initial population position: Among them, c s is a random number in the interval [0, 1]; Represents the maximum boundary of the search space; represents the minimum boundary of the search space; ρ represents the scale of the chaotic mapping; when ρ=4, the model is in a chaotic state. (2) Random Position Perturbation Strategy The positions of individuals in the tunicate group are dynamically disturbed through the random position perturbation strategy to ensure that the population maintains a high diversity throughout the entire solution process, thereby increasing the probability of the algorithm avoiding local extreme solutions and improving the solution effect of the scheduling model. The principle of the random position perturbation strategy is: during the iteration of the MTSA algorithm, the perturbation frequency f is set d By comparing f with a random number r in the interval [0, 1] d The value of and r determines whether to disturb the population position. Specifically, when f d ≤r t When f d >r t When , the individual positions in the quilt group are updated according to the formula. Among them, s c is the convergence coefficient, which is used to guide the individual convergence direction; s c min (s c min =0.3) is the minimum value of the convergence coefficient; s c max (s c max =0.7) is the maximum convergence coefficient; randn(0, σ 2 ) has a mean of 0 and a variance of σ 2 Gaussian distribution of T best represents the optimal individual position; f best represents the optimal individual fitness value; f T Represents the current individual fitness value.

4. The prediction method according to claim 1, characterized in that: The new energy power generation penalty mechanism systematically explores the impact of photovoltaic and wind power output forecast deviations on scheduling results from the two aspects of economy and environmental protection. In the process of establishing the objective function, specifically for the HDEED model, in addition to considering the fuel cost, it also includes the operation and maintenance costs of the photovoltaic power station and the wind farm, as well as the penalty cost caused by the prediction error. (1) Power generation cost objective function Among them, T dis is a scheduling cycle (24 hours in this paper); t is a scheduling period (1 hour in this paper); N f is the number of thermal power units; represents the minimum active output of the i-th thermal power unit at time t; a i 、b i 、c i 、e i 、f i They represent the fuel cost coefficient of the i-th thermal power unit respectively; represents the active power output of the i-th thermal power unit at time t; N pv is the number of photovoltaic power generation units; N w is the number of wind turbines; N soc is the number of energy storage units; represents the active output of the i-th photovoltaic power generation unit at time t; represents the active output of the i-th wind turbine at time t; represents the operation and maintenance cost coefficient of the i-th photovoltaic power generation unit at time t; represents the operation and maintenance cost coefficient of the i-th wind turbine at time t; The penalty cost function for underestimating the output of the i-th wind turbine; The reserve penalty cost function generated for overestimating the output of the i-th wind turbine; The penalty cost function for underestimating the output of the i-th photovoltaic power generation unit; The backup penalty cost function is generated for overestimating the output of the i-th PV unit. The mathematical model is as follows: in, is the operation and maintenance cost coefficient of the i-th wind turbine; e p，w Penalize cost coefficient for wind power output; is the rated output of the i-th wind turbine; f w (w) is the probability density function of wind power output; e r，w is the standby penalty cost coefficient; is the operation and maintenance cost coefficient of the ith photovoltaic power generation unit; e p，pv Penalize cost factor for PV output; is the rated output of the ith photovoltaic power generation unit; f pv (pv) is the probability density function of photovoltaic output; e r，pv The penalty cost coefficient for photovoltaic output reserve; is the charging power of the energy storage unit; is the discharge power of the energy storage unit. (2) Pollution gas emission objective function Among them, α i , β i , γ i , δ i , η i is the pollutant gas emission coefficient; is the penalty emission function caused by underestimating the output of the i-th wind turbine; is the reserve penalty emission function caused by overestimating the output of the i-th wind turbine; The penalty emission function generated for underestimating the output of the i-th photovoltaic power generation unit; Reserve penalty emission function generated for overestimating the output of the i-th PV unit. The mathematical model is Among them, d p，w Penalize emission factors for wind power output; is the rated output of the i-th wind turbine; f w (w) is the probability density function of wind power output; d r，w Penalize emission factor for wind power output reserve; d p，w Penalize emission factors for photovoltaic output; is the rated output of the ith photovoltaic power generation unit; f pv (pv) is the probability density function of photovoltaic output; d r，pv Penalize emission factors for photovoltaic output.

5. The specific steps of an optimization scheduling method based on artificial intelligence algorithm and renewable energy power generation penalty mechanism include: Construct a HDEED model with a penalty mechanism for renewable energy generation, and design the objective function and constraints of the HDEED model. In actual situations, due to the intermittent and random nature of photovoltaic and wind power output, there is a certain deviation between the model's predicted output and the actual output, and overestimation or underestimation of wind power and photovoltaic output will bring additional power generation costs and pollutant gas emissions, which will have a negative impact on the dispatch economy and environmental protection. Therefore, the present invention introduces a new energy power generation penalty mechanism into the HDEED model, and systematically explores the impact of photovoltaic and wind power output forecast deviation on the dispatch results from the two aspects of economy and environmental protection. 1) Objective function of HDEED model At present, countries around the world are setting strategic goals of "carbon peak" and "carbon neutrality". In this context, it is particularly necessary to improve the environmental protection of the system. Therefore, when constructing the HDEED model containing new energy power generation and energy storage units, the present invention considers the two indicators of system power generation cost and pollutant gas emissions in the objective function, and explores the dynamic scheduling scheme that optimizes the comprehensive economic and environmental benefits of the system. (1) Power generation cost objective function The power generation cost of the power system generally refers to the fuel cost consumed by thermal power units. For the HDEED model, in addition to the fuel cost, it also includes the operation and maintenance costs of photovoltaic power stations and wind farms, as well as the penalty costs caused by prediction errors. Among them, T dis is a scheduling cycle (24 hours in this paper); t is a scheduling period (1 hour in this paper); N f is the number of thermal power units; represents the minimum active output of the i-th thermal power unit at time t; a i 、b i 、c i 、e i 、f i They represent the fuel cost coefficient of the i-th thermal power unit respectively; represents the active power output of the i-th thermal power unit at time t; N pv is the number of photovoltaic power generation units; N w is the number of wind turbines; N soc is the number of energy storage units; represents the active output of the i-th photovoltaic power generation unit at time t; represents the active output of the i-th wind turbine at time t; represents the operation and maintenance cost coefficient of the i-th photovoltaic power generation unit at time t; represents the operation and maintenance cost coefficient of the i-th wind turbine at time t; The penalty cost function for underestimating the output of the i-th wind turbine; The reserve penalty cost function generated for overestimating the output of the i-th wind turbine; The penalty cost function for underestimating the output of the i-th photovoltaic power generation unit; The backup penalty cost function is generated for overestimating the output of the i-th PV unit. The mathematical model is as follows: in, is the operation and maintenance cost coefficient of the i-th wind turbine; e p，w Penalize cost coefficient for wind power output; is the rated output of the i-th wind turbine; f w (w) is the probability density function of wind power output; e r，w is the standby penalty cost coefficient; is the operation and maintenance cost coefficient of the ith photovoltaic power generation unit; e p，pv Penalize cost factor for PV output; is the rated output of the ith photovoltaic power generation unit; f pv (pv) is the probability density function of photovoltaic output; e r，pv The penalty cost coefficient for photovoltaic output reserve; is the charging power of the energy storage unit; is the discharge power of the energy storage unit. (2) Pollution gas emission objective function Among them, α i , β i , γ i , δ i , η i is the pollutant gas emission coefficient; is the penalty emission function caused by underestimating the output of the i-th wind turbine; is the reserve penalty emission function caused by overestimating the output of the i-th wind turbine; The penalty emission function generated for underestimating the output of the i-th photovoltaic power generation unit; Reserve penalty emission function generated for overestimating the output of the i-th PV unit. The mathematical model is Among them, d p，w Penalize emission factors for wind power output; is the rated output of the i-th wind turbine; f w (w) is the probability density function of wind power output; d r，w Penalize emission factor for wind power output reserve; d p，w Penalize emission factors for photovoltaic output; is the rated output of the ith photovoltaic power generation unit; f pv (pv) is the probability density function of photovoltaic output; d r，pv Penalize emission factors for photovoltaic output. 2) Constraints of the HDEED model In order to ensure the scientificity and feasibility of the obtained scheduling strategy, without significantly increasing the complexity of the model, this study focuses on the following constraints: (1) Power balance constraints Among them, P t load represents the system load demand at time t; P t loss represents the line loss at time t, and its expression is Among them, B * ij , B * i , B * is the line loss parameter. When solving the dynamic scheduling model, the loss parameter B * i With B * The impact on the system scheduling results is small. To simplify the model calculation, the above two parameters are ignored when solving the model. (2) Active power output constraints of thermal power units in, Indicates the lower limit of active output of the i-th thermal power unit at time t; It represents the upper limit of active output of the i-th thermal power unit at time t. (3) Active output constraints of wind farms and photovoltaic power stations in, and They are the maximum output of wind farm and photovoltaic power station respectively. (4) Climbing rate constraint The ramp rate constraint refers to the output that each thermal power unit can reduce or increase within a unit time. Among them, T * (T * =1) indicates the time interval of each period; are the ramp-down rate and ramp-up rate of the i-th thermal power unit respectively. (5) Positive and negative spinning reserve constraints Traditionally, spinning reserve constraints are used to deal with the adverse effects of unit shutdown and load forecast errors by setting a certain reserve capacity. However, for wind and solar grid-connected power systems, it is necessary not only to deal with the impact of load forecast errors, but also to deal with the impact of uncertainty in photovoltaic and wind power output. Positive and negative spinning reserve constraints are the most direct way to deal with wind and solar uncertainty. Positive spinning reserve constraints: in, represents the positive spinning reserve capacity provided by the i-th thermal power unit at time t; r l1 It represents the positive spinning reserve capacity factor of the thermal power unit in response to unit outage or load forecast error; r w1 It represents the positive spinning reserve capacity factor of thermal power units to cope with wind power output forecast error; r p1 It represents the positive spinning reserve capacity factor of thermal power units to cope with the photovoltaic output forecast error. Negative spinning reserve constraints: in, represents the negative spinning reserve capacity provided by the i-th thermal power unit at time t; r i2 It represents the negative spinning reserve capacity factor of the thermal power unit in response to unit outage or load forecast error; r w2 It represents the negative spinning reserve capacity factor of thermal power units to cope with the wind power output forecast error; r p2 It represents the negative spinning reserve capacity factor of thermal power units to cope with the photovoltaic output forecast error. (6) Charge and discharge power constraints of energy storage units Among them, P t char，max With P t dischar，max They are the maximum charging power and maximum discharging power of the energy storage unit respectively. In addition, the energy storage unit cannot be charged and discharged at the same time during operation, so it should also meet (7) Charge ratio constraints of energy storage units in, Q t soc，max is the upper limit of charge ratio; Q t soc，min is the lower limit of charge ratio; Q t soc is the charge ratio of the energy storage unit at time t, calculated as follows: Among them, Q t r is the remaining power of the energy storage unit; Q t max It is the maximum storage capacity of the energy storage unit. In addition, in order to ensure the sustainability of charging and discharging when the energy storage unit participates in scheduling, the sum of the charging power of the energy storage unit should be equal to the sum of the discharging power within a scheduling cycle, that is, 3) Conceptual scheme of HDEED multi-objective optimization The HDEED model constructed above belongs to a multi-objective minimization problem. Its purpose is to explore the optimal load distribution of the generator set based on the unit operation characteristic equation under given load levels and wind and solar grid conditions, and to seek a static dispatching scheme that optimizes the economic and environmental benefits of the system. Based on the analysis of 1) and 2), the HDEED mathematical model description formula taking into account the penalty mechanism for renewable energy generation is as follows, and its specific mathematical model is shown in Table A1. Among them, D cost With O emission They are the power generation cost target and pollutant gas emission target of the dynamic scheduling model taking into account the penalty mechanism for renewable energy power generation; E and I refer to the equality constraint and inequality constraint of the dynamic scheduling model respectively. Table A1. Relevant parameter settings of IEEE-10 unit test system (1) Based on the IEEE-10 unit 39-node thermal power system, a dynamic scheduling test system containing photovoltaic or wind power is established to obtain the load demand and the output of photovoltaic and wind farms at each scheduling time in the dynamic scheduling test system. Most of the current studies are based on the IEEE-10 unit 39-node thermal power system to establish a dynamic scheduling test system containing photovoltaic or wind power. To this end, the present invention is based on a photovoltaic power station and a wind farm in China, combined with the IEEE-10 unit thermal power system to establish a dynamic scheduling test system, and verify the constructed dynamic scheduling model and solution algorithm. The relevant parameter settings of the IEEE-10 unit 39-node thermal power system in the test system, the pollutant gas emission parameters and the transmission loss parameters are shown in Tables A2 and A3 below. Table A2. Pollutant gas emission parameters for IEEE-10 unit test system Table A3. Transmission loss parameters of IEEE-10 unit test system (2) The constructed dynamic scheduling model is solved by the MTSA algorithm to obtain the Pareto solution set of the dynamic scheduling model. The constructed dynamic scheduling model is solved by the MTSA algorithm. After the algorithm termination condition is reached, the individuals in the quilt group are the solutions in the Pareto solution set of the dynamic scheduling model. The objective function of the dynamic scheduling model is defined as the fitness function F of the MTSA algorithm, that is, F(p i )=[O cost (p i ),O emission (p i )]。

Images